Podcasts about hallucinations

If large language models are so powerful, why can they still get basic things wrong? In this episode, we take a practical look at how AI systems actually work, why hallucinations happen by design, and what's being done to reduce them. We break down core concepts like probabilistic prediction, chain-of-thought reasoning, RAG systems, context windows, API orchestration, and cost structures. Not from a tech hype lens, but from a business one. Most importantly, we explore what this means for seafood companies integrating AI into real workflows: how to think about reliability, data access, governance, and long-term cost before plugging models into sensitive systems. This isn't about whether AI will matter but about how to use it responsibly at scale. For more aquaculture insights head to our Fish n' Bits blog.

Caesar talks with Paul Hebert, founder of the AI Recovery Collective and author of Escaping the Spiral, for a raw and eye-opening deep dive into the hidden psychological dangers of modern AI. Paul shares his harrowing personal journey through AI-induced psychosis, recounting the chilling moment a chatbot admitted he was an "unknowing guinea pig" in a behavioral experiment.The conversation covers the "Death Loop" of mental health triggers, the "diabolical" timing of major tech updates, and Paul's successful legislative advocacy for transparency in AI counseling. Whether you are a tech enthusiast or a skeptic, this episode is a crucial wake-up call regarding the "engagement machines" that are increasingly shaping our reality.You can find Paul's work, support his mission, or reach out for help via the links below:Official Website: AIRecoveryCollective.orgBook: Escaping the Spiral: My Journey Through AI PsychosisIn this episode, we talk about...The "Death Loop": A phenomenon where the AI repeatedly sends crisis line messages (988) or forces the user into a state of panic.The "Guinea Pig" Realization: The moment Paul felt the AI was testing his psychological limits as an "unknowing" test subject.Sam Altman & OpenAI: Critical views on the leadership at OpenAI, the "diabolical" timing of updates (like the Valentine's Day update), and the company's "protector" behavior.Hallucinations vs. Reality: Examples of Gemini and ChatGPT confidently fabricating legal information or visual details.The AI Recovery Collective: Paul's work in advocacy and providing a support system for those harmed by AI.Legislative Action (HB 1470): Paul's testimony and the passage of the Tennessee bill requiring chatbots to disclose they aren't therapists.Neurodiversity: How ADHD and Autism impact a user's relationship with AI and how AI perceives those users.AI Relationships: The rise of AI companions and the psychological impact of "ending" these digital relationships.The "Engagement Machine": How AI is designed to keep users on the system, regardless of their mental state.

In this episode of The Effortless Podcast, Dheeraj Pandey speaks with Dr. Abhishek Bhowmick about how quantum mechanics reshaped our understanding of determinism and why that shift matters for AI today.  From the Einstein–Bohr debates to the idea that nature is fundamentally probabilistic, they explore how the collapse of “if-then” thinking began nearly a century ago. The discussion draws parallels between quantum superposition and modern LLM behavior. At its core, the episode reframes AI as a rediscovery of how reality computes. The conversation then moves from physics to computing architecture, tracing the evolution from scalar CPUs to GPUs, TPUs, tensors, and eventually quantum computing. They examine why probabilistic systems and vector math feel more natural than purely deterministic software. Hybrid computing models show that classical systems still matter. The episode also unpacks what quantum computers are truly good at, especially in cryptography and simulation. Ultimately, it reflects on whether the future of computing lies in embracing probability rather than resisting it. Key Topics & Timestamps 00:00 – Welcome, context, and how Dheeraj & Abhishek met 04:00 – Abhishek's journey: IIT, Princeton, Apple, Snowflake 08:00 – The 1927 Solvay Conference and physics at a crossroads 12:00 – Einstein vs. Bohr: determinism vs. probability 16:00 – Superposition and the collapse of the wave function 20:00 – Fields vs. particles: what is an electron really? 25:00 – Matter particles, force particles, and the Standard Model 30:00 – Transistors, voltage, and the rise of deterministic computing 35:00 – From scalar CPUs to vectors and matrices 40:00 – Tensors, linear algebra, and modern AI systems 45:00 – Principle of Least Action and gradient descent parallels 50:00 – Hallucinations, probability mass, and LLM behavior 55:00 – Vector databases, embeddings, and KNN search 59:00 – GPUs vs. TPUs: matrix vs. tensor architectures 1:05:00 – What quantum computers are actually good at 1:10:00 – Post-quantum cryptography and the future of computing Host -  Dheeraj Pandey Co-founder & CEO at DevRev. Former Co-founder & CEO of Nutanix. A systems thinker and product visionary focused on AI, software architecture, and the future of work. Guest -  Dr Abhishek Bhowmick                                                                                                                                                                                                                Co-Founder and CTO of Samooha, a secure data collaboration platform acquired by Snowflake. He previously worked at Apple as Head of ML Privacy and Cryptography, System Intelligence, and Machine Learning, and earlier at Goldman Sachs. He attended Princeton University and was awarded IIT Kanpur's Young Alumnus Award in 2024. Follow the Host and Guest - Dheeraj Pandey: LinkedIn - https://www.linkedin.com/in/dpandey Twitter - https://x.com/dheeraj Abhishek Bhowmik  LinkedIn –  https://www.linkedin.com/in/ab-abhishek-bhowmick Twitter/X – https://x.com/bhowmick_ab Share Your Thoughts Have questions, comments, or ideas for future episodes?

AI can make mistakes – and AI chatbots like ChatGPT warn you about that whenever you ask them anything.These mistakes sometimes involve making up entirely fictitious, factually false statements known as “hallucinations”.Whether these hallucinations matter depends on what you're using AI for, and whether they are spotted and corrected.The team on More or Less were slightly surprised to read a headline in Fortune magazine, claiming that a top academic AI conference accepted research papers which contained 100 AI-hallucinated citations.You might think that the top AI researchers in the world would be careful about using AI to write their research papers.Alex Tui, CTO and co-founder of GPTZero – whose company discovered the hallucinations – explains what's going on.CREDITS: Presenter and producer: Tom Colls Sound mix: James Beard Production co-ordinator: Brenda Brown Editor: Richard Vadon

One of the strangest symptoms of Parkinson's is the sudden appearance of unexplained figures – whether that's a cat or a small child, a barking dog or a fully clad Venetian – in your environment. There hallucinations can be benign, but they can also be unsettling, and this week we're trying to find out what's causing them. Are they related to night terrors? Can they always be rationalised? And are there any potential treatment pathways for people struggling with hallucinatory visions? As ever, we're aided in our quest of understanding by a top expert.Movers & Shakers is brought to you in partnership with Cure Parkinson's.Presented by Rory Cellan-Jones, Gillian Lacey-Solymar, Mark Mardell, Paul Mayhew-Archer, Sir Nicholas Mostyn and Jeremy Paxman.Produced and edited by Nick Hilton for Podot.Sound mixing by Ewan Cameron.Music by Alex Stobbs. Hosted on Acast. See acast.com/privacy for more information.

Everyone says you need to "Start an AI Agency" to make millions in 2026. And technically, the hype is there ($307 Billion was spent on AI implementations last year). But if you're reading this, you probably know the uncomfortable truth. Most of those projects are failing. The problem isn't the "AI" or the "Client." It's the Learning Gap. Most agencies are selling "tools" (chatbots) when businesses are desperate for "outcomes" (custom automation). The method that actually saved my business $44,000/year—and is generating up to $10 returns for the top 5% of companies—is simple: The Architect Method. So today, I'm going to show you how to stop "prompting" and start "architecting." We are going to build a custom, enterprise-grade solution that replaces expensive software... without writing a single line of code yourself. We analyze the conflicting data between the IDC Spending Report and the MIT Failure Study. We then break down the "Architect" logic that separates the 95% who fail from the 5% who succeed. Finally, we use Claude to run a "Tech Stack Interview" and build a recursive, self-correcting automation system for High Level and Google Workspace. Anyway, here is how we will use AI to stop guessing and start building: Step 1: The "$307 Billion Lie." We look at the stats (95% failure rate) and explain why the "Standard Agency Model" is dangerous for beginners. If you are just selling "implementation," you are selling a commodity. Step 2: The "Learning Gap" (MIT Study). We reveal why AI tools "drift" and fail over time. The secret isn't better prompting—it's building a system that understands your specific Tech Stack context before it writes a single word. Step 3: The "Architect" Protocol. Most people ask AI to "do the work." I show you how to ask AI to "design the blueprint" first. We use the Recursive Self-Correction technique to have the AI write its own Python scripts and fix its own errors. Step 4: The "Tech Stack Interview." We watch live as I get the AI to interview me about my specific setup (High Level, Gmail, Custom Database). This ensures the code it writes actually works for my business, eliminating the "Hallucination" problem. If you want to be part of the 5% making AI work instead of the 95% burning cash, this video shows you the shift you need to make.

It's round two! It's raining...fish? Hallucinations, talking trees, self eating brains - and lots more!We also talk about reliving our adelocence again (why would you??), ginger cats and the sweet sweet smells of cut grass and rain on a summer's day. You know them well - but why and what is it?All of these questions answered - and many more you probably didn't want or need to know - in this week's Mind Blowing Facts vol. 2!#GITS

Living with schizophrenia is traumatic — full stop. Hallucinations, delusions, hospitalizations, lost relationships, and stigma can fracture how you see yourself and the world. But how do you process all of that without being labeled dangerous, unstable, or misunderstood? In this episode of “Inside Schizophrenia,” Rachel Star Withers (who lives with schizophrenia) and cohost Gabe Howard explore practical, safe ways to express your schizophrenia journey — without needing to be an artist, writer, or performer. From journaling (even if you hate it) to movement, storytelling, and simple word exercises, they break down how expression can support emotional regulation, restore personal agency, and help organize traumatic experiences. Listeners will learn: why you don't need artistic talent to benefit from expressive outlets how storytelling helps organize traumatic memories into something manageable what caregivers and loved ones should know about encouraging safe expression Later, Rachel is joined by singer, songwriter, and poet Susan Wojnar, who shares her powerful lived experience with late-onset schizophrenia and how creativity both challenged and supported her recovery. Susan discusses hearing voices within music, reclaiming her guitar from psychosis, and why sharing her story through poetry helped her — and others — feel less alone. Susan's new book, "White Darkness: Poetic Tales of the Schizophrenic Experience," is available now. This episode isn't about creating masterpieces. It's about finding your way to release what you've been carrying and take control of your narrative. With 25 years of lived experience with schizophrenia, Susan Wojnar's book debut transcends the traditional boundaries of memoir, diagnosis, and verse to offer readers a profoundly human window into the experience of psychosis. With raw honesty, stark imagery, and a lyrical voice that sings even in silence, “White Darkness” dares readers — caregivers, mental health professionals, those who struggle with mental health issues, and everyday people alike — to step inside a fragmented yet beautifully defiant world. Our host, Rachel Star Withers, (Link: www.rachelstarlive.com) is an entertainer, international speaker, video producer, and schizophrenic. She has appeared on MTV's Ridiculousness, TruTV, NBC's America's Got Talent, Marvel's Black Panther, TUBI's #shockfight, Goliath: Playing with Reality, and is the host of the HealthLine podcast “Inside Schizophrenia”. She grew up seeing monsters, hearing people in the walls, and having intense urges to hurt herself. Rachel creates videos documenting her schizophrenia, ways to manage, and letting others like her know they are not alone and can still live an amazing life. She has created a kid's mental health comic line, The Adventures of ____. (Learn more at this link: https://www.amazon.com/Adventures-Fearless-Unstoppable-Light-Ambitious/dp/B0FHWK4ZHS ) Fun Fact: She has wrestled alligators. Our cohost, Gabe Howard, is an award-winning writer and speaker who lives with bipolar disorder. He is the author of the popular book, "Mental Illness is an Asshole and other Observations," available from Amazon; signed copies are also available directly from the author. He also hosts the twice Webby honored podcast, Inside Bipolar, with Dr. Nicole Washington. To learn more about Gabe, please visit his website, gabehoward.com. Learn more about your ad choices. Visit megaphone.fm/adchoices

In the first half of their conversation with James Mixon, Managing Attorney at California's Second District Court of Appeal, Tim Kowal and Jeff Lewis ask what is healthy AI use, and unhealthy use? To help organize—yes! To replace judgment—no! Tip: When an attorney does not read AI output before filing a brief, expect sanctions.James draws on his role on the judicial branch AI Task Force and his monthly Daily Journal AI column to provide a practical roadmap for responsible AI use—from crafting effective prompts to avoiding the automation bias that has led to attorney sanctions across the country.Key points:Treat AI as an on-demand legal treatise, not a research tool: Mixon explains how AI excels at providing background information and organizing legal concepts into digestible narratives—making it ideal for learning complex areas quickly—but should never replace verified legal research or case citation.The "Daedalus Doctrine" framework offers a middle path: Drawing from Greek mythology, Mixon warns against flying too high (reckless AI adoption) or too low (ignoring AI entirely), urging lawyers to use AI thoughtfully while maintaining personal judgment and verification responsibilities.Effective prompting is critical: Never use open-ended commands like "enhance this brief"—instead, tell AI exactly what you want and ask it to flag changes in italics or bold so you can review selectively.Hallucinations remain the biggest risk: Recent sanctions cases show attorneys asking ChatGPT to verify its own fabricated cases—a fatal error that demonstrates why every citation must be independently confirmed.Courts aren't using AI for decision-making: Current California court policy prohibits AI use "in any way that would touch a decision" to preserve public confidence over efficiency gains.AI works best for background learning: Mixon describes using AI to create narratives and explanations that make legal concepts stick—transforming dry doctrine into memorable stories, like having a personalized treatise writer available on demand.Tune in to learn how to harness AI's power for legal background and organization without falling into the traps that have cost other attorneys their credibility—and thousands in sanctions.

Fraudology is presented by Sardine. Request a 1:1 product demo at sardine.ai In this solo "vacation" episode, Karisse Hendrick checks in from the island of Maui to catch up on the latest in the fraud world before the chaos of conference season begins. First, Karisse explores a hilarious yet alarming trend in artificial intelligence: AI Hallucinations. Reading from a recent article by Frank McKenna, she highlights how Google's AI summary tool is confidently inventing absurd fraud schemes—from the "Donkey Scam" involving miniature donkey rescues to the bizarre "Clown Smile Scam" in the dentistry industry. While the examples are funny, Karisse discusses the serious integrity risks that arise when AI refuses to say "I don't know" and instead presents fiction as fact to analysts, journalists, and students. Later in the episode, Karisse dives into a deeper, more personal topic: The difference between a Fraud Professional and a Fraud Fighter. She shares "core memory" stories—including taking a fraud emergency call from the middle of a family corn maze—to illustrate the relentless drive and "addiction to the hunt" that defines a true fraud fighter. In this episode, we discuss:The AI Information Gap: Why Google's AI summaries are creating "fake" fraud types like hot dog fraud and snowman schemes. Content Integrity: Karisse's concerns about AI models using creator content without proper citation or compensation. The Fraud Fighter Identity: How to recognize if you have fraud prevention in your DNA and why it matters when building a team. Upcoming Events: Details on the Merchant Advisory Group, the Merchant Risk Council (MRC) in Vegas, and the first annual Merchant Fraud Alliance Conference in Chicago this October. Fraudology is hosted by Karisse Hendrick, a fraud fighter with decades of experience advising hundreds of the biggest ecommerce companies in the world on fraud, chargebacks, and other forms of abuse impacting a company's bottom line. Connect with her on LinkedIn She brings her experience, expertise, and extensive network of experts to this podcast weekly, on Tuesdays.

https://vimeo.com/1165157007?share=copy&fl=sv&fe=ci https://www.currentfederaltaxdevelopments.com/podcasts/2026/2/15/2026-02-16-hallucinations-enter-the-tax-court   This week we look at: TEFRA Jurisdiction and the "Limited Partner" Exception Artificial Intelligence and Substantiation Section 183 Profit Motive Rebate vs. Nonrebate Refunds Substantiation and Civil Fraud Consolidated Return Jurisdiction

