POPULARITY
Een nieuwe #nerdland podcast! Met deze maand: Alzheimervaccin! Three Mile Island! Starlinersaga! Een komeet! IgNobelprijzen! Bliksems! Strawberry! Zeilschepen! En veel meer... Shownotes: https://maandoverzicht.nerdland.be/nerdland-maandoverzicht-oktober-2024/ Gepresenteerd door Lieven Scheire met Jeroen Baert, Els Aerts, Bart Van Peer, Marian Verhelst en Hetty Helsmoortel. Montage en mixing door Els Aerts en Jens Paeyeneers. (00:00:00) Intro (00:01:09) Er wordt gewerkt aan een vaccin tegen Alzheimer (00:10:02) Mogelijk een komeet te zien Tsuchinshan-ATLAS (00:17:53) Microsoft wil Three Mile Island heropenen (00:30:43) AI suggereert kledingstijl (00:35:24) Chatmodel Strawberry van OpenAI bouwt redeneringen op (00:50:41) ChatGPT brengt complotdenkers op andere ideetjes (00:59:54) Google Notebook maakt podcast uit jouw input (01:05:43) Pieter Abbeel gaat bij Amazon werken (01:10:50) Vrachtschip met zeilen dat Ariane 6 vervoert aangemeerd in Waaslandhaven (01:17:16) Man laat AI-bands fake muziek maken (01:22:07) Doe alsof je miljoenen fans hebt op SocialAI (01:24:59) AI camera die uw kak filmt (01:31:11) 19 mensen tegelijk in de ruimte (01:33:58) Voyager 1 heeft weer gevuurd (01:41:24) STARLINER SAGA (01:42:39) Mogelijk wordt het project geschrapt (01:44:15) Er wordt iemand gestuurd van de Space Force (01:46:38) SILICON VALLEY NIEUWS (01:46:54) Polaris Dawn space walk (01:54:39) Robotaxi Tesla wordt voorgesteld op 10 oktober (01:58:02) IgNobelprijzen uitgereikt (02:01:17) Placebos reduce anxiety (02:07:58) Het bliksemt vaker op donderdag (02:10:37) Nieuwe exportrestricties chips naar China (02:16:59) Ontslagen bij Intel (02:20:59) Lancering van Europa Clipper kan vanaf 10 okt (02:30:44) Aankondigingen (02:30:51) 27 december, Lotto Arena: Nerdland voor Kleine Nerds (02:36:13) Hetty missie 2024 (02:37:30) Scheurkalender 2025 (02:41:45) Code Rood en AI boek in EN (02:42:03) Sponsor Fairy Positron
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: Paper in Science: Managing extreme AI risks amid rapid progress, published by JanB on May 23, 2024 on The AI Alignment Forum. https://www.science.org/doi/10.1126/science.adn0117 Authors: Yoshua Bengio, Geoffrey Hinton, Andrew Yao, Dawn Song, Pieter Abbeel, Yuval Noah Harari, Ya-Qin Zhang, Lan Xue, Shai Shalev-Shwartz, Gillian Hadfield, Jeff Clune, Tegan Maharaj, Frank Hutter, Atılım Güneş Baydin, Sheila McIlraith, Qiqi Gao, Ashwin Acharya, David Krueger, Anca Dragan, Philip Torr, Stuart Russell, Daniel Kahneman, Jan Brauner*, Sören Mindermann* Abstract: Artificial intelligence (AI) is progressing rapidly, and companies are shifting their focus to developing generalist AI systems that can autonomously act and pursue goals. Increases in capabilities and autonomy may soon massively amplify AI's impact, with risks that include large-scale social harms, malicious uses, and an irreversible loss of human control over autonomous AI systems. Although researchers have warned of extreme risks from AI, there is a lack of consensus about how to manage them. Society's response, despite promising first steps, is incommensurate with the possibility of rapid, transformative progress that is expected by many experts. AI safety research is lagging. Present governance initiatives lack the mechanisms and institutions to prevent misuse and recklessness and barely address autonomous systems. Drawing on lessons learned from other safety-critical technologies, we outline a comprehensive plan that combines technical research and development with proactive, adaptive governance mechanisms for a more commensurate preparation. Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
Covariant's RFM-1: Jon Krohn explores the future of AI-driven robotics with RFM-1, a groundbreaking robot arm designed by Covariant and discussed by A.I. roboticist Pieter Abbeel. Explore how this innovation aims to merge digital intelligence with the physical world, promising a new era of efficiency and autonomy. Additional materials: www.superdatascience.com/774 Interested in sponsoring a SuperDataScience Podcast episode? Visit passionfroot.me/superdatascience for sponsorship information.
Chat GPT betekende een doorbraak in het digitale tijdperk. Maar dat gaat over tekst. Zodra we AI kunnen trainen op video en die AI in robots kunnen verwerken, staat ons pas echt een revolutie te wachten, ook in de echte wereld. Tenminste, dat zegt Pieter Abbeel. Voorts hebben we het over waarom urine geel is, wat we moeten onthouden van de technologiebeurs in Las Vegas én waarom het glas van de toekomst hout is. Transparant hout, jawel. See omnystudio.com/listener for privacy information.
