POPULARITY
How do we figure out whether interpretability is doing its job? One way is to see if it helps us prove things about models that we care about knowing. In this episode, I speak with Jason Gross about his agenda to benchmark interpretability in this way, and his exploration of the intersection of proofs and modern machine learning. Patreon: https://www.patreon.com/axrpodcast Ko-fi: https://ko-fi.com/axrpodcast Transcript: https://axrp.net/episode/2025/03/28/episode-40-jason-gross-compact-proofs-interpretability.html Topics we discuss, and timestamps: 0:00:40 - Why compact proofs 0:07:25 - Compact Proofs of Model Performance via Mechanistic Interpretability 0:14:19 - What compact proofs look like 0:32:43 - Structureless noise, and why proofs 0:48:23 - What we've learned about compact proofs in general 0:59:02 - Generalizing 'symmetry' 1:11:24 - Grading mechanistic interpretability 1:43:34 - What helps compact proofs 1:51:08 - The limits of compact proofs 2:07:33 - Guaranteed safe AI, and AI for guaranteed safety 2:27:44 - Jason and Rajashree's start-up 2:34:19 - Following Jason's work Links to Jason: Github: https://github.com/jasongross Website: https://jasongross.github.io Alignment Forum: https://www.alignmentforum.org/users/jason-gross Links to work we discuss: Compact Proofs of Model Performance via Mechanistic Interpretability: https://arxiv.org/abs/2406.11779 Unifying and Verifying Mechanistic Interpretability: A Case Study with Group Operations: https://arxiv.org/abs/2410.07476 Modular addition without black-boxes: Compressing explanations of MLPs that compute numerical integration: https://arxiv.org/abs/2412.03773 Stage-Wise Model Diffing: https://transformer-circuits.pub/2024/model-diffing/index.html Causal Scrubbing: a method for rigorously testing interpretability hypotheses: https://www.lesswrong.com/posts/JvZhhzycHu2Yd57RN/causal-scrubbing-a-method-for-rigorously-testing Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition (aka the Apollo paper on APD): https://arxiv.org/abs/2501.14926 Towards Guaranteed Safe AI: https://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-45.pdf Episode art by Hamish Doodles: hamishdoodles.com
Welcome to Episode 3 of Structureless conversation "Jet Sweep" where Zac is joined by Steezy A Smith, an up and coming sports media personality out of Seattle. Steezy talks how he got his nickname, carving his place out in the sports media landscape, building an online presence, hop topics in the sporting world, fantasy picks and MUCH MORE
Each environment can be beneficial to learn about --- Support this podcast: https://podcasters.spotify.com/pod/show/change-griot/support
Welcome to episode two of Structureless Conversation presented by The No Structure Podcast. In this episode Zac is joined by Kwabi of the peace bus. Kwabi has devoted his life to philanthropical work and spreading peace amongst the human race. Kwabi joins Zac to talk the origins of his peace bus, being inspired, finding his life's purpose, traveling the country, his upcoming work in Africa, his kids show pilot and SO MUCH MORE. A huge thank you to Kwabi for coming on the podcast. Below you can find links to get connected to Kwabi and his peace movement: Instagram: https://instagram.com/thepeacebus?igshid=MzRlODBiNWFlZA== Website: https://www.thepeacebus.org Tik Tok: https://www.tiktok.com/@kwabiamoahforson?_t=8eM5u6RGdJ8&_r=1 On the podcast page you'll also find a playlist collection of some interviews Kwabi has done over the years and a link to his kids show pilot!!! Make sure to follow the podcast as well below: Website: https://www.thenostructurepodcast.com All things No Structure: https://linktr.ee/Nostructurepodcast?fbclid=PAAaZh8aozLZL6Nz_hF9HNFG5iknxYPACzu_QvI4qreYy1cv6XPRsCs1sn
Zac is joined by Tor for the first episode of "Structureless Conversation". Were talking Tors upbringing, norweigan influence, being a creative, new skills, his live show, new EP, and Much more. Episode includes preview of his new EP!! . To check out the visuals head here: FULL CIRCLE WITH TORBJORN | STRUCTURELESS CONVERSATION EP1 | THE NO STRUCTURE PODCAST #podcast https://youtu.be/dBZ_EZ0gKVg . Check out his new EP "Nordwest EP" dropping May 26th! . For all things Tor visit: https://www.beatsbytor.com/?fbclid=PAAaa2josTqDZ2J7nh6PBvG0NBZUThypkOVmmV2mbBbzyrfgXzGlZ6MitE9bY
The twins continue to host their guest, Jeff Trautman, to talk about how and why coaching relationships need structure. They wrestle with a host of issues: building trust, overcoming resistance, delivering bad news early, not "sandwiching" the client, and emotionally charged communication.