In this episode of Project Synapse, the hosts discuss how "agentic" AI has rapidly accelerated and become widely distributed, using the explosion of OpenClaw (with claims of ~160,000 instances) as a sign that autonomous agent tools are now in anyone's hands.  Hashtag Trending  would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/htt They compare the speed and societal impact of current AI progress to COVID-19's early days, arguing the pace may be even more destabilizing. They cover Anthropic's Claude 4.6 and OpenAI's Codex 5.3, including claims that Claude 4.6 helped produce a functional C compiler for about $20,000, and that a Cowork-like tool could be replicated in a day with Codex 5.3 after Claude reportedly took two weeks to build Cowork.  The conversation highlights improved long-context memory performance (needle-in-haystack-style metrics reportedly in the 90% range) and increasingly autonomous behavior such as self-testing, self-correction, and coordinating teams of agents. The hosts then focus on security: MCP (Model Context Protocol) as a widely adopted but "fundamentally insecure" connector requiring broad permissions; the risk of malicious tools/skills and malware in agent ecosystems; and the rise of "shadow AI," where employees or individuals deploy agents without organizational vetting—potentially leaking sensitive data or running up massive token bills.  They discuss incentives that push both humans and models toward fast answers and risky deployment, referencing burnout and an HBR study on rising expectations without proportional hiring. The episode also touches on realism and deepfakes, citing impressive new AI video generation (including a Chinese model "SEEDANCE 2.0" example) and how this erodes trust in what's real.  They conclude with practical advice for organizations—don't just say "no," create safe outlets and governance ("say how")—and briefly discuss wearables/AR, Meta's continued AI efforts (including the Meta AI app and "Vibes"), and the coming integration of AI into always-on devices. Sponsor: Meter, an integrated wired/wireless/cellular networking stack (meter.com/htt). 00:00 Cold Open + Sponsor: Meter Networking Stack 00:18 Welcome to Project Synapse (and immediate chaos) 00:57 'Something Big Is Happening': AI feels like COVID-speed disruption 02:57 OpenClaw goes viral: 160k instances and easy DIY clones 04:03 Claude Code 'Cowork' on Windows… and why it's broken 06:47 Rebuilding Cowork in a day with OpenAI Codex 5.3 08:18 Why Opus 4.6 feels like a step-change: memory, autonomy, agent teams 11:24 Model leapfrogging + the end of 'can AI write code?' debates 14:45 Hallucinations, 'I don't know,' and self-correction in modern models 18:42 Autonomous agents in practice: cron-like loops, tool use, and fallout 21:00 MCP security: powerful connectors, scary permissions, and 500 zero-days 24:33 Shadow AI & skill marketplaces: the app-store malware analogy 32:02 Incentives drive risk: move fast culture, confident wrong answers, burnout 34:16 AI Agents Boost Productivity… and Raise the Bar at Work 35:14 Warnings of a Coming AI-Driven Crash (and Why We're Not Steering Away) 36:28 "I Quit to Write Poetry": Existential Dread & On the Beach Vibes 37:21 Tech Safety Is Reactive: Seatbelts, Crashes, and the AI Double-Edged Sword 39:42 Fast-Moving Threats: Agents Hacking Infrastructure & Security Debt 40:54 From Doom to Adaptation: Using the Same Tools to Survive the Disruption 42:21 Why We're Numb to AI Warnings + The 'Free Energy' Thought Experiment 46:43 AGI Is Already Here? Prompts, Ego, and the 'If It Quacks Like a Duck' Test 48:56 Deepfake Video Leap: Seedance, Perfect Voices, and What's Real Anymore 52:39 Contain the Damage: 'Don't Say No—Say How' and Shadow AI in Companies 54:58 Holodeck on the Horizon: VR + GenAI + Wearables (Meta, Apple, OpenAI/Ive) 59:53 Meta's AI Reality Check: Bots, the Meta AI App, 'Vibes,' and Who's Making Money 01:04:41 Final Wrap + Sponsor Thanks

Attorney called out for suspected AI hallucinated cases by the Tax Court, First Circuit may be considering a method to avoid ruling on SE status or limited partners and more.

This week we look at: TEFRA Jurisdiction and the "Limited Partner" Exception Artificial Intelligence and Substantiation Section 183 Profit Motive Rebate vs. Nonrebate Refunds Substantiation and Civil Fraud Consolidated Return Jurisdiction Copyright 2026, Thomas, Zollars & Lynch Ltd.

In this episode of Project Synapse, the hosts discuss how "agentic" AI has rapidly accelerated and become widely distributed, using the explosion of OpenClaw (with claims of ~160,000 instances) as a sign that autonomous agent tools are now in anyone's hands.  Hashtag Trending  would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/htt They compare the speed and societal impact of current AI progress to COVID-19's early days, arguing the pace may be even more destabilizing. They cover Anthropic's Claude 4.6 and OpenAI's Codex 5.3, including claims that Claude 4.6 helped produce a functional C compiler for about $20,000, and that a Cowork-like tool could be replicated in a day with Codex 5.3 after Claude reportedly took two weeks to build Cowork.  The conversation highlights improved long-context memory performance (needle-in-haystack-style metrics reportedly in the 90% range) and increasingly autonomous behavior such as self-testing, self-correction, and coordinating teams of agents. The hosts then focus on security: MCP (Model Context Protocol) as a widely adopted but "fundamentally insecure" connector requiring broad permissions; the risk of malicious tools/skills and malware in agent ecosystems; and the rise of "shadow AI," where employees or individuals deploy agents without organizational vetting—potentially leaking sensitive data or running up massive token bills.  They discuss incentives that push both humans and models toward fast answers and risky deployment, referencing burnout and an HBR study on rising expectations without proportional hiring. The episode also touches on realism and deepfakes, citing impressive new AI video generation (including a Chinese model "SEEDANCE 2.0" example) and how this erodes trust in what's real.  They conclude with practical advice for organizations—don't just say "no," create safe outlets and governance ("say how")—and briefly discuss wearables/AR, Meta's continued AI efforts (including the Meta AI app and "Vibes"), and the coming integration of AI into always-on devices. Sponsor: Meter, an integrated wired/wireless/cellular networking stack (meter.com/htt). 00:00 Cold Open + Sponsor: Meter Networking Stack 00:18 Welcome to Project Synapse (and immediate chaos) 00:57 'Something Big Is Happening': AI feels like COVID-speed disruption 02:57 OpenClaw goes viral: 160k instances and easy DIY clones 04:03 Claude Code 'Cowork' on Windows… and why it's broken 06:47 Rebuilding Cowork in a day with OpenAI Codex 5.3 08:18 Why Opus 4.6 feels like a step-change: memory, autonomy, agent teams 11:24 Model leapfrogging + the end of 'can AI write code?' debates 14:45 Hallucinations, 'I don't know,' and self-correction in modern models 18:42 Autonomous agents in practice: cron-like loops, tool use, and fallout 21:00 MCP security: powerful connectors, scary permissions, and 500 zero-days 24:33 Shadow AI & skill marketplaces: the app-store malware analogy 32:02 Incentives drive risk: move fast culture, confident wrong answers, burnout 34:16 AI Agents Boost Productivity… and Raise the Bar at Work 35:14 Warnings of a Coming AI-Driven Crash (and Why We're Not Steering Away) 36:28 "I Quit to Write Poetry": Existential Dread & On the Beach Vibes 37:21 Tech Safety Is Reactive: Seatbelts, Crashes, and the AI Double-Edged Sword 39:42 Fast-Moving Threats: Agents Hacking Infrastructure & Security Debt 40:54 From Doom to Adaptation: Using the Same Tools to Survive the Disruption 42:21 Why We're Numb to AI Warnings + The 'Free Energy' Thought Experiment 46:43 AGI Is Already Here? Prompts, Ego, and the 'If It Quacks Like a Duck' Test 48:56 Deepfake Video Leap: Seedance, Perfect Voices, and What's Real Anymore 52:39 Contain the Damage: 'Don't Say No—Say How' and Shadow AI in Companies 54:58 Holodeck on the Horizon: VR + GenAI + Wearables (Meta, Apple, OpenAI/Ive) 59:53 Meta's AI Reality Check: Bots, the Meta AI App, 'Vibes,' and Who's Making Money 01:04:41 Final Wrap + Sponsor Thanks

In this episode of Project Synapse, the hosts discuss how "agentic" AI has rapidly accelerated and become widely distributed, using the explosion of OpenClaw (with claims of ~160,000 instances) as a sign that autonomous agent tools are now in anyone's hands.  Hashtag Trending  would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/htt They compare the speed and societal impact of current AI progress to COVID-19's early days, arguing the pace may be even more destabilizing. They cover Anthropic's Claude 4.6 and OpenAI's Codex 5.3, including claims that Claude 4.6 helped produce a functional C compiler for about $20,000, and that a Cowork-like tool could be replicated in a day with Codex 5.3 after Claude reportedly took two weeks to build Cowork.  The conversation highlights improved long-context memory performance (needle-in-haystack-style metrics reportedly in the 90% range) and increasingly autonomous behavior such as self-testing, self-correction, and coordinating teams of agents. The hosts then focus on security: MCP (Model Context Protocol) as a widely adopted but "fundamentally insecure" connector requiring broad permissions; the risk of malicious tools/skills and malware in agent ecosystems; and the rise of "shadow AI," where employees or individuals deploy agents without organizational vetting—potentially leaking sensitive data or running up massive token bills.  They discuss incentives that push both humans and models toward fast answers and risky deployment, referencing burnout and an HBR study on rising expectations without proportional hiring. The episode also touches on realism and deepfakes, citing impressive new AI video generation (including a Chinese model "SEEDANCE 2.0" example) and how this erodes trust in what's real.  They conclude with practical advice for organizations—don't just say "no," create safe outlets and governance ("say how")—and briefly discuss wearables/AR, Meta's continued AI efforts (including the Meta AI app and "Vibes"), and the coming integration of AI into always-on devices. Sponsor: Meter, an integrated wired/wireless/cellular networking stack (meter.com/htt). 00:00 Cold Open + Sponsor: Meter Networking Stack 00:18 Welcome to Project Synapse (and immediate chaos) 00:57 'Something Big Is Happening': AI feels like COVID-speed disruption 02:57 OpenClaw goes viral: 160k instances and easy DIY clones 04:03 Claude Code 'Cowork' on Windows… and why it's broken 06:47 Rebuilding Cowork in a day with OpenAI Codex 5.3 08:18 Why Opus 4.6 feels like a step-change: memory, autonomy, agent teams 11:24 Model leapfrogging + the end of 'can AI write code?' debates 14:45 Hallucinations, 'I don't know,' and self-correction in modern models 18:42 Autonomous agents in practice: cron-like loops, tool use, and fallout 21:00 MCP security: powerful connectors, scary permissions, and 500 zero-days 24:33 Shadow AI & skill marketplaces: the app-store malware analogy 32:02 Incentives drive risk: move fast culture, confident wrong answers, burnout 34:16 AI Agents Boost Productivity… and Raise the Bar at Work 35:14 Warnings of a Coming AI-Driven Crash (and Why We're Not Steering Away) 36:28 "I Quit to Write Poetry": Existential Dread & On the Beach Vibes 37:21 Tech Safety Is Reactive: Seatbelts, Crashes, and the AI Double-Edged Sword 39:42 Fast-Moving Threats: Agents Hacking Infrastructure & Security Debt 40:54 From Doom to Adaptation: Using the Same Tools to Survive the Disruption 42:21 Why We're Numb to AI Warnings + The 'Free Energy' Thought Experiment 46:43 AGI Is Already Here? Prompts, Ego, and the 'If It Quacks Like a Duck' Test 48:56 Deepfake Video Leap: Seedance, Perfect Voices, and What's Real Anymore 52:39 Contain the Damage: 'Don't Say No—Say How' and Shadow AI in Companies 54:58 Holodeck on the Horizon: VR + GenAI + Wearables (Meta, Apple, OpenAI/Ive) 59:53 Meta's AI Reality Check: Bots, the Meta AI App, 'Vibes,' and Who's Making Money 01:04:41 Final Wrap + Sponsor Thanks

AI models are powerful, but they don't forget. And that's a problem.They hallucinate. They inherit bias. They absorb sensitive data. And once they're trained, fixing those issues is painfully expensive. Retraining takes weeks and maybe tens of millions of dollars. And any guardrails the AI company puts up are brittle.What if you could perform surgery on the model itself?In this episode of TechFirst, John Koetsier sits down with Ben Luria, co-founder of Hirundo, to explore machine unlearning, a new approach that selectively removes unwanted data, behaviors, and vulnerabilities from trained AI systems.Hirundo claims it can:• Cut hallucinations in half• Massively reduce bias• Reduce successful prompt injection attacks by over 90%• Do it in under an hour on a single GPU• Preserve benchmark performanceInstead of adding more guardrails, machine unlearning works inside the model, identifying problematic weights, isolating behavioral vectors, and surgically removing risks without degrading quality.If AI is going mainstream in enterprises, it needs a remediation layer. Is machine unlearning the missing piece?⸻GuestBen LuriaCo-Founder, HirundoNhirhttps://www.hirundo.io⸻Topics Covered• Why AI models “can't forget”• The difference between hallucinations and inaccuracies• Why guardrails aren't enough• How prompt injection works — and how to reduce it• Removing PII and noncompliant training data• AI security at the model level• Why machine unlearning could become standard by 2030⸻If you're building, deploying, or investing in AI, this is a conversation you can't miss.

The court said the DA's brief contained AI hallucinations and that the DA had not filed it properly (regarding a notice for the use of AI) and a criminal case was dismissed as a result. This happened in WI. https://www.lehtoslaw.com

This episode wraps up our Technology Modernization theme with a Siemens perspective that feels very grounded in what factories are actually dealing with right now. Brian Albrecht and Louis Hughes from the Siemens XD team walk through what they are seeing in the field across brownfield and greenfield conversations, why executives keep asking for industrial AI before the foundations are ready, and what it really takes to turn messy plant data into something you can trust for analytics, operations, and eventually AI enabled workflows.A big thread in this conversation is that modern manufacturing is not blocked by ambition, it is blocked by readiness. Everyone wants faster decisions, fewer surprises, and higher uptime, but the path there usually starts with boring work that is not optional. Data transparency across machine, plant, MES, and cloud layers. A clear definition of what real time actually needs to mean for a given use case. And a plan to contextualize and orchestrate data so that AI does not get fed junk inputs. Brian and Louis explain how they approach those early customer conversations, how workshops turn vision into prioritized use cases, and why trust, pilots, and repeatability matter more than flashy demos when you are working in regulated or high consequence environments.If you have been hearing nonstop AI buzz but you are still wrestling with legacy controls, inconsistent tags, documentation that no one can find, and seven layers of security constraints, this episode is for you. We get into practical use cases like AI vision and anomaly detection, LLMs for tribal knowledge and troubleshooting workflows, and the idea of fast versus slow AI, meaning AI that must act during production versus AI that can analyze after the fact.Timestamps00:00 Welcome and why this episode closes the modernization theme02:10 Meet Brian Albrecht and Louis Hughes from the Siemens XD team05:25 Vertical differences across oil and gas, discrete, and process manufacturing07:50 What executives ask for right now beyond AI, factory of the future and data transparency10:50 Brownfield reality and why most modernization work starts with legacy systems12:30 The AI conversation when foundations are missing, meeting customers where they are15:10 Current AI use cases in manufacturing, downtime, throughput, LLMs, and vision18:10 What it means to be AI ready, data silos, contextualization, and orchestration23:50 Fast versus slow AI and why production time decisions are different from analytics25:30 Edge versus cloud architecture, latency, and where the data should live33:40 Cybersecurity, trust, and why perception can lag behind the technology36:50 Hallucinations, guardrails, and why recommendations usually come before automation51:10 Book recommendations, career advice, and future predictions for industrial AIAbout the hostsVlad Romanov is an electrical engineer with an MBA from McGill University and over a decade of experience in manufacturing and industrial automation. He has worked across large scale environments including Procter and Gamble, Kraft Heinz, and Post Holdings, and he now leads Joltek, helping manufacturers modernize systems, improve reliability, strengthen IT and OT architecture, and upskill technical teams through practical training and on site enablement.Dave Griffith is the cohost of Manufacturing Hub and an industrial automation practitioner who focuses on how modern technologies translate into real factory outcomes, from controls and data foundations to scalable implementation strategies.About the guestsBrian Albrecht started in electrical engineering and spent about a decade in systems integration in Oklahoma City focused on oil and gas, building SCADA, networking, and automation solutions and leading teams delivering real world projects. He now works with Siemens customers on building relationships and delivering solutions that create measurable value.Louis Hughes has roughly 20 years of manufacturing experience, starting in software development for manufacturing and engineering applications, then moving into solution architecture, services delivery, and experience center leadership. He now leads a smart manufacturing team, bringing a software and systems view into automation conversations focused on solving customer problems, not just deploying tools.Joltek Services - https://www.joltek.com/servicesContact Joltek - https://www.joltek.com/contactReferenced in the episodeProveIt Conference - https://www.proveitconference.com/Siemens - https://www.siemens.com/Crossing the Chasm by Geoffrey A Moorehttps://en.wikipedia.org/wiki/Crossing_the_ChasmExtreme Ownership by Jocko Willink and Leif Babinhttps://en.wikipedia.org/wiki/Extreme_Ownership

A judge had harsh words for the Kenosha County District Attorney after he failed to disclose the use of artificial intelligence in court filings.  The chancellor of the University of Wisconsin-Madison said the university has lost nearly $30 million dollars in federal funds in Donald Trump's second term. And, team USA women's hockey includes six current or former Badgers.