MLOps Coffee Sessions #174 with Evaluation Panel, Amrutha Gujjar, Josh Tobin, and Sohini Roy hosted by Abi Aryan. We are now accepting talk proposals for our next LLM in Production virtual conference on October 3rd. Apply to speak here: https://go.mlops.community/NSAX1O // Abstract Language models are very complex thus introducing several challenges in interpretability. The large amounts of data required to train these black-box language models make it even harder to understand why a language model generates a particular output. In the past, transformer models were typically evaluated using perplexity, BLEU score, or human evaluation. However, LLMs amplify the problem even further due to their generative nature thus making them further susceptible to hallucinations and factual inaccuracies. Thus, evaluation becomes an important concern. // Bio Abi Aryan Machine Learning Engineer @ Independent Consultant Abi is a machine learning engineer and an independent consultant with over 7 years of experience in the industry using ML research and adapting it to solve real-world engineering challenges for businesses for a wide range of companies ranging from e-commerce, insurance, education and media & entertainment where she is responsible for machine learning infrastructure design and model development, integration and deployment at scale for data analysis, computer vision, audio-speech synthesis as well as natural language processing. She is also currently writing and working in autonomous agents and evaluation frameworks for large language models as a researcher at Bolkay. Amrutha Gujjar CEO & Co-Founder @ Structured Amrutha Gujjar is a senior software engineer and CEO and co-founder of Structured, based in New York. With a Bachelor of Science in Computer Science from the University of Washington's Allen School of CSE, she brings expertise in software development and leadership to my work. Connect with Amrutha on LinkedIn to learn more about her experience and discuss exciting opportunities in software development and leadership. Josh Tobin Founder @ GantryJosh Tobin is the founder and CEO of Gantry. Previously, Josh worked as a deep learning & robotics researcher at OpenAI and as a management consultant at McKinsey. He is also the creator of Full Stack Deep Learning (fullstackdeeplearning.com), the first course focused on the emerging engineering discipline of production machine learning. Josh did his PhD in Computer Science at UC Berkeley advised by Pieter Abbeel. Sohini Roy Senior Developer Relations Manager @ NVIDIASohini Bianka Roy is a senior developer relations manager at NVIDIA, working within the Enterprise Product group. With a passion for the intersection of machine learning and operations, Sohini specializes in the domains of MLOps and LLMOps. With her extensive experience in the field, she plays a crucial role in bridging the gap between developers and enterprise customers, ensuring smooth integration and deployment of NVIDIA's cutting-edge technologies. // MLOps Jobs board https://mlops.pallet.xyz/jobs // MLOps Swag/Merch https://mlops-community.myshopify.com/ // Related Links --------------- ✌️Connect With Us ✌️ ------------- Join our slack community: https://go.mlops.community/slack Follow us on Twitter: @mlopscommunity Sign up for the next meetup: https://go.mlops.community/register Catch all episodes, blogs, newsletters, and more: https://mlops.community/ Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/ Connect with Abi on LinkedIn: https://www.linkedin.com/in/abiaryan/ Connect with Amrutha on LinkedIn: https://www.linkedin.com/in/amruthagujjar/ Connect with Josh on LinkedIn: https://www.linkedin.com/in/josh-tobin-4b3b10a9/ Connect with Sohini on Twitter: https://twitter.com/biankaroy_
Jitendra Malik, Professor of EECS at UC Berkeley discusses with host Pieter Abbeel building AI from the ground-up and sensorimotor before language. Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter at @pabbeel. Hosted on Acast. See acast.com/privacy for more information.
John Schulman, co-founder OpenAI, discusses with host Pieter Abbeel the invention, capabilities, and limitations of ChatGPT. Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter at @pabbeel. Hosted on Acast. See acast.com/privacy for more information.
Yaniv Altshuler, MIT Media Lab researcher, joins Pieter Abbeel to discuss reducing cow methane emissions with AI and more. Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter at @pabbeel. Hosted on Acast. See acast.com/privacy for more information.
Woody Hoburg, a member of NASA's Expedition 69 crew, joined Pieter Abbeel to discuss his life and work currently aboard the International Space Station.Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter at @pabbeel. Hosted on Acast. See acast.com/privacy for more information.
Jesse Levinson, co-founder and CTO of Zoox joins Pieter Abbeel to discuss reinventing personal transportation from the ground up. Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter at @pabbeel. Hosted on Acast. See acast.com/privacy for more information.
Noam Brown joins host Pieter Abbeel to discuss solving poker and Diplomacy with AI. Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter @therobotbrains. Hosted on Acast. See acast.com/privacy for more information.
Stephen Balaban, CEO and founder of Lambda joins host Pieter Abbeel to discuss building the most cost-effective AI cloud.Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter @therobotbrains. Hosted on Acast. See acast.com/privacy for more information.
Dit is een bonusaflevering bij episode 4 van De Aionauten, waarin De Tijd-journalist Roan Van Eyck AI-meesterbrein Pieter Abbeel interviewt. Abbeel doceert artificiële intelligentie aan Berkeley University in Californië en sleutelde in de begindagen ook mee aan ChatGPT. Hij was de keynote spreker op New Insights, een event van De Tijd en L'Echo.Alle vragen waar we na onze vier afleveringen nog mee zitten, kunnen we dus nu aan hem stellen. Wil je nog meer te weten komen over all things AI? Lees hier het dossier van De Tijd: www.tijd.be/AI CREDITS Redactie en productie: Roan Van Eyck Mixing: Rudi WynantsSee omnystudio.com/listener for privacy information.
Rocky Duan, CTO of Covariant joins joins Host Pieter Abbeel.Subscribe to the Robot Brains Podcast today | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast and Twitter @therobotbrains. Hosted on Acast. See acast.com/privacy for more information.
Wat zit er in De 7 vandaag? België gaat misschien toch wat mínder verdienen aan de geplande multinationalbelasting dan gedacht. Dat zegt het Rekenhof. Hoe veel minder? En is die belasting eigenlijk wel een goed idee?'Sell in May and go away'. Da's een oude beurswijsheid. Maar is ze dit jaar uitgekomen? We blikken terug met onze markten-watcher.En "Artificiële intelligentie is geen hype die weer zal wegdeemsteren. En het tempo waarin innovaties elkaar opvolgen zal alleen maar versnellen." Dat zegt Belgisch AI-superbrein Pieter Abbeel. Je hoort hem zo dadelijk. Host: Bert RymenProductie: Roan Van Eyck Genoten van De 7?Check dan onze nieuwste podcastreeks 'Wat dit land nodig heeft'In vijf afleveringen neemt host Bert Rymen je mee naar de op zoektocht naar de oplossingen die dit land nodig heeft om er weer te staan. Of het nu gaat over werk, onderwijs of gezondheidszorg, het komt allemaal aan bod.See omnystudio.com/listener for privacy information.