Structureless without Flaco and Will's humor, Johnny, Mooney, Mike and Leo tackle sensitive and controversial issues. The fella's discuss gas prices, politicians and Yoni shouts out to the good parents. The guys also debate the Supreme Court's ruling on Roe vs Wade which gets out of control. --- This episode is sponsored by · Anchor: The easiest way to make a podcast. https://anchor.fm/app Support this podcast: https://anchor.fm/gphomiespodcast/support
Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: The Hero With A Thousand Chances , published by Eliezer Yudkowsky on the AI Alignment Forum. "Allow me to make sure I have this straight," the hero said. "I've been untimely ripped from my home world to fight unspeakable horrors, and you say I'm here because I'm lucky?" Aerhien dipped her eyelashes in elegant acknowledgment; and quietly to herself, she thought: Thirty-seven. Thirty-seven heroes who'd said just that, more or less, on arrival. Not a sign of the thought showed on her outward face, where the hero could see, or the other council members of the Eerionnath take note. Over the centuries since her accidental immortality she'd built a reputation for serenity, more or less because it seemed to be expected. "There are kinds and kinds of luck," Aerhien said serenely. "Not every person desires their personal happiness above all else. Those who are lucky in aiding others, those whose luck is great in succor and in rescue, these ones are not always happy themselves. You are here, hero, because you have a hero's luck. The boy whose dusty heirloom sword proves to be magical. The peasant girl who finds herself the heir to a great kingdom. Those who discover, in time of sudden stress, an untrained wild magic within themselves. Success born not of learning, not of skill, not of determination, but unplanned coincidence and fortunes of birth: That is a hero's luck." "Gosh," said the hero after a long, awkward pause, "thanks for the compliment." "It is not a compliment," Aerhien said, "but this is: that you have taken good advantage of your luck. Our enemy does not speak, we do not know if there is any aliveness in it to think; but it learns, or seems to learn. We have never won against it using the same trick twice. It is rare now that a hero succeeds in conceiving a genuinely new trick, for we have fought this shadow long under our sun. For this reason we have taken to summoning heroes from distant dimensions with other modes of thought; sometimes one such knows a truly new technique, and at least they fight differently. But far more often, hero, the hero wins by luck." "Huh," said the hero. He frowned; more in thought, it seemed, than in displeasure. "How... very odd. I wonder why that is. What kind of enemy can be defeated only by luck?" "A nameless enemy and null," said Aerhien. "Structureless and empty, horrible and dark, the most terrifying thing imaginable: We call it Dust. That seems to be its only desire, to tear down every bit of structure in the world, grind it into specks of perfect chaos. Always the Dust is defeated, always it takes a new shape immune to its last defeat." "I wonder," murmured the hero, "if it will run out of shapes, and then end; or if it will finally become invincible." (One of the other Eerionnath shuddered.) "I do not know," Aerhien said simply. "I do not know the nature of the Dust, nor the nature of the Counter-Force that opposes it. The Dust is terrible and our world should long since have ended. We are not fools enough to believe we could be lucky so many times by chance alone. But the Counter-Force has never acted openly; it never reveals itself except in - a hero's luck. And so we, the council Eerionnath to prevent the world from destruction, are at your disposal to command; and all the power and resource that this world holds, for your battle." And she, Aerhien, and the council Eerionnath, bowed low. Then they waited to see if the hero would demand dominions or slaves as payment, before condescending to rescue a people in distress. If so they would dispose of him, and summon another. This one, though, seemed to have at least some qualities of a true hero; his face showed no avarice, only an abstracted puzzlement. "A hidden Counter-Force..." he murmured. "Excuse me, but this is all very vague. Can you give me a specific example of a h...
In this episode, the boys hit the studio with absolutely zero plan, no topics, and limited energy. Despite this, it’s a banger. New starches, NBA Finals, fantasy football, and much much more. Follow the podcast on Insta for clips and highlights: @2boys1cast. Each episode is available on literally every platform you could possibly think of (except youtube, but I mean come on).
Ahoy babes! Nicole moved this week, so we have a fly by the seat of our pants episode today, complete with food faves from the Middle Ages, birds on birds on birds action, and roosters that have been tied to bigs Human Centipede style! Plus, what's the deal with crepes?!? So turn on re-runs of Star Trek and shove a tiny bird inside an olive, it's Life's a Banquet the podcast!Heritage Radio Network is a listener supported nonprofit podcast network. Support Life's A Banquet by becoming a member!Life's A Banquet is Powered by Simplecast.