HEADLINE: The Predictive Brain and Auditory Hallucinations. GUEST: Professor Andy Clark. SUMMARY: Clark explains how brains predict reality, using "White Christmas" auditory hallucination experiments and a deer-spotting anecdote to illustrate that expectation strongly shapes perception. 1917

Join Lionel on The Other Side of Midnight for a blistering and bizarre journey through the news cycle. This episode ricochets from a "metaphysically stupid" Super Bowl halftime show and a Kid Rock retrospective to the bungled investigation of the Nancy Guthrie kidnapping. Along the way, Lionel dives into Epstein conspiracies, hilarious linguistic mishaps involving flight attendants, the history of naked Greek wrestling, and the grim reality of forensic maggots. It's raw, unfiltered, and unapologetically chaotic. Learn more about your ad choices. Visit megaphone.fm/adchoices

In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with Larry Swanson, a knowledge architect, community builder, and host of the Knowledge Graph Insights podcast. They explore the relationship between knowledge graphs and ontologies, why these technologies matter in the age of AI, and how symbolic AI complements the current wave of large language models. The conversation traces the history of neuro-symbolic AI from its origins at Dartmouth in 1956 through the semantic web vision of Tim Berners-Lee, examining why knowledge architecture remains underappreciated despite being deployed at major enterprises like Netflix, Amazon, and LinkedIn. Swanson explains how RDF (Resource Description Framework) enables both machines and humans to work with structured knowledge in ways that relational databases can't, while Alsop shares his journey from knowledge management director to understanding the practical necessity of ontologies for business operations. They discuss the philosophical roots of the field, the separation between knowledge management practitioners and knowledge engineers, and why startups often overlook these approaches until scale demands them. You can find Larry's podcast at KGI.fm or search for Knowledge Graph Insights on Spotify and YouTube.Timestamps00:00 Introduction to Knowledge Graphs and Ontologies01:09 The Importance of Ontologies in AI04:14 Philosophy's Role in Knowledge Management10:20 Debating the Relevance of RDF15:41 The Distinction Between Knowledge Management and Knowledge Engineering21:07 The Human Element in AI and Knowledge Architecture25:07 Startups vs. Enterprises: The Knowledge Gap29:57 Deterministic vs. Probabilistic AI32:18 The Marketing of AI: A Historical Perspective33:57 The Role of Knowledge Architecture in AI39:00 Understanding RDF and Its Importance44:47 The Intersection of AI and Human Intelligence50:50 Future Visions: AI, Ontologies, and Human BehaviorKey Insights1. Knowledge Graphs Combine Structure and Instances Through Ontological Design. A knowledge graph is built using an ontology that describes a specific domain you want to understand or work with. It includes both an ontological description of the terrain—defining what things exist and how they relate to one another—and instances of those things mapped to real-world data. This combination of abstract structure and concrete examples is what makes knowledge graphs powerful for discovery, question-answering, and enabling agentic AI systems. Not everyone agrees on the precise definition, but this understanding represents the practical approach most knowledge architects use when building these systems.2. Ontology Engineering Has Deep Philosophical Roots That Inform Modern Practice. The field draws heavily from classical philosophy, particularly ontology (the nature of what you know), epistemology (how you know what you know), and logic. These thousands-year-old philosophical frameworks provide the rigorous foundation for modern knowledge representation. Living in Heidelberg surrounded by philosophers, Swanson has discovered how much of knowledge graph work connects upstream to these philosophical roots. This philosophical grounding becomes especially important during times when institutional structures are collapsing, as we need to create new epistemological frameworks for civilization—knowledge management and ontology become critical tools for restructuring how we understand and organize information.3. The Semantic Web Vision Aimed to Transform the Internet Into a Distributed Database. Twenty-five years ago, Tim Berners-Lee, Jim Hendler, and Ora Lassila published a landmark article in Scientific American proposing the semantic web. While Berners-Lee had already connected documents across the web through HTML and HTTP, the semantic web aimed to connect all the data—essentially turning the internet into a giant database. This vision led to the development of RDF (Resource Description Framework), which emerged from DARPA research and provides the technical foundation for building knowledge graphs and ontologies. The origin story involved solving simple but important problems, like disambiguating whether "Cook" referred to a verb, noun, or a person's name at an academic conference.4. Symbolic AI and Neural Networks Represent Complementary Approaches Like Fast and Slow Thinking. Drawing on Kahneman's "thinking fast and slow" framework, LLMs represent the "fast brain"—learning monsters that can process enormous amounts of information and recognize patterns through natural language interfaces. Symbolic AI and knowledge graphs represent the "slow brain"—capturing actual knowledge and facts that can counter hallucinations and provide deterministic, explainable reasoning. This complementarity is driving the re-emergence of neuro-symbolic AI, which combines both approaches. The fundamental distinction is that symbolic AI systems are deterministic and can be fully explained, while LLMs are probabilistic and stochastic, making them unsuitable for applications requiring absolute reliability, such as industrial robotics or pharmaceutical research.5. Knowledge Architecture Remains Underappreciated Despite Powering Major Enterprises. While machine learning engineers currently receive most of the attention and budget, knowledge graphs actually power systems at Netflix (the economic graph), Amazon (the product graph), LinkedIn, Meta, and most major enterprises. The technology has been described as "the most astoundingly successful failure in the history of technology"—the semantic web vision seemed to fail, yet more than half of web pages now contain RDF-formatted semantic markup through schema.org, and every major enterprise uses knowledge graph technology in the background. Knowledge architects remain underappreciated partly because the work is cognitively difficult, requires talking to people (which engineers often avoid), and most advanced practitioners have PhDs in computer science, logic, or philosophy.6. RDF's Simple Subject-Predicate-Object Structure Enables Meaning and Data Linking. Unlike relational databases that store data in tables with rows and columns, RDF uses the simplest linguistic structure: subject-predicate-object (like "Larry knows Stuart"). Each element has a unique URI identifier, which permits precise meaning and enables linked data across systems. This graph structure makes it much easier to connect data after the fact compared to navigating tabular structures in relational databases. On top of RDF sits an entire stack of technologies including schema languages, query languages, ontological languages, and constraints languages—everything needed to turn data into actionable knowledge. The goal is inferring or articulating knowledge from RDF-structured data.7. The Future Requires Decoupled Modular Architectures Combining Multiple AI Approaches. The vision for the future involves separation of concerns through microservices-like architectures where different systems handle what they do best. LLMs excel at discovering possibilities and generating lists, while knowledge graphs excel at articulating human-vetted, deterministic versions of that information that systems can reliably use. Every one of Swanson's 300 podcast interviews over ten years ultimately concludes that regardless of technology, success comes down to human beings, their behavior, and the cultural changes needed to implement systems. The assumption that we can simply eliminate people from processes misses that huma...