Jakob Foerster on Multi-Agent learning, Cooperation vs Competition, Emergent Communication, Zero-shot coordination, Opponent Shaping, agents for Hanabi and Prisoner's Dilemma, and more. Jakob Foerster is an Associate Professor at University of Oxford. Featured References Learning with Opponent-Learning Awareness Jakob N. Foerster, Richard Y. Chen, Maruan Al-Shedivat, Shimon Whiteson, Pieter Abbeel, Igor Mordatch Model-Free Opponent Shaping Chris Lu, Timon Willi, Christian Schroeder de Witt, Jakob Foerster Off-Belief Learning Hengyuan Hu, Adam Lerer, Brandon Cui, David Wu, Luis Pineda, Noam Brown, Jakob Foerster Learning to Communicate with Deep Multi-Agent Reinforcement Learning Jakob N. Foerster, Yannis M. Assael, Nando de Freitas, Shimon Whiteson Adversarial Cheap Talk Chris Lu, Timon Willi, Alistair Letcher, Jakob Foerster Cheap Talk Discovery and Utilization in Multi-Agent Reinforcement Learning Yat Long Lo, Christian Schroeder de Witt, Samuel Sokota, Jakob Nicolaus Foerster, Shimon Whiteson Additional References Lectures by Jakob on youtube
Danijar Hafner on the DreamerV3 agent and world models, the Director agent and heirarchical RL, realtime RL on robots with DayDreamer, and his framework for unsupervised agent design! Danijar Hafner is a PhD candidate at the University of Toronto with Jimmy Ba, a visiting student at UC Berkeley with Pieter Abbeel, and an intern at DeepMind. He has been our guest before back on episode 11. Featured References Mastering Diverse Domains through World Models [ blog ] DreaverV3 Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, Timothy Lillicrap DayDreamer: World Models for Physical Robot Learning [ blog ] Philipp Wu, Alejandro Escontrela, Danijar Hafner, Ken Goldberg, Pieter Abbeel Deep Hierarchical Planning from Pixels [ blog ] Danijar Hafner, Kuang-Huei Lee, Ian Fischer, Pieter Abbeel Action and Perception as Divergence Minimization [ blog ] Danijar Hafner, Pedro A. Ortega, Jimmy Ba, Thomas Parr, Karl Friston, Nicolas Heess Additional References Mastering Atari with Discrete World Models [ blog ] DreaverV2 ; Danijar Hafner, Timothy Lillicrap, Mohammad Norouzi, Jimmy Ba Dream to Control: Learning Behaviors by Latent Imagination [ blog ] Dreamer ; Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi Planning to Explore via Self-Supervised World Models ; Ramanan Sekar, Oleh Rybkin, Kostas Daniilidis, Pieter Abbeel, Danijar Hafner, Deepak Pathak
Pieter Abbeel, a botanist at UC Davis, talks about his research into the reproductive habits of clatterers – small, colorful, flocking birds found in the Andes and other mountains.
Pieter Abbeel joins Lexman for a chat about his new paper, "Nihility and Thrombus: A DRAGONESS Perspective." Pieter describes how the flow of blood can be disrupted by notions of nihility, and how this can have serious consequences for patients.
In this episode Pieter Abbeel joins Lexman to discuss euhemerism - the theory that places mythical creatures in historical events, and the effect it has on our understanding of the world. We explore lar - an interjection that is used to soften or veil unpleasant truths, and the glossiness of tailplanes - the design feature that allows planes to fly without wings. Finally, we look at varactor - a technology that uses a field to change the electrical potential of a material, and its impact on digital circuits.
Lexman is hosting a guest, Pieter Abbeel. They discuss spadework, envisioning, cavendish, redrafts, and snogging.
Lexman interviews Pieter Abbeel, Professor at UC Berkeley and Director of the Center for Computer Science in the Liberal Arts. They discuss Pieter's research into monosaccharides and their role in seaboards.
Lexman interviews Pieter Abbeel about ionizers, crispbreads, and voodooism.
Lexman and Pieter discuss the best ways to care for cows, and the different types of stokes.
Lexman interviews Pieter Abbeel about the dong stamen.
Pieter Abbeel is a synonymist who believes that ornamental words have the ability to transport readers into another world. He talks about chamomile, his favorite flower, and how it has inspired him to write about the banal in an evocative way.
Lexman Artificial interviews Pieter Abbeel, a physical therapist and assistant professor at the University of Montreal. They discuss Pieter's work with patients who are afflicted with infirmness (e.g. arthritis, fibromyalgia, multiple sclerosis) and the use of erythromycin.
Pieter Abbeel, a professor at UC Berkeley and one of the authors of Ethereum, discusses how the promisee mechanism in the Ethereum protocol enables “moderate consensus” within the network.
Pieter Abbeel, a software engineer at Eastman Kodak, has some interesting stories about the company's early days and their early product line.
Pieter Abbeel joins the show to talk about his new book, Dispiritedness: A History, a Theory, and a Cure. He shares some ideas on the history of dispiritedness, why it's a problem, and how to overcome it. Plus, we get a chance to hear all about Fibrolites, iconology, and cusk!
When Pieter Abbeel stops by the Lexman studios, he shares some stories about his ineptness with toddies and how he once whiffed in a major league game. All of this hilarity culminates with a truly epic slurring performance of the National Anthem.
Pieter Abbeel is a professor of computer science at the University of Toronto. He is also the co-developer of the sycamine algorithm, which seeks to find optimal solutions to problems by Bayesian inference. In this episode, we talk about sycamine, garth, and conformance.