00:48 - Ben’s Superpower: Making Arancini (https://en.wikipedia.org/wiki/Arancini) and Reading A Lot of Books 02:00 - Starting Local Welcome (https://www.localwelcome.org/) and Helping Refugees Death of Alan Kurdi (https://en.wikipedia.org/wiki/Death_of_Alan_Kurdi) 09:37 - Humanization, Cognitive Biases, and Heuristics Contact Hypothesis (https://en.wikipedia.org/wiki/Contact_hypothesis) Social Constructionism (https://en.wikipedia.org/wiki/Social_constructionism) Social Conformity - Brain Games (https://www.youtube.com/watch?v=o8BkzvP19v4) In-group Bias (https://en.wikipedia.org/wiki/In-group_favoritism) Thinking, Fast and Slow by Daniel Kahneman (https://www.amazon.com/gp/product/0374533555/ref=as_li_qf_asin_il_tl?ie=UTF8&tag=therubyrep-20&creative=9325&linkCode=as2&creativeASIN=0374533555&linkId=8d08e4ff6f8ed87bf7b29239465ef9da) Rules for Radicals: A Practical Primer for Realistic Radicals by Saul D. Alinsky (https://www.amazon.com/gp/product/0679721134/ref=as_li_qf_asin_il_tl?ie=UTF8&tag=therubyrep-20&creative=9325&linkCode=as2&creativeASIN=0679721134&linkId=fa86d3e8a610bb665e32c57e718190e1) 21:25 - Empathy and Compassion; Humans Thriving Together The Compassionate Mind Foundation (https://compassionatemind.co.uk/) The Uninhabitable Earth: Life After Warming by David Wallace-Wells (https://www.amazon.com/gp/product/0525576703/ref=as_li_qf_asin_il_tl?ie=UTF8&tag=therubyrep-20&creative=9325&linkCode=as2&creativeASIN=0525576703&linkId=64c8492e8856e0c450e2570ffe8b6353) Doughnut Economics: Seven Ways to Think Like a 21st-Century Economist by Kate Raworth (https://www.amazon.com/gp/product/1603587969/ref=as_li_qf_asin_il_tl?ie=UTF8&tag=therubyrep-20&creative=9325&linkCode=as2&creativeASIN=1603587969&linkId=74466c30ad41ea179c4b86feb49665af) The 36 Questions That Lead to Love (https://www.nytimes.com/2015/01/11/fashion/no-37-big-wedding-or-small.html) 31:26 - Measuring Success Conversion Financial Resilience Social Contact Hours Net Promoter Score (NPS) (https://en.wikipedia.org/wiki/Net_Promoter) Language Ability 41:56 - Getting People to Connect with Refugees in a Personal Way Lump of Labour Fallacy (https://en.wikipedia.org/wiki/Lump_of_labour_fallacy) “We come together through our sameness and we grow through our differences.” ~ Virginia Satir Reflections: Rein: The idea of communicative praxis. The Self after Postmodernity by Calvin O. Schrag (https://www.amazon.com/gp/product/0300078765/ref=as_li_qf_asin_il_tl?ie=UTF8&tag=therubyrep-20&creative=9325&linkCode=as2&creativeASIN=0300078765&linkId=5527025502716e825c8a9e183aeaf558) Sam: The commodification of trust. Also, The Tyranny of Structureless (https://www.jofreeman.com/joreen/tyranny.htm): In any organization there are both explicit and implicit power structures. Changing the implicit ones. Ben: Having a safe space to reflect and have conversations. This episode was brought to you by @therubyrep (https://twitter.com/therubyrep) of DevReps, LLC (http://www.devreps.com/). To pledge your support and to join our awesome Slack community, visit patreon.com/greaterthancode (https://www.patreon.com/greaterthancode) To make a one-time donation so that we can continue to bring you more content and transcripts like this, please do so at paypal.me/devreps (https://www.paypal.me/devreps). You will also get an invitation to our Slack community this way as well. Amazon links may be affiliate links, which means you’re supporting the show when you purchase our recommendations. Thanks! Special Guest: Ben Pollard.
Oh boy. The first episode. I get why people do pilots of shows now. We learnt so many things after this one. This episode's mistakes include:Terrible audio quality. Yeesh. Sorry. So many creaks and clinks. Nick's too loud. Ben's too quiet. We know. Sorry.Structureless meandering. We were figuring it out as we went along. Rambling intro, abrupt ending.None of the hosts ever looked at the camera.Egregious factual inaccuracies.By almost every metric, this episode should have never been released. Enjoy!
The second installment of Slow Dancing Guys, where we witness an alpaca pee for five minutes straight before discussing all manner of meaningless things. We try to psycho-analyse a folder of Finn's assorted... um... works. We also have Down (Find Your Way), an original song by Finn.
The first installment of our podcast in which we discuss annoying friends, the Irish potato famine and giving constructive criticism, as well as a story about Timothy with Mitch and Callum and everybody joining together to play daft Punk's "Get Lucky". We're sorry about us eating the whole way through and I am especially sorry about Finn's "Surprise".