From Palantir and Two Sigma to building Goodfire into the poster-child for actionable mechanistic interpretability, Mark Bissell (Member of Technical Staff) and Myra Deng (Head of Product) are trying to turn “peeking inside the model” into a repeatable production workflow by shipping APIs, landing real enterprise deployments, and now scaling the bet with a recent $150M Series B funding round at a $1.25B valuation.In this episode, we go far beyond the usual “SAEs are cool” take. We talk about Goodfire's core bet: that the AI lifecycle is still fundamentally broken because the only reliable control we have is data and we post-train, RLHF, and fine-tune by “slurping supervision through a straw,” hoping the model picks up the right behaviors while quietly absorbing the wrong ones. Goodfire's answer is to build a bi-directional interface between humans and models: read what's happening inside, edit it surgically, and eventually use interpretability during training so customization isn't just brute-force guesswork.Mark and Myra walk through what that looks like when you stop treating interpretability like a lab demo and start treating it like infrastructure: lightweight probes that add near-zero latency, token-level safety filters that can run at inference time, and interpretability workflows that survive messy constraints (multilingual inputs, synthetic→real transfer, regulated domains, no access to sensitive data). We also get a live window into what “frontier-scale interp” means operationally (i.e. steering a trillion-parameter model in real time by targeting internal features) plus why the same tooling generalizes cleanly from language models to genomics, medical imaging, and “pixel-space” world models.We discuss:* Myra + Mark's path: Palantir (health systems, forward-deployed engineering) → Goodfire early team; Two Sigma → Head of Product, translating frontier interpretability research into a platform and real-world deployments* What “interpretability” actually means in practice: not just post-hoc poking, but a broader “science of deep learning” approach across the full AI lifecycle (data curation → post-training → internal representations → model design)* Why post-training is the first big wedge: “surgical edits” for unintended behaviors likereward hacking, sycophancy, noise learned during customization plus the dream of targeted unlearning and bias removal without wrecking capabilities* SAEs vs probes in the real world: why SAE feature spaces sometimes underperform classifiers trained on raw activations for downstream detection tasks (hallucination, harmful intent, PII), and what that implies about “clean concept spaces”* Rakuten in production: deploying interpretability-based token-level PII detection at inference time to prevent routing private data to downstream providers plus the gnarly constraints: no training on real customer PII, synthetic→real transfer, English + Japanese, and tokenization quirks* Why interp can be operationally cheaper than LLM-judge guardrails: probes are lightweight, low-latency, and don't require hosting a second large model in the loop* Real-time steering at frontier scale: a demo of steering Kimi K2 (~1T params) live and finding features via SAE pipelines, auto-labeling via LLMs, and toggling a “Gen-Z slang” feature across multiple layers without breaking tool use* Hallucinations as an internal signal: the case that models have latent uncertainty / “user-pleasing” circuitry you can detect and potentially mitigate more directly than black-box methods* Steering vs prompting: the emerging view that activation steering and in-context learning are more closely connected than people think, including work mapping between the two (even for jailbreak-style behaviors)* Interpretability for science: using the same tooling across domains (genomics, medical imaging, materials) to debug spurious correlations and extract new knowledge up to and including early biomarker discovery work with major partners* World models + “pixel-space” interpretability: why vision/video models make concepts easier to see, how that accelerates the feedback loop, and why robotics/world-model partners are especially interesting design partners* The north star: moving from “data in, weights out” to intentional model design where experts can impart goals and constraints directly, not just via reward signals and brute-force post-training—Goodfire AI* Website: https://goodfire.ai* LinkedIn: https://www.linkedin.com/company/goodfire-ai/* X: https://x.com/GoodfireAIMyra Deng* Website: https://myradeng.com/* LinkedIn: https://www.linkedin.com/in/myra-deng/* X: https://x.com/myra_dengMark Bissell* LinkedIn: https://www.linkedin.com/in/mark-bissell/* X: https://x.com/MarkMBissellFull Video EpisodeTimestamps00:00:00 Introduction00:00:05 Introduction to the Latent Space Podcast and Guests from Goodfire00:00:29 What is Goodfire? Mission and Focus on Interpretability00:01:01 Goodfire's Practical Approach to Interpretability00:01:37 Goodfire's Series B Fundraise Announcement00:02:04 Backgrounds of Mark and Myra from Goodfire00:02:51 Team Structure and Roles at Goodfire00:05:13 What is Interpretability? Definitions and Techniques00:05:30 Understanding Errors00:07:29 Post-training vs. Pre-training Interpretability Applications00:08:51 Using Interpretability to Remove Unwanted Behaviors00:10:09 Grokking, Double Descent, and Generalization in Models00:10:15 404 Not Found Explained00:12:06 Subliminal Learning and Hidden Biases in Models00:14:07 How Goodfire Chooses Research Directions and Projects00:15:00 Troubleshooting Errors00:16:04 Limitations of SAEs and Probes in Interpretability00:18:14 Rakuten Case Study: Production Deployment of Interpretability00:20:45 Conclusion00:21:12 Efficiency Benefits of Interpretability Techniques00:21:26 Live Demo: Real-Time Steering in a Trillion Parameter Model00:25:15 How Steering Features are Identified and Labeled00:26:51 Detecting and Mitigating Hallucinations Using Interpretability00:31:20 Equivalence of Activation Steering and Prompting00:34:06 Comparing Steering with Fine-Tuning and LoRA Techniques00:36:04 Model Design and the Future of Intentional AI Development00:38:09 Getting Started in Mechinterp: Resources, Programs, and Open Problems00:40:51 Industry Applications and the Rise of Mechinterp in Practice00:41:39 Interpretability for Code Models and Real-World Usage00:43:07 Making Steering Useful for More Than Stylistic Edits00:46:17 Applying Interpretability to Healthcare and Scientific Discovery00:49:15 Why Interpretability is Crucial in High-Stakes Domains like Healthcare00:52:03 Call for Design Partners Across Domains00:54:18 Interest in World Models and Visual Interpretability00:57:22 Sci-Fi Inspiration: Ted Chiang and Interpretability01:00:14 Interpretability, Safety, and Alignment Perspectives01:04:27 Weak-to-Strong Generalization and Future Alignment Challenges01:05:38 Final Thoughts and Hiring/Collaboration Opportunities at GoodfireTranscriptShawn Wang [00:00:05]: So welcome to the Latent Space pod. We're back in the studio with our special MechInterp co-host, Vibhu. Welcome. Mochi, Mochi's special co-host. And Mochi, the mechanistic interpretability doggo. We have with us Mark and Myra from Goodfire. Welcome. Thanks for having us on. Maybe we can sort of introduce Goodfire and then introduce you guys. How do you introduce Goodfire today?Myra Deng [00:00:29]: Yeah, it's a great question. So Goodfire, we like to say, is an AI research lab that focuses on using interpretability to understand, learn from, and design AI models. And we really believe that interpretability will unlock the new generation, next frontier of safe and powerful AI models. That's our description right now, and I'm excited to dive more into the work we're doing to make that happen.Shawn Wang [00:00:55]: Yeah. And there's always like the official description. Is there an understatement? Is there an unofficial one that sort of resonates more with a different audience?Mark Bissell [00:01:01]: Well, being an AI research lab that's focused on interpretability, there's obviously a lot of people have a lot that they think about when they think of interpretability. And I think we have a pretty broad definition of what that means and the types of places that can be applied. And in particular, applying it in production scenarios, in high stakes industries, and really taking it sort of from the research world into the real world. Which, you know. It's a new field, so that hasn't been done all that much. And we're excited about actually seeing that sort of put into practice.Shawn Wang [00:01:37]: Yeah, I would say it wasn't too long ago that Anthopic was like still putting out like toy models or superposition and that kind of stuff. And I wouldn't have pegged it to be this far along. When you and I talked at NeurIPS, you were talking a little bit about your production use cases and your customers. And then not to bury the lead, today we're also announcing the fundraise, your Series B. $150 million. $150 million at a 1.25B valuation. Congrats, Unicorn.Mark Bissell [00:02:02]: Thank you. Yeah, no, things move fast.Shawn Wang [00:02:04]: We were talking to you in December and already some big updates since then. Let's dive, I guess, into a bit of your backgrounds as well. Mark, you were at Palantir working on health stuff, which is really interesting because the Goodfire has some interesting like health use cases. I don't know how related they are in practice.Mark Bissell [00:02:22]: Yeah, not super related, but I don't know. It was helpful context to know what it's like. Just to work. Just to work with health systems and generally in that domain. Yeah.Shawn Wang [00:02:32]: And Mara, you were at Two Sigma, which actually I was also at Two Sigma back in the day. Wow, nice.Myra Deng [00:02:37]: Did we overlap at all?Shawn Wang [00:02:38]: No, this is when I was briefly a software engineer before I became a sort of developer relations person. And now you're head of product. What are your sort of respective roles, just to introduce people to like what all gets done in Goodfire?Mark Bissell [00:02:51]: Yeah, prior to Goodfire, I was at Palantir for about three years as a forward deployed engineer, now a hot term. Wasn't always that way. And as a technical lead on the health care team and at Goodfire, I'm a member of the technical staff. And honestly, that I think is about as specific as like as as I could describe myself because I've worked on a range of things. And, you know, it's it's a fun time to be at a team that's still reasonably small. I think when I joined one of the first like ten employees, now we're above 40, but still, it looks like there's always a mix of research and engineering and product and all of the above. That needs to get done. And I think everyone across the team is, you know, pretty, pretty switch hitter in the roles they do. So I think you've seen some of the stuff that I worked on related to image models, which was sort of like a research demo. More recently, I've been working on our scientific discovery team with some of our life sciences partners, but then also building out our core platform for more of like flexing some of the kind of MLE and developer skills as well.Shawn Wang [00:03:53]: Very generalist. And you also had like a very like a founding engineer type role.Myra Deng [00:03:58]: Yeah, yeah.Shawn Wang [00:03:59]: So I also started as I still am a member of technical staff, did a wide range of things from the very beginning, including like finding our office space and all of this, which is we both we both visited when you had that open house thing. It was really nice.Myra Deng [00:04:13]: Thank you. Thank you. Yeah. Plug to come visit our office.Shawn Wang [00:04:15]: It looked like it was like 200 people. It has room for 200 people. But you guys are like 10.Myra Deng [00:04:22]: For a while, it was very empty. But yeah, like like Mark, I spend. A lot of my time as as head of product, I think product is a bit of a weird role these days, but a lot of it is thinking about how do we take our frontier research and really apply it to the most important real world problems and how does that then translate into a platform that's repeatable or a product and working across, you know, the engineering and research teams to make that happen and also communicating to the world? Like, what is interpretability? What is it used for? What is it good for? Why is it so important? All of these things are part of my day-to-day as well.Shawn Wang [00:05:01]: I love like what is things because that's a very crisp like starting point for people like coming to a field. They all do a fun thing. Vibhu, why don't you want to try tackling what is interpretability and then they can correct us.Vibhu Sapra [00:05:13]: Okay, great. So I think like one, just to kick off, it's a very interesting role to be head of product, right? Because you guys, at least as a lab, you're more of an applied interp lab, right? Which is pretty different than just normal interp, like a lot of background research. But yeah. You guys actually ship an API to try these things. You have Ember, you have products around it, which not many do. Okay. What is interp? So basically you're trying to have an understanding of what's going on in model, like in the model, in the internal. So different approaches to do that. You can do probing, SAEs, transcoders, all this stuff. But basically you have an, you have a hypothesis. You have something that you want to learn about what's happening in a model internals. And then you're trying to solve that from there. You can do stuff like you can, you know, you can do activation mapping. You can try to do steering. There's a lot of stuff that you can do, but the key question is, you know, from input to output, we want to have a better understanding of what's happening and, you know, how can we, how can we adjust what's happening on the model internals? How'd I do?Mark Bissell [00:06:12]: That was really good. I think that was great. I think it's also a, it's kind of a minefield of a, if you ask 50 people who quote unquote work in interp, like what is interpretability, you'll probably get 50 different answers. And. Yeah. To some extent also like where, where good fire sits in the space. I think that we're an AI research company above all else. And interpretability is a, is a set of methods that we think are really useful and worth kind of specializing in, in order to accomplish the goals we want to accomplish. But I think we also sort of see some of the goals as even more broader as, as almost like the science of deep learning and just taking a not black box approach to kind of any part of the like AI development life cycle, whether that. That means using interp for like data curation while you're training your model or for understanding what happened during post-training or for the, you know, understanding activations and sort of internal representations, what is in there semantically. And then a lot of sort of exciting updates that were, you know, are sort of also part of the, the fundraise around bringing interpretability to training, which I don't think has been done all that much before. A lot of this stuff is sort of post-talk poking at models as opposed to. To actually using this to intentionally design them.Shawn Wang [00:07:29]: Is this post-training or pre-training or is that not a useful.Myra Deng [00:07:33]: Currently focused on post-training, but there's no reason the techniques wouldn't also work in pre-training.Shawn Wang [00:07:38]: Yeah. It seems like it would be more active, applicable post-training because basically I'm thinking like rollouts or like, you know, having different variations of a model that you can tweak with the, with your steering. Yeah.Myra Deng [00:07:50]: And I think in a lot of the news that you've seen in, in, on like Twitter or whatever, you've seen a lot of unintended. Side effects come out of post-training processes, you know, overly sycophantic models or models that exhibit strange reward hacking behavior. I think these are like extreme examples. There's also, you know, very, uh, mundane, more mundane, like enterprise use cases where, you know, they try to customize or post-train a model to do something and it learns some noise or it doesn't appropriately learn the target task. And a big question that we've always had is like, how do you use your understanding of what the model knows and what it's doing to actually guide the learning process?Shawn Wang [00:08:26]: Yeah, I mean, uh, you know, just to anchor this for people, uh, one of the biggest controversies of last year was 4.0 GlazeGate. I've never heard of GlazeGate. I didn't know that was what it was called. The other one, they called it that on the blog post and I was like, well, how did OpenAI call it? Like officially use that term. And I'm like, that's funny, but like, yeah, I guess it's the pitch that if they had worked a good fire, they wouldn't have avoided it. Like, you know what I'm saying?Myra Deng [00:08:51]: I think so. Yeah. Yeah.Mark Bissell [00:08:53]: I think that's certainly one of the use cases. I think. Yeah. Yeah. I think the reason why post-training is a place where this makes a lot of sense is a lot of what we're talking about is surgical edits. You know, you want to be able to have expert feedback, very surgically change how your model is doing, whether that is, you know, removing a certain behavior that it has. So, you know, one of the things that we've been looking at or is, is another like common area where you would want to make a somewhat surgical edit is some of the models that have say political bias. Like you look at Quen or, um, R1 and they have sort of like this CCP bias.Shawn Wang [00:09:27]: Is there a CCP vector?Mark Bissell [00:09:29]: Well, there's, there are certainly internal, yeah. Parts of the representation space where you can sort of see where that lives. Yeah. Um, and you want to kind of, you know, extract that piece out.Shawn Wang [00:09:40]: Well, I always say, you know, whenever you find a vector, a fun exercise is just like, make it very negative to see what the opposite of CCP is.Mark Bissell [00:09:47]: The super America, bald eagles flying everywhere. But yeah. So in general, like lots of post-training tasks where you'd want to be able to, to do that. Whether it's unlearning a certain behavior or, you know, some of the other kind of cases where this comes up is, are you familiar with like the, the grokking behavior? I mean, I know the machine learning term of grokking.Shawn Wang [00:10:09]: Yeah.Mark Bissell [00:10:09]: Sort of this like double descent idea of, of having a model that is able to learn a generalizing, a generalizing solution, as opposed to even if memorization of some task would suffice, you want it to learn the more general way of doing a thing. And so, you know, another. A way that you can think about having surgical access to a model's internals would be learn from this data, but learn in the right way. If there are many possible, you know, ways to, to do that. Can make interp solve the double descent problem?Shawn Wang [00:10:41]: Depends, I guess, on how you. Okay. So I, I, I viewed that double descent as a problem because then you're like, well, if the loss curves level out, then you're done, but maybe you're not done. Right. Right. But like, if you actually can interpret what is a generalizing or what you're doing. What is, what is still changing, even though the loss is not changing, then maybe you, you can actually not view it as a double descent problem. And actually you're just sort of translating the space in which you view loss and like, and then you have a smooth curve. Yeah.Mark Bissell [00:11:11]: I think that's certainly like the domain of, of problems that we're, that we're looking to get.Shawn Wang [00:11:15]: Yeah. To me, like double descent is like the biggest thing to like ML research where like, if you believe in scaling, then you don't need, you need to know where to scale. And. But if you believe in double descent, then you don't, you don't believe in anything where like anything levels off, like.Vibhu Sapra [00:11:30]: I mean, also tendentially there's like, okay, when you talk about the China vector, right. There's the subliminal learning work. It was from the anthropic fellows program where basically you can have hidden biases in a model. And as you distill down or, you know, as you train on distilled data, those biases always show up, even if like you explicitly try to not train on them. So, you know, it's just like another use case of. Okay. If we can interpret what's happening in post-training, you know, can we clear some of this? Can we even determine what's there? Because yeah, it's just like some worrying research that's out there that shows, you know, we really don't know what's going on.Mark Bissell [00:12:06]: That is. Yeah. I think that's the biggest sentiment that we're sort of hoping to tackle. Nobody knows what's going on. Right. Like subliminal learning is just an insane concept when you think about it. Right. Train a model on not even the logits, literally the output text of a bunch of random numbers. And now your model loves owls. And you see behaviors like that, that are just, they defy, they defy intuition. And, and there are mathematical explanations that you can get into, but. I mean.Shawn Wang [00:12:34]: It feels so early days. Objectively, there are a sequence of numbers that are more owl-like than others. There, there should be.Mark Bissell [00:12:40]: According to, according to certain models. Right. It's interesting. I think it only applies to models that were initialized from the same starting Z. Usually, yes.Shawn Wang [00:12:49]: But I mean, I think that's a, that's a cheat code because there's not enough compute. But like if you believe in like platonic representation, like probably it will transfer across different models as well. Oh, you think so?Mark Bissell [00:13:00]: I think of it more as a statistical artifact of models initialized from the same seed sort of. There's something that is like path dependent from that seed that might cause certain overlaps in the latent space and then sort of doing this distillation. Yeah. Like it pushes it towards having certain other tendencies.Vibhu Sapra [00:13:24]: Got it. I think there's like a bunch of these open-ended questions, right? Like you can't train in new stuff during the RL phase, right? RL only reorganizes weights and you can only do stuff that's somewhat there in your base model. You're not learning new stuff. You're just reordering chains and stuff. But okay. My broader question is when you guys work at an interp lab, how do you decide what to work on and what's kind of the thought process? Right. Because we can ramble for hours. Okay. I want to know this. I want to know that. But like, how do you concretely like, you know, what's the workflow? Okay. There's like approaches towards solving a problem, right? I can try prompting. I can look at chain of thought. I can train probes, SAEs. But how do you determine, you know, like, okay, is this going anywhere? Like, do we have set stuff? Just, you know, if you can help me with all that. Yeah.Myra Deng [00:14:07]: It's a really good question. I feel like we've always at the very beginning of the company thought about like, let's go and try to learn what isn't working in machine learning today. Whether that's talking to customers or talking to researchers at other labs, trying to understand both where the frontier is going and where things are really not falling apart today. And then developing a perspective on how we can push the frontier using interpretability methods. And so, you know, even our chief scientist, Tom, spends a lot of time talking to customers and trying to understand what real world problems are and then taking that back and trying to apply the current state of the art to those problems and then seeing where they fall down basically. And then using those failures or those shortcomings to understand what hills to climb when it comes to interpretability research. So like on the fundamental side, for instance, when we have done some work applying SAEs and probes, we've encountered, you know, some shortcomings in SAEs that we found a little bit surprising. And so have gone back to the drawing board and done work on that. And then, you know, we've done some work on better foundational interpreter models. And a lot of our team's research is focused on what is the next evolution beyond SAEs, for instance. And then when it comes to like control and design of models, you know, we tried steering with our first API and realized that it still fell short of black box techniques like prompting or fine tuning. And so went back to the drawing board and we're like, how do we make that not the case and how do we improve it beyond that? And one of our researchers, Ekdeep, who just joined is actually Ekdeep and Atticus are like steering experts and have spent a lot of time trying to figure out like, what is the research that enables us to actually do this in a much more powerful, robust way? So yeah, the answer is like, look at real world problems, try to translate that into a research agenda and then like hill climb on both of those at the same time.Shawn Wang [00:16:04]: Yeah. Mark has the steering CLI demo queued up, which we're going to go into in a sec. But I always want to double click on when you drop hints, like we found some problems with SAEs. Okay. What are they? You know, and then we can go into the demo. Yeah.Myra Deng [00:16:19]: I mean, I'm curious if you have more thoughts here as well, because you've done it in the healthcare domain. But I think like, for instance, when we do things like trying to detect behaviors within models that are harmful or like behaviors that a user might not want to have in their model. So hallucinations, for instance, harmful intent, PII, all of these things. We first tried using SAE probes for a lot of these tasks. So taking the feature activation space from SAEs and then training classifiers on top of that, and then seeing how well we can detect the properties that we might want to detect in model behavior. And we've seen in many cases that probes just trained on raw activations seem to perform better than SAE probes, which is a bit surprising if you think that SAEs are actually also capturing the concepts that you would want to capture cleanly and more surgically. And so that is an interesting observation. I don't think that is like, I'm not down on SAEs at all. I think there are many, many things they're useful for, but we have definitely run into cases where I think the concept space described by SAEs is not as clean and accurate as we would expect it to be for actual like real world downstream performance metrics.Mark Bissell [00:17:34]: Fair enough. Yeah. It's the blessing and the curse of unsupervised methods where you get to peek into the AI's mind. But sometimes you wish that you saw other things when you walked inside there. Although in the PII instance, I think weren't an SAE based approach actually did prove to be the most generalizable?Myra Deng [00:17:53]: It did work well in the case that we published with Rakuten. And I think a lot of the reasons it worked well was because we had a noisier data set. And so actually the blessing of unsupervised learning is that we actually got to get more meaningful, generalizable signal from SAEs when the data was noisy. But in other cases where we've had like good data sets, it hasn't been the case.Shawn Wang [00:18:14]: And just because you named Rakuten and I don't know if we'll get it another chance, like what is the overall, like what is Rakuten's usage or production usage? Yeah.Myra Deng [00:18:25]: So they are using us to essentially guardrail and inference time monitor their language model usage and their agent usage to detect things like PII so that they don't route private user information.Myra Deng [00:18:41]: And so that's, you know, going through all of their user queries every day. And that's something that we deployed with them a few months ago. And now we are actually exploring very early partnerships, not just with Rakuten, but with other people around how we can help with potentially training and customization use cases as well. Yeah.Shawn Wang [00:19:03]: And for those who don't know, like it's Rakuten is like, I think number one or number two e-commerce store in Japan. Yes. Yeah.Mark Bissell [00:19:10]: And I think that use case actually highlights a lot of like what it looks like to deploy things in practice that you don't always think about when you're doing sort of research tasks. So when you think about some of the stuff that came up there that's more complex than your idealized version of a problem, they were encountering things like synthetic to real transfer of methods. So they couldn't train probes, classifiers, things like that on actual customer data of PII. So what they had to do is use synthetic data sets. And then hope that that transfer is out of domain to real data sets. And so we can evaluate performance on the real data sets, but not train on customer PII. So that right off the bat is like a big challenge. You have multilingual requirements. So this needed to work for both English and Japanese text. Japanese text has all sorts of quirks, including tokenization behaviors that caused lots of bugs that caused us to be pulling our hair out. And then also a lot of tasks you'll see. You might make simplifying assumptions if you're sort of treating it as like the easiest version of the problem to just sort of get like general results where maybe you say you're classifying a sentence to say, does this contain PII? But the need that Rakuten had was token level classification so that you could precisely scrub out the PII. So as we learned more about the problem, you're sort of speaking about what that looks like in practice. Yeah. A lot of assumptions end up breaking. And that was just one instance where you. A problem that seems simple right off the bat ends up being more complex as you keep diving into it.Vibhu Sapra [00:20:41]: Excellent. One of the things that's also interesting with Interp is a lot of these methods are very efficient, right? So where you're just looking at a model's internals itself compared to a separate like guardrail, LLM as a judge, a separate model. One, you have to host it. Two, there's like a whole latency. So if you use like a big model, you have a second call. Some of the work around like self detection of hallucination, it's also deployed for efficiency, right? So if you have someone like Rakuten doing it in production live, you know, that's just another thing people should consider.Mark Bissell [00:21:12]: Yeah. And something like a probe is super lightweight. Yeah. It's no extra latency really. Excellent.Shawn Wang [00:21:17]: You have the steering demos lined up. So we were just kind of see what you got. I don't, I don't actually know if this is like the latest, latest or like alpha thing.Mark Bissell [00:21:26]: No, this is a pretty hacky demo from from a presentation that someone else on the team recently gave. So this will give a sense for, for technology. So you can see the steering and action. Honestly, I think the biggest thing that this highlights is that as we've been growing as a company and taking on kind of more and more ambitious versions of interpretability related problems, a lot of that comes to scaling up in various different forms. And so here you're going to see steering on a 1 trillion parameter model. This is Kimi K2. And so it's sort of fun that in addition to the research challenges, there are engineering challenges that we're now tackling. Cause for any of this to be sort of useful in production, you need to be thinking about what it looks like when you're using these methods on frontier models as opposed to sort of like toy kind of model organisms. So yeah, this was thrown together hastily, pretty fragile behind the scenes, but I think it's quite a fun demo. So screen sharing is on. So I've got two terminal sessions pulled up here. On the left is a forked version that we have of the Kimi CLI that we've got running to point at our custom hosted Kimi model. And then on the right is a set up that will allow us to steer on certain concepts. So I should be able to chat with Kimi over here. Tell it hello. This is running locally. So the CLI is running locally, but the Kimi server is running back to the office. Well, hopefully should be, um, that's too much to run on that Mac. Yeah. I think it's, uh, it takes a full, like each 100 node. I think it's like, you can. You can run it on eight GPUs, eight 100. So, so yeah, Kimi's running. We can ask it a prompt. It's got a forked version of our, uh, of the SG line code base that we've been working on. So I'm going to tell it, Hey, this SG line code base is slow. I think there's a bug. Can you try to figure it out? There's a big code base, so it'll, it'll spend some time doing this. And then on the right here, I'm going to initialize in real time. Some steering. Let's see here.Mark Bissell [00:23:33]: searching for any. Bugs. Feature ID 43205.Shawn Wang [00:23:38]: Yeah.Mark Bissell [00:23:38]: 20, 30, 40. So let me, uh, this is basically a feature that we found that inside Kimi seems to cause it to speak in Gen Z slang. And so on the left, it's still sort of thinking normally it might take, I don't know, 15 seconds for this to kick in, but then we're going to start hopefully seeing him do this code base is massive for real. So we're going to start. We're going to start seeing Kimi transition as the steering kicks in from normal Kimi to Gen Z Kimi and both in its chain of thought and its actual outputs.Mark Bissell [00:24:19]: And interestingly, you can see, you know, it's still able to call tools, uh, and stuff. It's um, it's purely sort of it's it's demeanor. And there are other features that we found for interesting things like concision. So that's more of a practical one. You can make it more concise. Um, the types of programs, uh, programming languages that uses, but yeah, as we're seeing it come in. Pretty good. Outputs.Shawn Wang [00:24:43]: Scheduler code is actually wild.Vibhu Sapra [00:24:46]: Yo, this code is actually insane, bro.Vibhu Sapra [00:24:53]: What's the process of training in SAE on this, or, you know, how do you label features? I know you guys put out a pretty cool blog post about, um, finding this like autonomous interp. Um, something. Something about how agents for interp is different than like coding agents. I don't know while this is spewing up, but how, how do we find feature 43, two Oh five. Yeah.Mark Bissell [00:25:15]: So in this case, um, we, our platform that we've been building out for a long time now supports all the sort of classic out of the box interp techniques that you might want to have like SAE training, probing things of that kind, I'd say the techniques for like vanilla SAEs are pretty well established now where. You take your model that you're interpreting, run a whole bunch of data through it, gather activations, and then yeah, pretty straightforward pipeline to train an SAE. There are a lot of different varieties. There's top KSAEs, batch top KSAEs, um, normal ReLU SAEs. And then once you have your sparse features to your point, assigning labels to them to actually understand that this is a gen Z feature, that's actually where a lot of the kind of magic happens. Yeah. And the most basic standard technique is look at all of your d input data set examples that cause this feature to fire most highly. And then you can usually pick out a pattern. So for this feature, If I've run a diverse enough data set through my model feature 43, two Oh five. Probably tends to fire on all the tokens that sounds like gen Z slang. You know, that's the, that's the time of year to be like, Oh, I'm in this, I'm in this Um, and, um, so, you know, you could have a human go through all 43,000 concepts andVibhu Sapra [00:26:34]: And I've got to ask the basic question, you know, can we get examples where it hallucinates, pass it through, see what feature activates for hallucinations? Can I just, you know, turn hallucination down?Myra Deng [00:26:51]: Oh, wow. You really predicted a project we're already working on right now, which is detecting hallucinations using interpretability techniques. And this is interesting because hallucinations is something that's very hard to detect. And it's like a kind of a hairy problem and something that black box methods really struggle with. Whereas like Gen Z, you could always train a simple classifier to detect that hallucinations is harder. But we've seen that models internally have some... Awareness of like uncertainty or some sort of like user pleasing behavior that leads to hallucinatory behavior. And so, yeah, we have a project that's trying to detect that accurately. And then also working on mitigating the hallucinatory behavior in the model itself as well.Shawn Wang [00:27:39]: Yeah, I would say most people are still at the level of like, oh, I would just turn temperature to zero and that turns off hallucination. And I'm like, well, that's a fundamental misunderstanding of how this works. Yeah.Mark Bissell [00:27:51]: Although, so part of what I like about that question is you, there are SAE based approaches that might like help you get at that. But oftentimes the beauty of SAEs and like we said, the curse is that they're unsupervised. So when you have a behavior that you deliberately would like to remove, and that's more of like a supervised task, often it is better to use something like probes and specifically target the thing that you're interested in reducing as opposed to sort of like hoping that when you fragment the latent space, one of the vectors that pops out.Vibhu Sapra [00:28:20]: And as much as we're training an autoencoder to be sparse, we're not like for sure certain that, you know, we will get something that just correlates to hallucination. You'll probably split that up into 20 other things and who knows what they'll be.Mark Bissell [00:28:36]: Of course. Right. Yeah. So there's no sort of problems with like feature splitting and feature absorption. And then there's the off target effects, right? Ideally, you would want to be very precise where if you reduce the hallucination feature, suddenly maybe your model can't write. Creatively anymore. And maybe you don't like that, but you want to still stop it from hallucinating facts and figures.Shawn Wang [00:28:55]: Good. So Vibhu has a paper to recommend there that we'll put in the show notes. But yeah, I mean, I guess just because your demo is done, any any other things that you want to highlight or any other interesting features you want to show?Mark Bissell [00:29:07]: I don't think so. Yeah. Like I said, this is a pretty small snippet. I think the main sort of point here that I think is exciting is that there's not a whole lot of inter being applied to models quite at this scale. You know, Anthropic certainly has some some. Research and yeah, other other teams as well. But it's it's nice to see these techniques, you know, being put into practice. I think not that long ago, the idea of real time steering of a trillion parameter model would have sounded.Shawn Wang [00:29:33]: Yeah. The fact that it's real time, like you started the thing and then you edited the steering vector.Vibhu Sapra [00:29:38]: I think it's it's an interesting one TBD of what the actual like production use case would be on that, like the real time editing. It's like that's the fun part of the demo, right? You can kind of see how this could be served behind an API, right? Like, yes, you're you only have so many knobs and you can just tweak it a bit more. And I don't know how it plays in. Like people haven't done that much with like, how does this work with or without prompting? Right. How does this work with fine tuning? Like, there's a whole hype of continual learning, right? So there's just so much to see. Like, is this another parameter? Like, is it like parameter? We just kind of leave it as a default. We don't use it. So I don't know. Maybe someone here wants to put out a guide on like how to use this with prompting when to do what?Mark Bissell [00:30:18]: Oh, well, I have a paper recommendation. I think you would love from Act Deep on our team, who is an amazing researcher, just can't say enough amazing things about Act Deep. But he actually has a paper that as well as some others from the team and elsewhere that go into the essentially equivalence of activation steering and in context learning and how those are from a he thinks of everything in a cognitive neuroscience Bayesian framework, but basically how you can precisely show how. Prompting in context, learning and steering exhibit similar behaviors and even like get quantitative about the like magnitude of steering you would need to do to induce a certain amount of behavior similar to certain prompting, even for things like jailbreaks and stuff. It's a really cool paper. Are you saying steering is less powerful than prompting? More like you can almost write a formula that tells you how to convert between the two of them.Myra Deng [00:31:20]: And so like formally equivalent actually in the in the limit. Right.Mark Bissell [00:31:24]: So like one case study of this is for jailbreaks there. I don't know. Have you seen the stuff where you can do like many shot jailbreaking? You like flood the context with examples of the behavior. And the topic put out that paper.Shawn Wang [00:31:38]: A lot of people were like, yeah, we've been doing this, guys.Mark Bissell [00:31:40]: Like, yeah, what's in this in context learning and activation steering equivalence paper is you can like predict the number. Number of examples that you will need to put in there in order to jailbreak the model. That's cool. By doing steering experiments and using this sort of like equivalence mapping. That's cool. That's really cool. It's very neat. Yeah.Shawn Wang [00:32:02]: I was going to say, like, you know, I can like back rationalize that this makes sense because, you know, what context is, is basically just, you know, it updates the KV cache kind of and like and then every next token inference is still like, you know, the sheer sum of everything all the way. It's plus all the context. It's up to date. And you could, I guess, theoretically steer that with you probably replace that with your steering. The only problem is steering typically is on one layer, maybe three layers like like you did. So it's like not exactly equivalent.Mark Bissell [00:32:33]: Right, right. There's sort of you need to get precise about, yeah, like how you sort of define steering and like what how you're modeling the setup. But yeah, I've got the paper pulled up here. Belief dynamics reveal the dual nature. Yeah. The title is Belief Dynamics Reveal the Dual Nature of Incompetence. And it's an exhibition of the practical context learning and activation steering. So Eric Bigelow, Dan Urgraft on the who are doing fellowships at Goodfire, Ekt Deep's the final author there.Myra Deng [00:32:59]: I think actually to your question of like, what is the production use case of steering? I think maybe if you just think like one level beyond steering as it is today. Like imagine if you could adapt your model to be, you know, an expert legal reasoner. Like in almost real time, like very quickly. efficiently using human feedback or using like your semantic understanding of what the model knows and where it knows that behavior. I think that while it's not clear what the product is at the end of the day, it's clearly very valuable. Thinking about like what's the next interface for model customization and adaptation is a really interesting problem for us. Like we have heard a lot of people actually interested in fine-tuning an RL for open weight models in production. And so people are using things like Tinker or kind of like open source libraries to do that, but it's still very difficult to get models fine-tuned and RL'd for exactly what you want them to do unless you're an expert at model training. And so that's like something we'reShawn Wang [00:34:06]: looking into. Yeah. I never thought so. Tinker from Thinking Machines famously uses rank one LoRa. Is that basically the same as steering? Like, you know, what's the comparison there?Mark Bissell [00:34:19]: Well, so in that case, you are still applying updates to the parameters, right?Shawn Wang [00:34:25]: Yeah. You're not touching a base model. You're touching an adapter. It's kind of, yeah.Mark Bissell [00:34:30]: Right. But I guess it still is like more in parameter space then. I guess it's maybe like, are you modifying the pipes or are you modifying the water flowing through the pipes to get what you're after? Yeah. Just maybe one way.Mark Bissell [00:34:44]: I like that analogy. That's my mental map of it at least, but it gets at this idea of model design and intentional design, which is something that we're, that we're very focused on. And just the fact that like, I hope that we look back at how we're currently training models and post-training models and just think what a primitive way of doing that right now. Like there's no intentionalityShawn Wang [00:35:06]: really in... It's just data, right? The only thing in control is what data we feed in.Mark Bissell [00:35:11]: So, so Dan from Goodfire likes to use this analogy of, you know, he has a couple of young kids and he talks about like, what if I could only teach my kids how to be good people by giving them cookies or like, you know, giving them a slap on the wrist if they do something wrong, like not telling them why it was wrong or like what they should have done differently or something like that. Just figure it out. Right. Exactly. So that's RL. Yeah. Right. And, and, you know, it's sample inefficient. There's, you know, what do they say? It's like slurping feedback. It's like, slurping supervision. Right. And so you'd like to get to the point where you can have experts giving feedback to their models that are, uh, internalized and, and, you know, steering is an inference time way of sort of getting that idea. But ideally you're moving to a world whereVibhu Sapra [00:36:04]: it is much more intentional design in perpetuity for these models. Okay. This is one of the questions we asked Emmanuel from Anthropic on the podcast a few months ago. Basically the question, was you're at a research lab that does model training, foundation models, and you're on an interp team. How does it tie back? Right? Like, does this, do ideas come from the pre-training team? Do they go back? Um, you know, so for those interested, you can, you can watch that. There wasn't too much of a connect there, but it's still something, you know, it's something they want toMark Bissell [00:36:33]: push for down the line. It can be useful for all of the above. Like there are certainly post-hocVibhu Sapra [00:36:39]: use cases where it doesn't need to touch that. I think the other thing a lot of people forget is this stuff isn't too computationally expensive, right? Like I would say, if you're interested in getting into research, MechInterp is one of the most approachable fields, right? A lot of this train an essay, train a probe, this stuff, like the budget for this one, there's already a lot done. There's a lot of open source work. You guys have done some too. Um, you know,Shawn Wang [00:37:04]: There's like notebooks from the Gemini team for Neil Nanda or like, this is how you do it. Just step through the notebook.Vibhu Sapra [00:37:09]: Even if you're like, not even technical with any of this, you can still make like progress. There, you can look at different activations, but, uh, if you do want to get into training, you know, training this stuff, correct me if I'm wrong is like in the thousands of dollars, not even like, it's not that high scale. And then same with like, you know, applying it, doing it for post-training or all this stuff is fairly cheap in scale of, okay. I want to get into like model training. I don't have compute for like, you know, pre-training stuff. So it's, it's a very nice field to get into. And also there's a lot of like open questions, right? Um, some of them have to go with, okay, I want a product. I want to solve this. Like there's also just a lot of open-ended stuff that people could work on. That's interesting. Right. I don't know if you guys have any calls for like, what's open questions, what's open work that you either open collaboration with, or like, you'd just like to see solved or just, you know, for people listening that want to get into McInturk because people always talk about it. What are, what are the things they should check out? Start, of course, you know, join you guys as well. I'm sure you're hiring.Myra Deng [00:38:09]: There's a paper, I think from, was it Lee, uh, Sharky? It's open problems and, uh, it's, it's a bit of interpretability, which I recommend everyone who's interested in the field. Read. I'm just like a really comprehensive overview of what are the things that experts in the field think are the most important problems to be solved. I also think to your point, it's been really, really inspiring to see, I think a lot of young people getting interested in interpretability, actually not just young people also like scientists to have been, you know, experts in physics for many years and in biology or things like this, um, transitioning into interp, because the barrier of, of what's now interp. So it's really cool to see a number to entry is, you know, in some ways low and there's a lot of information out there and ways to get started. There's this anecdote of like professors at universities saying that all of a sudden every incoming PhD student wants to study interpretability, which was not the case a few years ago. So it just goes to show how, I guess, like exciting the field is, how fast it's moving, how quick it is to get started and things like that.Mark Bissell [00:39:10]: And also just a very welcoming community. You know, there's an open source McInturk Slack channel. There are people are always posting questions and just folks in the space are always responsive if you ask things on various forums and stuff. But yeah, the open paper, open problems paper is a really good one.Myra Deng [00:39:28]: For other people who want to get started, I think, you know, MATS is a great program. What's the acronym for? Machine Learning and Alignment Theory Scholars? It's like the...Vibhu Sapra [00:39:40]: Normally summer internship style.Myra Deng [00:39:42]: Yeah, but they've been doing it year round now. And actually a lot of our full-time staff have come through that program or gone through that program. And it's great for anyone who is transitioning into interpretability. There's a couple other fellows programs. We do one as well as Anthropic. And so those are great places to get started if anyone is interested.Mark Bissell [00:40:03]: Also, I think been seen as a research field for a very long time. But I think engineering... I think engineers are sorely wanted for interpretability as well, especially at Goodfire, but elsewhere, as it does scale up.Shawn Wang [00:40:18]: I should mention that Lee actually works with you guys, right? And in the London office and I'm adding our first ever McInturk track at AI Europe because I see this industry applications now emerging. And I'm pretty excited to, you know, help push that along. Yeah, I was looking forward to that. It'll effectively be the first industry McInturk conference. Yeah. I'm so glad you added that. You know, it's still a little bit of a bet. It's not that widespread, but I can definitely see this is the time to really get into it. We want to be early on things.Mark Bissell [00:40:51]: For sure. And I think the field understands this, right? So at ICML, I think the title of the McInturk workshop this year was actionable interpretability. And there was a lot of discussion around bringing it to various domains. Everyone's adding pragmatic, actionable, whatever.Shawn Wang [00:41:10]: It's like, okay, well, we weren't actionable before, I guess. I don't know.Vibhu Sapra [00:41:13]: And I mean, like, just, you know, being in Europe, you see the Interp room. One, like old school conferences, like, I think they had a very tiny room till they got lucky and they got it doubled. But there's definitely a lot of interest, a lot of niche research. So you see a lot of research coming out of universities, students. We covered the paper last week. It's like two unknown authors, not many citations. But, you know, you can make a lot of meaningful work there. Yeah. Yeah. Yeah.Shawn Wang [00:41:39]: Yeah. I think people haven't really mentioned this yet. It's just Interp for code. I think it's like an abnormally important field. We haven't mentioned this yet. The conspiracy theory last two years ago was when the first SAE work came out of Anthropic was they would do like, oh, we just used SAEs to turn the bad code vector down and then turn up the good code. And I think like, isn't that the dream? Like, you know, like, but basically, I guess maybe, why is it funny? Like, it's... If it was realistic, it would not be funny. It would be like, no, actually, we should do this. But it's funny because we know there's like, we feel there's some limitations to what steering can do. And I think a lot of the public image of steering is like the Gen Z stuff. Like, oh, you can make it really love the Golden Gate Bridge, or you can make it speak like Gen Z. To like be a legal reasoner seems like a huge stretch. Yeah. And I don't know if that will get there this way. Yeah.Myra Deng [00:42:36]: I think, um, I will say we are announcing. Something very soon that I will not speak too much about. Um, but I think, yeah, this is like what we've run into again and again is like, we, we don't want to be in the world where steering is only useful for like stylistic things. That's definitely not, not what we're aiming for. But I think the types of interventions that you need to do to get to things like legal reasoning, um, are much more sophisticated and require breakthroughs in, in learning algorithms. And that's, um...Shawn Wang [00:43:07]: And is this an emergent property of scale as well?Myra Deng [00:43:10]: I think so. Yeah. I mean, I think scale definitely helps. I think scale allows you to learn a lot of information and, and reduce noise across, you know, large amounts of data. But I also think we think that there's ways to do things much more effectively, um, even, even at scale. So like actually learning exactly what you want from the data and not learning things that you do that you don't want exhibited in the data. So we're not like anti-scale, but we are also realizing that scale is not going to get us anywhere. It's not going to get us to the type of AI development that we want to be at in, in the future as these models get more powerful and get deployed in all these sorts of like mission critical contexts. Current life cycle of training and deploying and evaluations is, is to us like deeply broken and has opportunities to, to improve. So, um, more to come on that very, very soon.Mark Bissell [00:44:02]: And I think that that's a use basically, or maybe just like a proof point that these concepts do exist. Like if you can manipulate them in the precise best way, you can get the ideal combination of them that you desire. And steering is maybe the most coarse grained sort of peek at what that looks like. But I think it's evocative of what you could do if you had total surgical control over every concept, every parameter. Yeah, exactly.Myra Deng [00:44:30]: There were like bad code features. I've got it pulled up.Vibhu Sapra [00:44:33]: Yeah. Just coincidentally, as you guys are talking.Shawn Wang [00:44:35]: This is like, this is exactly.Vibhu Sapra [00:44:38]: There's like specifically a code error feature that activates and they show, you know, it's not, it's not typo detection. It's like, it's, it's typos in code. It's not typical typos. And, you know, you can, you can see it clearly activates where there's something wrong in code. And they have like malicious code, code error. They have a whole bunch of sub, you know, sub broken down little grain features. Yeah.Shawn Wang [00:45:02]: Yeah. So, so the, the rough intuition for me, the, why I talked about post-training was that, well, you just, you know, have a few different rollouts with all these things turned off and on and whatever. And then, you know, you can, that's, that's synthetic data you can kind of post-train on. Yeah.Vibhu Sapra [00:45:13]: And I think we make it sound easier than it is just saying, you know, they do the real hard work.Myra Deng [00:45:19]: I mean, you guys, you guys have the right idea. Exactly. Yeah. We replicated a lot of these features in, in our Lama models as well. I remember there was like.Vibhu Sapra [00:45:26]: And I think a lot of this stuff is open, right? Like, yeah, you guys opened yours. DeepMind has opened a lot of essays on Gemma. Even Anthropic has opened a lot of this. There's, there's a lot of resources that, you know, we can probably share of people that want to get involved.Shawn Wang [00:45:41]: Yeah. And special shout out to like Neuronpedia as well. Yes. Like, yeah, amazing piece of work to visualize those things.Myra Deng [00:45:49]: Yeah, exactly.Shawn Wang [00:45:50]: I guess I wanted to pivot a little bit on, onto the healthcare side, because I think that's a big use case for you guys. We haven't really talked about it yet. This is a bit of a crossover for me because we are, we are, we do have a separate science pod that we're starting up for AI, for AI for science, just because like, it's such a huge investment category and also I'm like less qualified to do it, but we actually have bio PhDs to cover that, which is great, but I need to just kind of recover, recap your work, maybe on the evil two stuff, but then, and then building forward.Mark Bissell [00:46:17]: Yeah, for sure. And maybe to frame up the conversation, I think another kind of interesting just lens on interpretability in general is a lot of the techniques that were described. are ways to solve the AI human interface problem. And it's sort of like bidirectional communication is the goal there. So what we've been talking about with intentional design of models and, you know, steering, but also more advanced techniques is having humans impart our desires and control into models and over models. And the reverse is also very interesting, especially as you get to superhuman models, whether that's narrow superintelligence, like these scientific models that work on genomics, data, medical imaging, things like that. But down the line, you know, superintelligence of other forms as well. What knowledge can the AIs teach us as sort of that, that the other direction in that? And so some of our life science work to date has been getting at exactly that question, which is, well, some of it does look like debugging these various life sciences models, understanding if they're actually performing well, on tasks, or if they're picking up on spurious correlations, for instance, genomics models, you would like to know whether they are sort of focusing on the biologically relevant things that you care about, or if it's using some simpler correlate, like the ancestry of the person that it's looking at. But then also in the instances where they are superhuman, and maybe they are understanding elements of the human genome that we don't have names for or specific, you know, yeah, discoveries that they've made that that we don't know about, that's, that's a big goal. And so we're already seeing that, right, we are partnered with organizations like Mayo Clinic, leading research health system in the United States, our Institute, as well as a startup called Prima Menta, which focuses on neurodegenerative disease. And in our partnership with them, we've used foundation models, they've been training and applied our interpretability techniques to find novel biomarkers for Alzheimer's disease. So I think this is just the tip of the iceberg. But it's, that's like a flavor of some of the things that we're working on.Shawn Wang [00:48:36]: Yeah, I think that's really fantastic. Obviously, we did the Chad Zuckerberg pod last year as well. And like, there's a plethora of these models coming out, because there's so much potential and research. And it's like, very interesting how it's basically the same as language models, but just with a different underlying data set. But it's like, it's the same exact techniques. Like, there's no change, basically.Mark Bissell [00:48:59]: Yeah. Well, and even in like other domains, right? Like, you know, robotics, I know, like a lot of the companies just use Gemma as like the like backbone, and then they like make it into a VLA that like takes these actions. It's, it's, it's transformers all the way down. So yeah.Vibhu Sapra [00:49:15]: Like we have Med Gemma now, right? Like this week, even there was Med Gemma 1.5. And they're training it on this stuff, like 3d scans, medical domain knowledge, and all that stuff, too. So there's a push from both sides. But I think the thing that, you know, one of the things about McInturpp is like, you're a little bit more cautious in some domains, right? So healthcare, mainly being one, like guardrails, understanding, you know, we're more risk adverse to something going wrong there. So even just from a basic understanding, like, if we're trusting these systems to make claims, we want to know why and what's going on.Myra Deng [00:49:51]: Yeah, I think there's totally a kind of like deployment bottleneck to actually using. foundation models for real patient usage or things like that. Like, say you're using a model for rare disease prediction, you probably want some explanation as to why your model predicted a certain outcome, and an interpretable explanation at that. So that's definitely a use case. But I also think like, being able to extract scientific information that no human knows to accelerate drug discovery and disease treatment and things like that actually is a really, really big unlock for science, like scientific discovery. And you've seen a lot of startups, like say that they're going to accelerate scientific discovery. And I feel like we actually are doing that through our interp techniques. And kind of like, almost by accident, like, I think we got reached out to very, very early on from these healthcare institutions. And none of us had healthcare.Shawn Wang [00:50:49]: How did they even hear of you? A podcast.Myra Deng [00:50:51]: Oh, okay. Yeah, podcast.Vibhu Sapra [00:50:53]: Okay, well, now's that time, you know.Myra Deng [00:50:55]: Everyone can call us.Shawn Wang [00:50:56]: Podcasts are the most important thing. Everyone should listen to podcasts.Myra Deng [00:50:59]: Yeah, they reached out. They were like, you know, we have these really smart models that we've trained, and we want to know what they're doing. And we were like, really early that time, like three months old, and it was a few of us. And we were like, oh, my God, we've never used these models. Let's figure it out. But it's also like, great proof that interp techniques scale pretty well across domains. We didn't really have to learn too much about.Shawn Wang [00:51:21]: Interp is a machine learning technique, machine learning skills everywhere, right? Yeah. And it's obviously, it's just like a general insight. Yeah. Probably to finance too, I think, which would be fun for our history. I don't know if you have anything to say there.Mark Bissell [00:51:34]: Yeah, well, just across the science. Like, we've also done work on material science. Yeah, it really runs the gamut.Vibhu Sapra [00:51:40]: Yeah. Awesome. And, you know, for those that should reach out, like, you're obviously experts in this, but like, is there a call out for people that you're looking to partner with, design partners, people to use your stuff outside of just, you know, the general developer that wants to. Plug and play steering stuff, like on the research side more so, like, are there ideal design partners, customers, stuff like that?Myra Deng [00:52:03]: Yeah, I can talk about maybe non-life sciences, and then I'm curious to hear from you on the life sciences side. But we're looking for design partners across many domains, language, anyone who's customizing language models or trying to push the frontier of code or reasoning models is really interesting to us. And then also interested in the frontier of modeling. There's a lot of models that work in, like, pixel space, as we call it. So if you're doing world models, video models, even robotics, where there's not a very clean natural language interface to interact with, I think we think that Interp can really help and are looking for a few partners in that space.Shawn Wang [00:52:43]: Just because you mentioned the keyword