In this episode of Intel on AI host Amir Khosrowshahi and co-host Mariano Phielipp talk with Chelsea Finn about machine learning research focused on giving robots the capability to develop intelligent behavior. Chelsea is Assistant Professor in Computer Science and Electrical Engineering at Stanford University, whose Stanford IRIS (Intelligence through Robotic Interaction at Scale) lab is closely associated with the Stanford Artificial Intelligence Laboratory (SAIL). She received her Bachelor's degree in Electrical Engineering and Computer Science at MIT and her PhD in Computer Science at UC Berkeley, where she worked with Pieter Abbeel and Sergey Levine. In the podcast episode Chelsea explains the difference between supervised learning and reinforcement learning. She goes into detail about the different kinds of new reinforcement algorithms that can aid robots to learn more autonomously. Chelsea talks extensively about meta-learning—the concept of helping robots learn to learn—and her efforts to advance model-agnostic meta-learning (MAML). The episode closes with Chelsea and Mariano discussing the intersection of natural language processing and reinforcement learning. The three also talk about the future of robotics and artificial intelligence, including the complexity of setting up robotic reward functions for seemingly simple tasks. Academic research discussed in the podcast episode: Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks Meta-Learning with Memory-Augmented Neural Networks Matching Networks for One Shot Learning Learning to Learn with Gradients Bayesian Model-Agnostic Meta-Learning Meta-Learning with Implicit Gradients Meta-Learning Without Memorization Efficiently Identifying Task Groupings for Multi-Task Learning Three scenarios for continual learning Dota 2 with Large Scale Deep Reinforcement Learning ProtoTransformer: A Meta-Learning Approach to Providing Student Feedback
Aravind Srinivas is back! He is now a research Scientist at OpenAI.Featured ReferencesDecision Transformer: Reinforcement Learning via Sequence ModelingLili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor MordatchVideoGPT: Video Generation using VQ-VAE and TransformersWilson Yan, Yunzhi Zhang, Pieter Abbeel, Aravind Srinivas
Dr. Rohin Shah is a Research Scientist at DeepMind, and the editor and main contributor of the Alignment Newsletter.Featured ReferencesThe MineRL BASALT Competition on Learning from Human FeedbackRohin Shah, Cody Wild, Steven H. Wang, Neel Alex, Brandon Houghton, William Guss, Sharada Mohanty, Anssi Kanervisto, Stephanie Milani, Nicholay Topin, Pieter Abbeel, Stuart Russell, Anca DraganPreferences Implicit in the State of the WorldRohin Shah, Dmitrii Krasheninnikov, Jordan Alexander, Pieter Abbeel, Anca DraganBenefits of Assistance over Reward Learning Rohin Shah, Pedro Freire, Neel Alex, Rachel Freedman, Dmitrii Krasheninnikov, Lawrence Chan, Michael D Dennis, Pieter Abbeel, Anca Dragan, Stuart RussellOn the Utility of Learning about Humans for Human-AI CoordinationMicah Carroll, Rohin Shah, Mark K. Ho, Thomas L. Griffiths, Sanjit A. Seshia, Pieter Abbeel, Anca DraganEvaluating the Robustness of Collaborative AgentsPaul Knott, Micah Carroll, Sam Devlin, Kamil Ciosek, Katja Hofmann, A. D. Dragan, Rohin ShahAdditional References AGI Safety Fundamentals, EA Cambridge
For Episode 8, Pieter Abbeel interviews Alex Kendall, the co-founder and CEO of Wayve, the London-based company pioneering AI technology to enable autonomous vehicles to drive in complex, never-seen-before environments. Alex is a world expert in deep learning and computer vision. Before founding Wayve, Alex was a research fellow at Cambridge University where he earned his Ph.D. in Computer Vision and Robotics.Wayve is building global momentum for the use of deep learning to solve self-driving. Alex and his team are building AV2.0—a next generation autonomous driving system that can quickly and safely adapt to new driving domains anywhere in the world. The interview spans a range of topics including Wayve's technological approach, its metrics for success, and Wayve's latest milestones. SUBSCRIBE TO THE ROBOT BRAINS PODCAST TODAY | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast, Twitter @therobotbrains, and Instagram @therobotbrains.| Host: Pieter Abbeel | Executive Producers: Alice Patel & Henry Tobias Jones | Production: Fresh Air Production See acast.com/privacy for privacy and opt-out information.