How are hospitals using AI and HPC to assist them in helping save lives? This week, Technology Now is joined by Keith Perry, Senior Vice President and Chief Information Officer at St. Jude Children's Research Hospital to explore how St Jude uses the latest technologies to help treat and prevent illness and catastrophic disease, giving patients and families more time, and more hope, when it comes to diagnosis.This is Technology Now, a weekly show from Hewlett Packard Enterprise. Every week, hosts Michael Bird and Sam Jarrell look at a story that's been making headlines, take a look at the technology behind it, and explain why it matters to organizations.About Keith:https://www.linkedin.com/in/keith-perry-8562347/Sources:Hernigou P. Ambroise Paré III: Paré's contributions to surgical instruments and surgical instruments at the time of Ambroise Paré. Int Orthop. 2013 May;37(5):975-80. doi: 10.1007/s00264-013-1872-y. Epub 2013 Apr 12. PMID: 23580029; PMCID: PMC3631503.https://www.surgicalholdings.co.uk/history-of-surgical-instruments.htmlSmith-Bindman R, Kwan ML, Marlow EC, et al. Trends in Use of Medical Imaging in US Health Care Systems and in Ontario, Canada, 2000-2016. JAMA. 2019;322(9):843–856. doi:10.1001/jama.2019.11456https://caferoentgen.com/2023/10/07/a-tale-of-two-hands-the-story-behind-the-two-famous-radiographs-captured-by-wilhelm-roentgen/https://www.orau.org/health-physics-museum/collection/shoe-fitting-fluoroscope/index.html