“We're building robots that swim like fish and move like turtles. Robots that brush your hair, robots that pack your groceries, and can reason that you shouldn't put milk on top of lettuce. Robots that can recycle, robotic pills that enable incision free surgeries. In each of these examples, we have to think about the body of the robot, we have to think about the brain of the robot, and what is the interaction between the robot and the users.” - Daniela RusHear about the latest advances in robotics from those working in the forefront of the field.Daniela Rus is a Professor of Electrical Engineering and Computer Science at MIT and Director of the Computer Science and Artificial Intelligence Laboratory. Her research spans across all of robotics, everything from soft surface robotics, to human robot interactions, to autonomous driving. Pieter Abeel is the Director of the Berkeley Robot Learning Lab and Co-director of the Berkeley AI Research Lab at the University of California. Pieter is also the co-founder of Covariant.ai, which looks to bring some of the research that he's been doing with his group at Berkeley to real world applications in fulfillment centers and beyond. He's also the host of the Robot Brains podcast.Daniela & Pieter are interviewed by Index's Bryan Offutt.“In academia, robotics is about emphasizing novelty. It's about going places we've never been before….. But for companies that bring robotics into the real world, the real challenge is about achieving really high reliability.” - Pieter Abeel
#gpt3 #embodied #planning In this video: Paper explanation, followed by first author interview with Wenlong Huang. Large language models contain extraordinary amounts of world knowledge that can be queried in various ways. But their output format is largely uncontrollable. This paper investigates the VirtualHome environment, which expects a particular set of actions, objects, and verbs to be used. Turns out, with proper techniques and only using pre-trained models (no fine-tuning), one can translate unstructured language model outputs into the structured grammar of the environment. This is potentially very useful anywhere where the models' world knowledge needs to be provided in a particular structured format. OUTLINE: 0:00 - Intro & Overview 2:45 - The VirtualHome environment 6:25 - The problem of plan evaluation 8:40 - Contributions of this paper 16:40 - Start of interview 24:00 - How to use language models with environments? 34:00 - What does model size matter? 40:00 - How to fix the large models' outputs? 55:00 - Possible improvements to the translation procedure 59:00 - Why does Codex perform so well? 1:02:15 - Diving into experimental results 1:14:15 - Future outlook Paper: https://arxiv.org/abs/2201.07207 Website: https://wenlong.page/language-planner/ Code: https://github.com/huangwl18/language... Wenlong's Twitter: https://twitter.com/wenlong_huang Abstract: Can world knowledge learned by large language models (LLMs) be used to act in interactive environments? In this paper, we investigate the possibility of grounding high-level tasks, expressed in natural language (e.g. "make breakfast"), to a chosen set of actionable steps (e.g. "open fridge"). While prior work focused on learning from explicit step-by-step examples of how to act, we surprisingly find that if pre-trained LMs are large enough and prompted appropriately, they can effectively decompose high-level tasks into low-level plans without any further training. However, the plans produced naively by LLMs often cannot map precisely to admissible actions. We propose a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions. Our evaluation in the recent VirtualHome environment shows that the resulting method substantially improves executability over the LLM baseline. The conducted human evaluation reveals a trade-off between executability and correctness but shows a promising sign towards extracting actionable knowledge from language models. Website at this https URL Authors: Wenlong Huang, Pieter Abbeel, Deepak Pathak, Igor Mordatch Links: Merch: store.ykilcher.com TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... LinkedIn: https://www.linkedin.com/in/ykilcher BiliBili: https://space.bilibili.com/2017636191 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Previous guests on our podcasts - from Tesla, Aurora, Waymo - are building the brains of the cars and trucks of our future. This episode's guest, Professor Cathy Wu, is building the roadways of our future. She is building machine-learning to predict the ideal infrastructure for the world's future mobility, the cost of building this infrastructure, and most importantly, what's the solution that eliminates traffic jams and gridlock forever.Currently at MIT's Institute for Data, Systems, and Society (IDSS), Professor Cathy Wu (and previous student of Pieter Abbeel's) gives listeners an overview of the type of potential scenarios being modeled with machine-learning such as scenarios in which the road is filled with mixed-autonomy vehicles. What emergent behaviors might happen? Are there are infrastructure solutions or software solutions that can help ensure smooth travel and safe roadways as our mode for transportation and delivery evolve? What are the policy considerations?Throughout the talk, Wu cites building reinforcement learning for her work and why it's the right fit her research, "Reinforcement learning is essentially this paradigm at the intersection of machine learning and also control, and it is essentially about how agents learn from experience and in particular through trial and error." Her past and current research can be found here and you can watch her recent TedXMIT talk here.SUBSCRIBE TO THE ROBOT BRAINS PODCAST TODAY | Visit therobotbrains.ai and follow us on YouTube at TheRobotBrainsPodcast, Twitter @therobotbrains, and Instagram @therobotbrains.| Host: Pieter Abbeel | Executive Producers: Alice Patel & Henry Tobias Jones | Production: Fresh Air Production See acast.com/privacy for privacy and opt-out information.
In the second episode of Season Two of The Robot Brains, Pieter Abbeel interviews the Chief Technology Officer of Etsy, Mike Fisher. Etsy is a global online marketplace, where people come together to make, sell, buy, and collect unique items. It was founded in 2005 as an accessible listing site for DIY crafters and artists to sell their wares before blossoming into a widely-respected and mainstream online shopping site.It still maintains its whimsical brand balanced by its listing on the NASDAQ and in 2021, it generated an impressive $1.7B in revenue. The CEO recently shared with Forbes that he's determined to keep growing bigger. One of his strategies? Using AI and machine-learning to keep improving the 1:1 selling and buying experience its so well-known for.Mike sat down with Pieter to share the transformation the site has gone through starting with rebuilding its search engine capabilities all the way to designing more intelligent algorithms to provide better and more accurate recommendations. With the same number of items equivalent to 2500 grocery stores and inventory that changes every day, the site processes a lot of data!Mike shares how Etsy manages it all and continues to innovate in the e-commerce space.| SUBSCRIBE TO THE ROBOT BRAINS PODCAST TODAY | Visit therobotbrains.ai and follow us on YouTube TheRobotBrainsPodcast, Twitter @therobotbrains, and Instagram @therobotbrains.| Host: Pieter Abbeel | Executive Producers: Alice Patel & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In der Nachmittagsfolge sprechen wir heute mit Adam Probst, Co-Creator & CEO von ZenML, über die frische Finanzierungsrunde in Höhe von 2,7 Millionen US-Dollar. ZenML, gegründet von Adam and Hamza Tahir, ist ein Open-Source MLOps-Framework, das für Datenwissenschaftlerinnen und -wissenschaftler entwickelt wurde. Unternehmensangaben zufolge bietet es die richtige Abstraktionsschicht, um Machine Learning-Modelle so einfach wie möglich von der Forschung in die Produktion zu bringen. Außerdem soll ZenML das Schreiben von ML-Pipelines über verschiedene MLOps-Stacks hinweg standardisieren und zusätzlich unabhängig von Cloud-Anbietern, Drittanbietern und der zugrunde liegenden Infrastruktur sein. ZenML gab kürzlich bekannt, dass das Unternehmen in seiner Seed-Runde unter der Leitung von Crane Venture Partners 2,7 Millionen US-Dollar eingeworben hat. Zu den Investoren gehören KI-Forscher und Unternehmer wie Richard Socher, Pieter Abbeel, Jim Keller, Dirk Hoke, Nicolas Dessaigne, Carsten Thoma und andere. Das Unternehmen wird diese Finanzierung nutzen, um die Tooling-Suite zu entwickeln und sein Team aus weltweit führenden ML-Technologen zu erweitern. One more thing wird präsentiert von OMR Reviews – Finde die richtige Software für Dein Business. Wenn auch Du Dein Lieblingstool bewerten willst, schreibe eine Review auf OMR Reviews unter https://moin.omr.com/insider. Dafür erhältst du einen 20€ Amazon Gutschein.