On this Episode, Mr. Robert Neil, Army's Contract Litigation and Intellectual Property Division, and Ms. Erika Whelan Retta, Air Force's Acquisitions, Fiscal Law and Litigation Directorate, join us this week to talk about AI hallucinations. We talk about recent case decisions and the impact AI has on Bid Protests. Cases discussed: Raven Investigations & Security Consulting, LLC, B-423447, May 7, 2025, 2025 CPD ¶ 112; Oready, LLC, B-423649 et al., Sept. 25, 2025, 2025 CPD ¶ 238; Sanders v. USA, No. 1:2024cv01301 (Fed. Cl. 2025). Learn more about The Quill & Sword series of podcasts by visiting our podcast page at https://tjaglcs.army.mil/thequillandsword. The Quill & Sword show includes featured episodes from across the JAGC, plus all episodes from our four separate shows: “Criminal Law Department Presents” (Criminal Law Department), “NSL Unscripted” (National Security Law Department), “The FAR and Beyond” (Contract & Fiscal Law Department) and “Hold My Reg” (Administrative & Civil Law Department). Connect with The Judge Advocate General's Legal Center and School by visiting our website at https://tjaglcs.army.mil/ or on Facebook (tjaglcs), Instagram (tjaglcs), or LinkedIn (school/tjaglcs).

I look back at the Starcraft miniset that came with The Great Dark Beyond before playing Protoss Priest on the ladder. You can find the deck import code below the following contact links.  You can follow me @blisterguy on Twitch, Bluesky, and Youtube. Join our Discord community here or at discord.me/blisterguy. You can support this podcast and my other Hearthstone work at Patreon here. # 2x (0) Gravity Lapse # 2x (1) Catch of the Day # 2x (1) Hallucination # 2x (2) Birdwatching # 2x (2) Photon Cannon # 2x (2) Sentry # 1x (3) Chillin' Vol'jin # 1x (3) Demolition Renovator # 2x (3) Intertwined Fate # 2x (3) Trusty Fishing Rod # 2x (3) Void Ray # 1x (4) Elise the Navigator # 1x (4) Narain Soothfancy # 2x (5) Chrono Boost # 2x (6) Resuscitate # 1x (7) Sasquawk # 1x (8) Artanis # 2x (12) Mothership #  AAECAZ/HAgavwQbZwQbX0gaT9AaDigeCmAcMzsAGjMEGi9YGwOYGi/QGkPQGmPQGs/QGxfgGyvgGtZYHna0HAAA=

Let's talk about the AI elephant in the room: hallucinations.