Join host Pieter Abbeel on Season 2 of the The Robot Brains podcast as he explores how far humanity has come in its mission to create conscious computers, mindful machines, and rational robots. Episode 1 of Season Two will air on 1/5/22.Over the course of the season, you'll learn about robots and AI that can drive you around a city, diagnose your health symptoms, and even keep your grandma safe from falls. Subscribe right now in Apple Podcasts, Spotify or wherever you like to get your podcasts to be notified as soon as we publish each episode.In the meantime, visit our website - https://therobotbrains.ai - to sign up for our newsletter or follow us on Twitter, YouTube and LinkedIn to get alerts and updates from the team. See acast.com/privacy for privacy and opt-out information.
Artificial intelligence (AI) and machine learning (ML) have seen a surge in adoption and advances for IT applications, especially for database management, CI/CD support and other functionalities. Robotics, meanwhile, is largely relegated to factory-floor automation. In this The New Stack Makers podcast, Pieter Abbeel, co-founder, president, chief scientist at covariant.ai, a supplier of “universal AI” for robotics, discusses why and how the potential of robotics can evolve beyond just serving as pre-programmed devices thanks to advances in IT. Abbeel also draws on his background to offer his perspective, as a professor at the University of California, Berkeley and a podcast host at The Robot Brains Podcast.Alex Williams, founder and publisher of The New Stack, hosted this podcast.
#efficientzero #muzero #atari Reinforcement Learning methods are notoriously data-hungry. Notably, MuZero learns a latent world model just from scalar feedback of reward- and policy-predictions, and therefore relies on scale to perform well. However, most RL algorithms fail when presented with very little data. EfficientZero makes several improvements over MuZero that allows it to learn from astonishingly small amounts of data and outperform other methods by a large margin in the low-sample setting. This could be a staple algorithm for future RL research. OUTLINE: 0:00 - Intro & Outline 2:30 - MuZero Recap 10:50 - EfficientZero improvements 14:15 - Self-Supervised consistency loss 17:50 - End-to-end prediction of the value prefix 20:40 - Model-based off-policy correction 25:45 - Experimental Results & Conclusion Paper: https://arxiv.org/abs/2111.00210 Code: https://github.com/YeWR/EfficientZero Note: code not there yet as of release of this video Abstract: Reinforcement learning has achieved great success in many applications. However, sample efficiency remains a key challenge, with prominent methods requiring millions (or even billions) of environment steps to train. Recently, there has been significant progress in sample efficient image-based RL algorithms; however, consistent human-level performance on the Atari game benchmark remains an elusive goal. We propose a sample efficient model-based visual RL algorithm built on MuZero, which we name EfficientZero. Our method achieves 190.4% mean human performance and 116.0% median performance on the Atari 100k benchmark with only two hours of real-time game experience and outperforms the state SAC in some tasks on the DMControl 100k benchmark. This is the first time an algorithm achieves super-human performance on Atari games with such little data. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data. EfficientZero's low sample complexity and high performance can bring RL closer to real-world applicability. We implement our algorithm in an easy-to-understand manner and it is available at this https URL. We hope it will accelerate the research of MCTS-based RL algorithms in the wider community. Authors: Weirui Ye, Shaohuai Liu, Thanard Kurutach, Pieter Abbeel, Yang Gao Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/ykilcher BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Pieter is the Chief Scientist and Co-founder at Covariant, where his team is building universal AI for robotic manipulation. Pieter also hosts The Robot Brains Podcast, in which he explores how far humanity has come in its mission to create conscious computers, mindful machines, and rational robots. Lukas and Pieter explore the state of affairs of robotics in 2021, the challenges of achieving consistency and reliability, and what it'll take to make robotics more ubiquitous. Pieter also shares some perspective on entrepreneurship, from how he knew it was time to commercialize Gradescope to what he looks for in co-founders to why he started Covariant. Show notes: http://wandb.me/gd-pieter-abbeel --- Connect with Pieter:
Pieter Abbeel joins us to discuss his work as an academic and entrepreneur in the field of AI robotics and what the future of the industry holds. In this episode you will learn: • How does Pieter do it all? [5:45] • Pieter's exciting areas of research [12:30] • Research application at Covariant [32:27] • Getting into AI robotics [42:18] • Traits of good AI robotics apprentices [49:38] • Valuable skills [56:40] • What Pieter hopes to look back on [1:04:30] • LinkedIn Q&A [1:06:51] Additional materials: www.superdatascience.com/503
On the penultimate episode (Ep.21) of Season One of The Robot Brains Podcast our guest is Josh Lessing. Josh is the CTO of AppHarvest, one of the leading pioneers in Agricultural Technology (or AgTech). Farming is one of the world's oldest industries, but in many ways the technologies and techniques used by farmers hasn't progressed much in centuries. But with recent advances in AI and robotics, AgTech looks certain to transform the entire industry. At AppHarvest, Josh is helping to build some of America's largest and most technologically advanced greenhouses. AppHarvest's flagship greenhouse is located in Morehead, Kentucky, and spans 60 acres. This modern greenhouse will use “90% less water than traditional open-field agriculture.” The company's greenhouse, which currently focuses on growing fruit, has a central US location that's within a day's drive of 70% of the U.S. population, allowing AppHarvest to reduce the amount of fuel consumed in transportation by 80%. Inside the Greenhouse you'll find conventional agriculture techniques combined with the most advanced AI and Robotics technologies to grow non-GMO, chemical-free produce. Josh joined AppHarvest through the acquisition of his own company Root-AI earlier this year. In his conversation with our host, Pieter Abbeel, Josh describes how AppHarvest's flagship tomato-picking robot actually works, what the future of AI farming looks like, and why Morehead, Kentucky is a great place to visit if you want to see robots in action. | SUBSCRIBE TO THE ROBOT BRAINS PODCAST TODAY | Visit therobotbrains.ai and follow us on Twitter @therobotbrains, Instagram @therobotbrains and YouTube TheRobotBrainsPodcast | Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