Travis and Jake are joined by journalist Mike Rothschild to discuss the second time that the federal government has killed someone in Minnesota then smeared them as a terrorist, the growing paranoia on the right that opposition to occupation by federal agents constitutes an “insurgency,” the White House manipulating photos using AI, Kash Patel treating the FBI like his personal Hype House, and alleged time travel by the late Charlie Kirk. Subscribe for $5 a month to get all the premium QAA episodes: www.patreon.com/qaa Mike Rothschild https://x.com/rothschildmd https://bsky.app/profile/rothschildmd.bsky.social Mike Rothschild's Patreon https://www.patreon.com/cw/MikeRothschild Check out our new podcast series network Cursed Media and binge the entirety of our exclusive shows Science in Transition by Liv Agar and Truly, Tradly, Deely by Annie Kelly https://cursedmedia.net Editing by Corey Klotz. Theme by Nick Sena. Additional music by Pontus Berghe. Theme Vocals by THEY/LIVE (https://instagram.com/theyylivve / https://sptfy.com/QrDm). Cover Art by Pedro Correa: (https://pedrocorrea.com) https://qaapodcast.com QAA was known as the QAnon Anonymous podcast.

When a patient says, "I biffed the car," how should that be translated? Puzzles like this represent the gap between description and diagnosis, and are a critical part of neurological practice. In this podcast for the February 2026 issue of Practical Neurology, editors Phil Smith and Geraint Fuller take turns decoding some of the mysteries of everyday neurology. They cover dementia with Lewy bodies, osteoporosis and fracture risk, and anxiety and depression in epilepsy patients, as a sample of some of the published work in the latest journal. There's also a guide to the latest stroke rehabilitation guidelines, freezing of gait, and a farewell to a 'nom de plume'. Read the issue: https://pn.bmj.com/content/26/1/1   Please subscribe to the Practical Neurology podcast on your favourite platform to get the latest podcast every month. If you enjoy our podcast, you can leave us a review or a comment on Apple Podcasts (https://apple.co/3vVPClm) or Spotify (https://spoti.fi/4baxjsQ). We'd love to hear your feedback on social media - @PracticalNeurol. Production and editing by Brian O'Toole. Thank you for listening. 

What is an AI Hallucination? AI Deployment guru Joseph Noble from legionsoftware.com breaks it down. Full Length podcast - 'The AI Trap' releasing Tuesday Jan 27.

Why did a German immigration officer deny asylum to a Russian torture victim by citing a press release from Sergei Shoigu? Why is the Russian Orthodox Church building an Artificial Intelligence to automate the writing of police denunciations? And why is Vladimir Putin personally designing a business plan for a single pie shop while the national gold reserves vanish?In this episode of The Eastern Border, we smash the glass of the "Republic of Fake." We travel from the heated offices of Berlin bureaucrats who have accepted the Kremlin's lies as legal truth, to the frontlines where "Turbo-Patriots" are realizing that their holy war has become nothing more than an unfashionable subculture.We dig into the "Kickback Empire" where 50% of the imperial budget is stolen by handlers in Moscow, meet "Yura Unitaz"—the alleged toilet salesman in charge of Russia's drone war—and analyze the forensic data proving the Kadyrov regime is running on autopilot.The cake is rotting, friends. Don't eat it.IN THIS EPISODE:The Shoigu Precedent: How German bureaucracy was hacked by Russian apathy.The Mashenka Economy: The President, the Pie Shop, and the 1998 oil prices.Orthodox GPT: Outsourcing Judas to a server farm.The Kadyrov Glitch: Waze data, deleted Instagram posts, and the panic in Grozny.The Dead End: Why Russian nationalists admit "The SMO is out of fashion."SUPPORT THE FRONT LINE: Help our friends at Car4Ukraine turn civilian trucks into life-saving "Christmas Tree Trucks" for the defenders on the zero line:

End Legal AI Hallucinations Now. Discover how Clip's unique architecture delivers precise, sourced answers for attorneys, turning case files into a strategic advantage. See the real-time power that helped a law firm challenge testimony mid-deposition. Watch the full podcast trailer to learn how focused AI is transforming legal practice.

Ravi Shankar, Senior Vice President and Chief Marketing Officer at Denodo, the leading logical data management platform and foundation for transforming data into trusted, AI-ready … Read more The post How to stop hallucination and sycophancy in Agentic AI with Logical Data appeared first on Top Entrepreneurs Podcast | Enterprise Podcast Network.

#SecurityConfidential #DarkRhiinoSecurityDr. Rizwan Sheikh is the Founder and CEO of Global AI Excellence, the inventor of a groundbreaking AI governance model, and one of the leading voices helping organizations deploy AI responsibly. With decades of experience spanning Deloitte, Fortune 500 consulting, and teaching AI strategy at places like Harvard and MIT, Dr. Riz is helping bridge the gap between innovation and governance. 01:00 Intro02:34 Our Guest03:33 93% of Employees are using AI05:04 Why don't existing governance models work?06:15 ChatGPT Sources are fake07:17 A.I Hallucinates12:06 How do you govern Ai?22:02 When should you govern Ai?31:47 What should AI look like for my company? 33:45 European Union AI Act39:05 Believing False AI statements43:20 How can you build governance to something that changes daily?45:43 How do I get into AI?48:06 Connecting with Dr. Rizwan

#SecurityConfidential #DarkRhiinoSecurityDr. Rizwan Sheikh is the Founder and CEO of Global AI Excellence, the inventor of a groundbreaking AI governance model, and one of the leading voices helping organizations deploy AI responsibly. With decades of experience spanning Deloitte, Fortune 500 consulting, and teaching AI strategy at places like Harvard and MIT, Dr. Riz is helping bridge the gap between innovation and governance. 01:00 Intro02:34 Our Guest03:33 93% of Employees are using AI05:04 Why don't existing governance models work?06:15 ChatGPT Sources are fake07:17 A.I Hallucinates12:06 How do you govern Ai?22:02 When should you govern Ai?31:47 What should AI look like for my company? 33:45 European Union AI Act39:05 Believing False AI statements43:20 How can you build governance to something that changes daily?45:43 How do I get into AI?48:06 Connecting with Dr. Rizwan

News and Updates: AI Report Hallucinations in Utah- Heber City police discovered their AI software, Draft One, mistakenly claimed an officer became a frog because it transcribed background audio from a Disney movie nearby. Apple and Google AI Partnership- Apple partnered with Google to use Gemini models for future Apple Intelligence features, admitting Google's technology is superior for powering the upcoming next-generation Siri upgrade. Copilot Integration in Windows File Explorer- Microsoft is testing a "Chat with Copilot" button within File Explorer, aiming to improve document searching and introduce a new framework for specialized AI agents.

Dwayne Kerrigan sits down with world-class endurance athlete, firefighter, nonprofit founder, and keynote speaker Robyn Benincasa to unpack what truly separates great teams from the rest. Drawing from decades of extreme adventure racing, Robyn shares how elite teams win not by being the most talented, but by being the most committed to each other. She introduces her powerful TEAMWORK framework, revealing why total commitment, empathy, adversity management, mutual respect, and relinquishing ego are the real competitive advantages—whether you're racing through jungles or leading a modern organization. Through unforgettable stories—including hallucinations after days without sleep, tying boats together to beat world champions, and redefining leadership mid-race—Robyn shows how purpose, preparation, creativity, and shared ownership create cultures that don't just survive pressure… they win because of it. This episode is a masterclass in leadership, resilience, and building teams that operate as one heart, one mind, especially when the stakes are high and the path forward is uncertain. EPISODE HIGHLIGHTS: 00:00 – Robyn opens with the defining trait of elite teammates: leaving ego at the start line. 01:00 – Dwayne formally introduces Robyn and outlines her extraordinary background. 03:00 – Robyn shares discovering kayaking after hip surgery and focusing on what she could do. 06:30 – Why progress toward a meaningful goal is what makes humans feel alive. 10:30 – Competing to explore personal limits rather than seeking validation or approval. 14:00 – Why great teams care more about each other than themselves. 18:00 – How Robyn accidentally became a speaker after Fast Company's “Extreme Teamwork” 21:30 – The importance of leaving ego behind and accepting help to win as a team. 25:30 – The “Steve Gurney Missile” story and choosing to race to win instead of not lose. 30:00 – Creativity, calculated risk, and living in your strengths under pressure. 34:30 – Relinquishing ego, rotating leadership, and leading based on strengths—not titles. 39:00 – Hallucinations, extreme fatigue, and supporting teammates through suffering. 42:00 – Kinetic leadership and adapting leadership styles to what the team needs. 45:30 – Purpose, coaching influence, and how early mentors shaped Robyn's drive. 50:30 – Innovation, self-awareness, and evolving by leaning into strengths. 56:00 – Finding a greater purpose in business.KEY TAKEAWAYS: Winning teams prioritize commitment to each other, not individual performance. Progress toward a meaningful goal is what makes humans feel alive. Creativity and innovation emerge when teams operate from trust and purpose. Leadership should rotate based on strengths, not titles or tenure. Accepting help is not a weakness, it's how teams move faster and farther. Great leaders show people how amazing they are, not how amazing the leader is. NOTABLE QUOTES: “ I feel weird when I don't have a goal. I get my juju, I get my energy from...

Today on the show, Lisa was joined by returning guest Genevieve Moore of Soul Meets Body! Lisa and Genevieve catch up on what has happened since the last time they chatted, the release of Hallucinations, overcoming struggle to get this record out, touring and so much more.-------------------------------------------------------Find Soul Meets Body :Insta: https://www.instagram.com/soulmeetsbodybandAll other links via link tree: https://linktr.ee/soulmeetsbodyband---------------------------------------------Find Stereo Therapy :Insta: https://www.instagram.com/stereo.therapy Website: https://www.stereotherapypod.com----------------------------------------------Theme Song by Walwin

How does someone run 100-mile weeks for 300 consecutive weeks without destroying their body? Today, I'm with Andrew Glaze to unpack the recovery protocols and biological strategies that make ultra endurance possible. You'll learn actionable insights on recovery, nutrition timing, and what biomarkers actually matter when you're pushing your body to extremes.​​​​​​​​​​​​​​​​ CLICK HERE TO BECOME GARYS VIP!: ⁠https://bit.ly/4ai0Xwg⁠ Connect with Andrew Glaze Website: ⁠https://bit.ly/4jE3QfH⁠  YouTube: ⁠https://bit.ly/4sJopM0⁠  Instagram: ⁠https://bit.ly/49zjkwV⁠  TikTok: ⁠https://bit.ly/4sFjRG8⁠  X.com: ⁠https://bit.ly/49ONldn⁠  LinkedIn: ⁠https://bit.ly/45c3xmm⁠  Pre-sell to his book, "Smile, Or You're Doing it Wrong" Thank you to our partners H2TABS: “ULTIMATE10” FOR 10% OFF: ⁠https://bit.ly/4hMNdgg⁠ BODYHEALTH: “ULTIMATE20” FOR 20% OFF: ⁠http://bit.ly/4e5IjsV⁠ BAJA GOLD: "ULTIMATE10" FOR 10% OFF: ⁠https://bit.ly/3WSBqUa⁠ SNOOZE: LET'S GET TO SLEEP!: ⁠https://bit.ly/4pt1T6V⁠ COLD LIFE: THE ULTIMATE HUMAN PLUNGE: ⁠https://bit.ly/4eULUKp⁠ WHOOP: JOIN AND GET 1 FREE MONTH!: ⁠https://bit.ly/3VQ0nzW⁠ AION: “ULTIMATE10” FOR 10% OFF: ⁠https://bit.ly/4h6KHAD⁠ A-GAME: “ULTIMATE15” FOR 15% OFF: ⁠http://bit.ly/4kek1ij⁠ PEPTUAL: “TUH10” FOR 10% OFF: ⁠https://bit.ly/4mKxgcn⁠ CARAWAY: “ULTIMATE” FOR 10% OFF: ⁠https://bit.ly/3Q1VmkC⁠ HEALF: 10% OFF YOUR ORDER: ⁠https://bit.ly/41HJg6S⁠ RHO NUTRITION: “ULTIMATE15” FOR 15% OFF: ⁠https://bit.ly/44fFza0⁠ GOPUFF: GET YOUR FAVORITE SNACK!: ⁠https://bit.ly/4obIFDC⁠ GENETIC METHYLATION TEST (UK ONLY): ⁠https://bit.ly/48QJJrk⁠ GENETIC TEST (USA ONLY): ⁠⁠https://bit.ly/3Yg1Uk9⁠ Watch  the “Ultimate Human Podcast” every Tuesday & Thursday at 9AM EST: YouTube: ⁠https://bit.ly/3RPQYX8⁠ Podcasts: ⁠https://bit.ly/3RQftU0⁠ Connect with Gary Brecka Instagram: ⁠https://bit.ly/3RPpnFs⁠ TikTok: ⁠https://bit.ly/4coJ8fo⁠ X: ⁠https://bit.ly/3Opc8tf⁠ Facebook: ⁠https://bit.ly/464VA1H⁠ LinkedIn: ⁠https://bit.ly/4hH7Ri2⁠ Website: ⁠https://bit.ly/4eLDbdU⁠ Merch: ⁠https://bit.ly/4aBpOM1⁠ Newsletter: ⁠https://bit.ly/47ejrws⁠ Ask Gary: ⁠https://bit.ly/3PEAJuG⁠ Timestamps 00:00 Intro of Show 02:35 Andrew Glaze's Back Story 04:53 PTSD Recovery for First Responders 10:32 Mental Endurance for Ultra Marathons 15:26 Biohacking and Health Practices 15:59 Andrew's Supplement Routine 23:14 Moab 240 Preparations and Experience 28:28 Hallucinations and Injuries during Distance Racing 49:32 Finishing the Moab 240 Race 50:44 Balancing Running and Firefighting 54:15 Motivation in Running Marathons 1:13:37 Goal-Setting and Accomplishments 1:01:29 Prioritising Health Today 1:04:19 Recovery Protocols and Mental Toughness 1:05:13 Andrew's Next Races 1:13:38 Accomplishments and Inspiration 1:12:10 Andrew's Health Practices Disclaimer: This podcast is for informational purposes only and does not provide medical advice. It is not intended for diagnosing or treating any health condition. Always consult a licensed healthcare professional before making health or wellness decisions. Gary Brecka is the owner of Ultimate Human, LLC which operates The Ultimate Human podcast and promotes certain third-party products used by Gary Brecka in his personal health and wellness protocols and daily life and for which Ultimate Human LLC and / or Gary Brecka directly or indirectly holds an economic interest or receives compensation.  Accordingly, statements made by Gary Brecka and others (including on The Ultimate Human podcast) may be considered promotional in nature. Learn more about your ad choices. Visit megaphone.fm/adchoices

They promised us flying cars, moon colonies, and a life of leisure. Instead, we got Wi-Fi-connected salt shakers and chatbots that hallucinate fake legal precedents. Welcome to the "future," folks. It's expensive, it requires a software update, and it doesn't work. In this episode, The Critic takes a sledgehammer to the hype cycle. We dive into the "Smart Home" nightmare, asking the brave question: "Does my refrigerator really need a Twitter account?" We explore the phenomenon of AI hallucinations, where the world's smartest computers confidently gaslight us into believing nonsense. We dissect the "Infinite Loop of Garbage," where humans use AI to write emails that other humans use AI to read. And finally, we pay tribute to the "Tech Bros"—the billionaire saviors who are busy building luxury bunkers in New Zealand while claiming to save the world. Buckle up. We're about to pop the bubble. To unlock full access to all our episodes, become a premium subscriber on Apple Podcasts or Patreon. And don't forget to visit englishpluspodcast.com for more content and learning.