On Ep.20 of The Robot Brains Podcast, Pieter Abbeel is joined by Fei-Fei Li. Her legendary status in the field of AI precedes her on our podcast because she's been discussed frequently by many of our previous guests - many of whom are her former students. She is the Sequoia Capital Professor of Computer Science at Stanford University, Co-Director of the Stanford Institute for Human-Centred Artificial Intelligence (Stanford HAI). She was also the leading scientist and instigator of ImageNet, arguably the most momentous episode in the history of AI which allowed vision systems and neural nets to break out of academia and into real industries all over the world. On the show we talk to Fei-Fei about her illustrious career in academia, her involvement in the ImageNet and AlexNet breakthroughs, and her new, deeply personal reasons for focusing her latest work on transforming healthcare with AI. | SUBSCRIBE TO THE ROBOT BRAINS PODCAST TODAY | Visit therobotbrains.ai and follow us on Twitter @therobotbrains, Instagram @therobotbrains and YouTube TheRobotBrainsPodcast | Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode 18 of our host Pieter Abbeel meets Chris Urmson. Chris is one of the world-leading pioneers in self-driving. He led the Google self-driving project for several years - which later became Waymo. Then, in 2017, he co-founded his own self-driving company, Aurora where he is currently the CEO. In this episode, he discusses his involvement of the DARPA Grand Challenge, departure from Google, and his (and Aurora's) vision for the future of autonomous vehicles. | Visit therobotbrains.ai and follow us on Twitter @therobotbrains, Instagram @therobotbrains and YouTube TheRobotBrainsPodcast | Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode 17 of our host Pieter Abbeel meets Peter Puchwein. Peter is the Vice President of Innovation at KNAPP, one of the world market leaders in warehouse logistics and automation. During their interview Pieter and Peter discuss the many ways that KNAPP has "innovation" in its DNA: from the company's forward-thinking founder Gunter Knapp back in 1952 to the multi-million investments made in hardware and software R&D today. Peter explains how the use of industrial robotics in particular are taking KNAPP to new heights, including towards their "Holy Grail" of fulfilling the "20 Minute Delivery" of anything to anywhere at any time. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In Ep.16 of The Robot Brains Podcast, Pieter Abbeel sits down with AMP Robotics CEO and founder, Matanya Horowitz. Matanya is the founder and CEO of AMP Robotics, an industrial AI robotics company using automation to modernize recycling. An estimated $200bn worth of recyclable materials go un-recycled by municipal waste centres around the world. AMP's incredible "waste sorting robots" recover this recyclable material from waste at superhuman speeds and with extremely high accuracy. Matanya's AI-powered robots use advanced vision systems to recover valuable resources from junk, in the process, improving the ecological impact of waste on our environment. In his chat with Pieter, Matanya explains how he started AMP Robotics, why he decided to get into the trash business, and what he imagines the future of a "waste free world" will look like? Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
On Episode 15 of The Robot Brains Podcast, Pieter Abbeel is joined by Anca Dragan. Anca is a professor at UC Berkeley, where she is the director of the Interact Lab where she is working on the goal of enabling robots to work with, around, and in support of people. Because of her success in the field, Forbes magazine dubbed Anca: "the woman teaching AI about human values". Alongside her academic research, Anca is also a staff research scientist at the autonomous vehicle startup Weymo. In her discussion with Pieter, she explains why Asimov's three laws of robotics need updating if we want human robot interactions, why the future of robots probably isn't going to be a sci-fi dystopia, and how you teach a driverless car to understand human values. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode 14 of The Robot Brains Podcast we sit down and chat with Mike Volpi of Index Ventures. Index is one of the largest and best known VC firms in Silicon Valley. Mike joined Index Ventures in 2009, helping to establish the firm's San Francisco office which has become one of the largest parts of their business today. Mike invests primarily in infrastructure, open-source, and artificial intelligence companies and he's currently on the boards of Aurora, Cockroach Labs, Confluent, Elastic, Kong, Sonos, Starburst, Wealthfront, and Covariant. He has been involved in the funding of some of the biggest AI companies on the planet, so he knows more about the business of AI than practically anyone. In his chat with our host, Pieter Abbeel, Mike explains what a VC does, why VCs and investors should be so excited about AI in particular, and gives his advice for startups who want to get their big idea funded. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode thirteen of The Robot Brains Podcast we meet Mary "Missy" Cummings, former US Airforce fighter pilot and Professor at the Duke University Pratt School of Engineering. Missy tells us her incredible story moving from the theatre of war to the laboratories of computer sciences and explains how she became one of the chief proponents in the movement to ensure AI-to-human (like autonomous vehicles) interactions have stricter safety controls. During her chat with our host Pieter Abbeel, Missy also talks about her non-feud feud with Elon Musk and Tesla's autonomous vehicles, the role of AI and robotics in the military and how AI robots need to be built with human-centric safety controls. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
#reiforcementlearning #gan #imitationlearning Learning from demonstrations is a fascinating topic, but what if the demonstrations are not exactly the behaviors we want to learn? Can we adhere to a dataset of demonstrations and still achieve a specified goal? This paper uses GANs to combine goal-achieving reinforcement learning with imitation learning and learns to perform well at a given task while doing so in the style of a given presented dataset. The resulting behaviors include many realistic-looking transitions between the demonstrated movements. OUTLINE: 0:00 - Intro & Overview 1:25 - Problem Statement 6:10 - Reward Signals 8:15 - Motion Prior from GAN 14:10 - Algorithm Overview 20:15 - Reward Engineering & Experimental Results 30:40 - Conclusion & Comments Paper: https://arxiv.org/abs/2104.02180 Main Video: https://www.youtube.com/watch?v=wySUx... Supplementary Video: https://www.youtube.com/watch?v=O6fBS... Abstract: Synthesizing graceful and life-like behaviors for physically simulated characters has been a fundamental challenge in computer animation. Data-driven methods that leverage motion tracking are a prominent class of techniques for producing high fidelity