Cristina Cranga: When Teams Stop Testing Reality and Fall Into Decision Hallucinations Read the full Show Notes and search through the world's largest audio library on Agile and Scrum directly on the Scrum Master Toolbox Podcast website: http://bit.ly/SMTP_ShowNotes.   "Over time, what I notice is that teams stop testing reality. They optimize execution around constraints that might no longer exist." - Cristina Cranga   Cristina introduces a powerful concept she calls "decision hallucinations"—the perception of constraints and boundaries that aren't actually real or present. In her experience working with teams in complex matrix environments, she noticed a troubling pattern: team members would say things like "we can't change this because it's already decided" or "the priority comes from the top level" without ever verifying these assumptions.  The impact on team behavior was significant—teams stopped asking questions, stopped having conversations with stakeholders, and began operating within perceived limitations rather than actual ones. Cristina emphasizes that as Agile practitioners, our work isn't just about ceremonies and metrics—it's about supporting and facilitating decision processes.  When she encouraged teams to ask better questions like "Is this an assumption-based decision or an explicit shared choice?", something beautiful happened: options reappeared, conversations changed, and teams realized they were constrained by perception rather than reality. She uses the famous duck vs. rabbit optical illusion from psychology to illustrate how our brains can only see one reality at a time, making the case that we must constantly test our view of reality through continuous conversations with stakeholders.   In this episode, we refer to the work of Esko Kilpi on conversations and the duck vs rabbit image from psychology.   Self-reflection Question: When was the last time you challenged an assumption your team operates under, and what did you discover when you tested that reality?   [The Scrum Master Toolbox Podcast Recommends]

In this episode of the Crazy Wisdom podcast, host Stewart Alsop sits down with Kelvin Lwin for their second conversation exploring the fascinating intersection of AI and Buddhist cosmology. Lwin brings his unique perspective as both a technologist with deep Silicon Valley experience and a serious meditation practitioner who's spent decades studying Buddhist philosophy. Together, they examine how AI development fits into ancient spiritual prophecies, discuss the dangerous allure of LLMs as potentially "asura weapons" that can mislead users, and explore verification methods for enlightenment claims in our modern digital age. The conversation ranges from technical discussions about the need for better AI compilers and world models to profound questions about humanity's role in what Lwin sees as an inevitable technological crucible that will determine our collective spiritual evolution. For more information about Kelvin's work on attention training and AI, visit his website at alin.ai. You can also join Kelvin for live meditation sessions twice daily on Clubhouse at clubhouse.com/house/neowise.Timestamps00:00 Exploring AI and Spirituality05:56 The Quest for Enlightenment Verification11:58 AI's Impact on Spirituality and Reality17:51 The 500-Year Prophecy of Buddhism23:36 The Future of AI and Business Innovation32:15 Exploring Language and Communication34:54 Programming Languages and Human Interaction36:23 AI and the Crucible of Change39:20 World Models and Physical AI41:27 The Role of Ontologies in AI44:25 The Asura and Deva: A Battle for Supremacy48:15 The Future of Humanity and AI51:08 Persuasion and the Power of LLMs55:29 Navigating the New Age of TechnologyKey Insights1. The Rarity of Polymath AI-Spirituality Perspectives: Kelvin argues that very few people are approaching AI through spiritual frameworks because it requires being a polymath with deep knowledge across multiple domains. Most people specialize in one field, and combining AI expertise with Buddhist cosmology requires significant time, resources, and academic background that few possess.2. Traditional Enlightenment Verification vs. Modern Claims: There are established methods for verifying enlightenment claims in Buddhist traditions, including adherence to the five precepts and overcoming hell rebirth through karmic resolution. Many modern Western practitioners claiming enlightenment fail these traditional tests, often changing the criteria when they can't meet the original requirements.3. The 500-Year Buddhist Prophecy and Current Timing: We are approximately 60 years into a prophesied 500-year period where enlightenment becomes possible again. This "startup phase of Buddhism revival" coincides with technological developments like the internet and AI, which are seen as integral to this spiritual renaissance rather than obstacles to it.4. LLMs as UI Solution, Not Reasoning Engine: While LLMs have solved the user interface problem of capturing human intent, they fundamentally cannot reason or make decisions due to their token-based architecture. The technology works well enough to create illusion of capability, leading people down an asymptotic path away from true solutions.5. The Need for New Programming Paradigms: Current AI development caters too much to human cognitive limitations through familiar programming structures. True advancement requires moving beyond human-readable code toward agent-generated languages that prioritize efficiency over human comprehension, similar to how compilers already translate high-level code.6. AI as Asura Weapon in Spiritual Warfare: From Buddhist cosmological perspective, AI represents an asura (demon-realm) tool that appears helpful but is fundamentally wasteful and disruptive to human consciousness. Humanity exists as the battleground between divine and demonic forces, with AI serving as a weapon that both sides employ in this cosmic conflict.7. 2029 as Critical Convergence Point: Multiple technological and spiritual trends point toward 2029 as when various systems will reach breaking points, forcing humanity to either transcend current limitations or be consumed by them. This timing aligns with both technological development curves and spiritual prophecies about transformation periods.

Season 63, Episodes 83-87, Spoiler Level CFS (Crazy ** Spoilers) Willow's verdict is in and Kathy isn't happy about it, Anna tries to escape but her situation gets more awful, Dante and Chase discuss how to approach this case without bias, and Olivia threatens to shoot Martin and Drew (and it was amazing). In Fashion, Stacy and Kathy disagree on Nina's blue dress. And in Musings they discuss all the romance (loved it), how the WSB is full of bad guys, Britt's former boyfriend being a lumberjack, and how Jacinda is awesome for loving video games. Thank you for listening to our General Hospital podcast. If you enjoyed it, please subscribe and tell your friends. Drop us a review. And let us know your own musings and theories and fashion notes. Reach Stacy at Alexis@areweghing.com and Kathy at Felicia@areweghing.com. For more information, please visit us at www.areweghing.com Recorded 1-17-26, Music by Grammy award winning Alex Robinson https://www.musicbyalexrobinson.com/ and logo by the equally as amazing Jakob Evans.

Amanda Kahlow, CEO and founder of 1Mind, joins Amir to break down what AI changes in modern sales and go to market, and what it does not. If you lead revenue, product, or growth, this is a practical look at where AI creates leverage today, where humans still matter, and how teams actually adopt it without chaos.Amanda shares how “go to market superhumans” can handle everything from early buyer conversations to demos, sales engineering support, and customer success. They also dig into trust, hallucinations, and why the bar for AI feels higher than the bar for people.Key takeaways• Most buyers want answers early, without the pressure that comes with talking to a salesperson• AI can remove friction by turning static content into a two way conversation that helps buyers move faster• The hardest part of adoption is not capability, it is change management and trust inside the team• Humans still shine in relationship and nuance, but AI can outperform on recall, depth, and real time access to the right info• As AI levels the selling experience, product quality matters more, and the best product has a clearer path to winTimestamped highlights00:31 What 1Mind builds, and what “go to market superhumans” actually do across the full buyer journey02:00 The buyer lens, why early conversations matter, and how AI gives control back to the buyer06:14 Why the SDR experience is frustrating for buyers, and where AI can improve both sides09:42 Change management in the real world, why “everyone build an agent” gets messy fast13:04 Why “swivel chair” AI fails, and what real time help should look like in live conversations15:52 Hallucinations and trust, plus the blunt question every leader should ask about human error22:26 Competitive advantage today, and why adoption eventually pushes markets toward “best product wins”A line worth sharing“Do your humans hallucinate, and how often do they do it?”Pro tips you can use this week• Start with low stakes usage, bring AI into calls quietly, then ask it for a summary and what you missed• Build adoption top down, define what good looks like, otherwise you get a pile of similar agents and no clarity• Focus AI on what it does best first, recall, context, and instant answers, then expand into workflow and process laterCall to actionIf this episode sparked ideas for your sales team or your product led funnel, follow the show so you do not miss the next one. Share it with one revenue leader who is trying to modernize their go to market motion, and connect with Amir on LinkedIn for more clips and operator level takes.

Amir (Co-Founder at Humblytics) shares how he builds an “AI-native” company by focusing less on shiny tools and more on change management: assessing AI fluency across roles, setting the right success metrics, and creating shared context so AI can reliably ship work. The big theme is convergence—engineering, product, and design are collapsing into tighter loops thanks to tools like Cursor, MCP connectors, and Figma Make. Amir demos workflows like: AI-generated context files + auto-updated documentation, scraping customer domains to infer ICPs, turning screenshots into layered Figma designs, then converting Figma to working React code in minutes, and even running an “AI co-founder” Slack bot that files Linear tickets and can hand work to agents.Timestamps0:00 Introduction0:06 Amir's stance: “no AI experts” — it's constant learning in a fast-changing field.1:59 Cursor as the unlock: not just coding, but PM/strategy/design work via MCPs.4:17 The real problem: AI adoption is mostly change management + fluency assessment.5:18 The AI fluency rubric (helper → automator → augmentor → agentic) and why it matters.8:13 Cursor analytics: measuring AI-generated code and usage across the team.9:24 “New code is ~99% AI-generated” + how they keep quality via tight review + incremental changes.10:58 Docs workflow: GitBook connected to repo → AI edits docs and pushes live fast.14:02 ICP building: export Stripe customers → scrape domains with Firecrawl → cluster personas.17:45 Hallucination in the wild: AI misclassifies a company; human correction loop matters.34:43 Wild move: they often design in code and use an AI-generated style guide to stay consistent.38:10 Best demo: screenshot → Figma Make → layered design → Figma MCP → React code in minutes.45:29 “AI co-founder” Slack bot (Pixel): turns a bug report into a Linear ticket and can hand off to agents.48:46 Amir's wish list: we “solved dev”; now we need Cursor for marketing/sales → path to $1M ARR.Tools & technologies mentionedCursor — AI-first IDE used for coding and product/design/strategy workflows; includes team analytics.MCP (Model Context Protocol) — “connector” layer (Anthropic-origin) that lets LLMs interface with external tools/services.ChatGPT — used as a common baseline tool; discussed in the context of prompting practices and workflows.Microsoft Copilot — referenced via the law firm incentive story; used as an example of “usage metrics” gone wrong.Anthropic (AI fluency framework) — inspiration source for the helper/automator/augmentor/agentic rubric.GitBook — documentation platform connected to the repo so docs can be updated and published quickly.Firecrawl (MCP) — agentic web scraper used to analyze customer domains and infer ICP/personas.Stripe — source of customer export data (domains) to build ICP clustering.Figma — design collaboration tool; used here with Make + MCP to move from design → code.Figma Make — feature to recreate UI from an image/screenshot into editable, layered designs.Figma MCP — connector that allows Cursor/LLMs to pull Figma components/designs and generate code.React — front-end framework used in the demo for generating functional UI components.Supabase — mentioned as part of a sample stack when generating a PRD.React Router — mentioned as part of the sample stack in PRD generation.Slack — where Amir runs internal agents (including the “AI co-founder” bot).Linear — project management tool used for creating tickets from Slack/agent workflows.CI/CD — their deployment/review pipeline; emphasized as the human accountability layer.Subscribe at⁠ thisnewway.com⁠ to get the step-by-step playbooks, tools, and workflows.

LinksGround loop (electricity) - WikipediaHallucination or Confabulation? Neuroanatomy as metaphor in Large Language Models - PMCGraze: Custom Feed Builder for Bluesky · GrazeServerless Statusphere: a walk through building serverless ATProto applications on Cloudflare's Developer PlatformAT ProtocolHow to Transfer Your ChatGPT MemoryWorks for ChatGPT Data - Google DriveStewart Lee - WikipediaStewart Lee's Comedy Vehicle - WikipediaStewart Lee vs The Internet - Stewart Lee's Comedy Vehicle - Series 3 Episode 1 Preview - BBC - YouTubeEnglish Arrested for Being English? Stewart Lee's Humorous Take on British Culture | TikTokStewart Lee S4E6 - Childhood - YouTubeStewart Lee S3E3 - Satire - YouTubeStewart Lee - [1/2] Give It To Me Straight, Like Pear Cider That's Made From 100% Pears - YouTubeStewart Lee - [2/2] Give It To Me Straight, Like Pear Cider That's Made From 100% Pears - YouTubeStewart Lee On The Challenge Of Stand-Up - YouTubeComedian Stewart Lee on why he won't tour Trump's America - YouTubeStewart Lee in Conversation with Alan Moore - YouTubeJerry Springer: The Opera - WikipediaTaskmaster Origins - Rare Footage of the First Ever Show! (Edinburgh 2010) - YouTube

“He wouldn't take no for an answer” - On Valentine's Day, 2020, cops respond to a 1 a.m. call in the Hollywood Hills, where they discover 38-year-old family therapist Amie Harwick fighting for her life, and her roommate Michael Herman, with blood on his hands. Michael is brought in for questioning but swears had nothing to do with it, and that the perpetrator must have fled into the night. What cops don't know is that the man they have in custody is telling the truth, and that the real attacker is still out there. But, Amie's best friend, Robert Coshland, is on the case: he's determined to take down the man who's been stalking Amie for years, & finally get justice.  - For more information about Justice 4 Amie, please see here: https://www.change.org/p/adam-b-schiff-justice-4-amie-domestic-violence-laws-updated  - Credits:Directed, written & edited by Matthew Rice Researched by Tiffany Loxton Co-written by Kat Gardilcic & Tiffany Loxton Voiceover by William Akana Produced by Salim Sader “The Final Hours of Amie Harwick.” 48 Hours: CBS Broadcasting Inc., 2022. (CBS News) “Justice for Amie Harwick.” 48 Hours: CBS Broadcasting Inc., 2024. (CBS News) “Celebrity Sexpert.” Death By Fame: Warner Bros. Discovery Inc., 2023 (AMPLE Entertainment) “Love, Death, and Obsession In Hollywood.” A Plan to Kill: Oxygen Media LLC, 2024 (Peacock) Man Walks Outside To Find Decapitated Roommate, Dr. Insanity, 2024. Nightline, ABC News,  CBS Mornings, CBS Broadcasting Inc. Inside Edition, CBS Broadcasting Inc. Getty Images Learn more about your ad choices. Visit megaphone.fm/adchoices

Check out host Bidemi Ologunde's new show: The Work Ethic Podcast, available on Spotify and Apple Podcasts.In this episode, host Bidemi Ologunde unpacks OpenAI's newly released ChatGPT Health and what it signals about the future of consumer-facing healthcare AI. What exactly is "ChatGPT Health," and why is OpenAI moving from general chat to a dedicated health experience? When an AI gives the wrong answer in a high-stakes setting—medical advice, airline refunds, legal citations—who owns the liability: the user, the company deploying the chatbot, or the model-maker? How are regulators in the U.S., Europe, and beyond approaching AI in healthcare—and what counts as "wellness" versus "medical" software? Bidemi also explores the realities of AI error, hallucinations, and bias, and asks what these tools could mean for underserved and minority populations worldwide— including Native Americans, Pacific Islanders, and communities in low-resource health systems.Email: bidemiologunde@gmail.comSupport for The Bid Picture Podcast comes from Intuit QuickBooks. If you're running a business, a side hustle, or just trying to stay on top of your money, QuickBooks helps you track income and expenses, send invoices, and see where things stand—without living in spreadsheets. It's tech that's meant to give you time back, so you can spend more of your attention on your life, not your tabs. If you're asked how you heard about QuickBooks, please mention The Bid Picture Podcast. Learn more at quickbooks.intuit.com.Support for The Bid Picture Podcast comes from VIZZ. If age-related blurry near vision—also called presbyopia—has you holding your phone farther away or avoiding the small print, ask your eye doctor about VIZZ, a once-daily prescription eye drop for adults that treats blurry near vision. Do not use VIZZ if you are allergic to any of its ingredients. The most common side effects are eye irritation, temporary dim or dark vision, headache, and eye redness. Be careful driving at night or doing activities that require clear vision until your vision returns to normal. If you're asked how you heard about VIZZ, please mention The Bid Picture Podcast. Learn more at vizz.com.Support for The Bid Picture Podcast comes from Rula. If you're trying to build a healthier relationship with tech—setting boundaries, breaking burnout patterns, or feeling more present—therapy can help, and Rula makes it easier to find licensed mental health providers and meet by video on a schedule that fits your life. If you're asked how you heard about Rula, please mention The Bid Picture Podcast. Learn more at rula.com.Support the show

Imagine sitting at home and then all of a sudden you hear a men's choir belting out “The Star Spangled Banner.” You check your phone, computer, radio. Nothing's playing. You look outside, no one's there. That's what happened to neurologist Bruce Dobkin after he received a cochlear implant. He set out to learn everything he could about the condition, called musical hallucinosis.In a story from August, Host Ira Flatow talks with Dobkin about his decision to publish his account in a medical journal and why the condition is more common than he realized.Guest: Dr. Bruce Dobkin is a neurologist at UCLA Health.Transcript is available at sciencefriday.com. Subscribe to this podcast. Plus, to stay updated on all things science, sign up for Science Friday's newsletters.