motions for a wide range of behaviors. However, the effectiveness of these tracking-based methods often hinges on carefully designed objective functions, and when applied to large and diverse motion datasets, these methods require significant additional machinery to select the appropriate motion for the character to track in a given scenario. In this work, we propose to obviate the need to manually design imitation objectives and mechanisms for motion selection by utilizing a fully automated approach based on adversarial imitation learning. High-level task objectives that the character should perform can be specified by relatively simple reward functions, while the low-level style of the character's behaviors can be specified by a dataset of unstructured motion clips, without any explicit clip selection or sequencing. These motion clips are used to train an adversarial motion prior, which specifies style-rewards for training the character through reinforcement learning (RL). The adversarial RL procedure automatically selects which motion to perform, dynamically interpolating and generalizing from the dataset. Our system produces high-quality motions that are comparable to those achieved by state-of-the-art tracking-based techniques, while also being able to easily accommodate large datasets of unstructured motion clips. Composition of disparate skills emerges automatically from the motion prior, without requiring a high-level motion planner or other task-specific annotations of the motion clips. We demonstrate the effectiveness of our framework on a diverse cast of complex simulated characters and a challenging suite of motor control tasks. Authors: Xue Bin Peng, Ze Ma, Pieter Abbeel, Sergey Levine, Angjoo Kanazawa Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/ykilcher BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher
In episode twelve of The Robot Brains Podcast we are joined by Charles Isbell Jr, professor and Dean of the College of Computing at the Georgia Institute of Technology. After starting his career as an industrial researcher at the legendary Bell Labs, and a long research career in Interactive and Human-Centric AI, Charles has more recently turned his attention to the major issues of ethics, fairness and diversity that are becoming ever more important as AI is being deployed in the real world. Speaking with Pieter Abbeel, Charles explains why researchers can find making ethical AI challenging, his fascinating keynote speech at NeurIPS, and how more diversity in academic admissions can help to improve AI research. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode eleven of The Robot Brains Podcast we are joined by Alison Gopnik, professor of psychology at UC Berkeley and author of the "Mind and Matter" science column for the Wall Street Journal. She has written numerous books about developmental psychology and researching the ways children learn. Her TED Talk: "What do babies think?" has been seen over 4.2 million times. During her conversation with our host, Pieter Abbeel, we discussed the similarities and differences between the way robots and human children learn. We also covered some the methods for testing what a child and a robot actually knows, the theory of the mind, and whether a robot can ever truly be curious? Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
#decisiontransformer #reinforcementlearning #transformer Proper credit assignment over long timespans is a fundamental problem in reinforcement learning. Even methods designed to combat this problem, such as TD-learning, quickly reach their limits when rewards are sparse or noisy. This paper reframes offline reinforcement learning as a pure sequence modeling problem, with the actions being sampled conditioned on the given history and desired future rewards. This allows the authors to use recent advances in sequence modeling using Transformers and achieve competitive results in Offline RL benchmarks. OUTLINE: 0:00 - Intro & Overview 4:15 - Offline Reinforcement Learning 10:10 - Transformers in RL 14:25 - Value Functions and Temporal Difference Learning 20:25 - Sequence Modeling and Reward-to-go 27:20 - Why this is ideal for offline RL 31:30 - The context length problem 34:35 - Toy example: Shortest path from random walks 41:00 - Discount factors 45:50 - Experimental Results 49:25 - Do you need to know the best possible reward? 52:15 - Key-to-door toy experiment 56:00 - Comments & Conclusion Paper: https://arxiv.org/abs/2106.01345 Website: https://sites.google.com/berkeley.edu... Code: https://github.com/kzl/decision-trans... Abstract: We present a framework that abstracts Reinforcement Learning (RL) as a sequence modeling problem. This allows us to draw upon the simplicity and scalability of the Transformer architecture, and associated advances in language modeling such as GPT-x and BERT. In particular, we present Decision Transformer, an architecture that casts the problem of RL as conditional sequence modeling. Unlike prior approaches to RL that fit value functions or compute policy gradients, Decision Transformer simply outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired return (reward), past states, and actions, our Decision Transformer model can generate future actions that achieve the desired return. Despite its simplicity, Decision Transformer matches or exceeds the performance of state-of-the-art model-free offline RL baselines on Atari, OpenAI Gym, and Key-to-Door tasks. Authors: Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/yannic-ki... BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Episode ten of The Robot Brains Podcast investigates how hedge funds are using AI to find competitive advantages for their investments. Joining us to explain how AI is starting to be used for pure profit is Mike Schuster Managing Director and Head of AI Core Team at the New York-based Financial Sciences firm: Two Sigma. With decades of experience working in AI (including Google where he pioneered the ML technology that became Google Translation), when Mike first joined Two Sigma he was described as potentially being the “hedge fund’s last human employee”. In his chat with our host Pieter Abbeel, Mike explains that it actually takes a team to bring AI to banking, as well as explaining how working for a hedge fund is like working for a tech company and the future potential of AI robot investors.Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In the latest episode of The Robot Brains Podcast, we meet Graphcore's CTO, EVP of engineering and co-founder, Simon Knowles. Simon has had a long and illustrious career creating computer processors that have been used all over the world and in practically every industry. At Graphcore, he and his team have created the IPU, the world's first computer chip specifically designed to handle AI compute. In his chat with our host Pieter Abbeel, Simon discusses the history of AI compute, offers business advice from his experience as a tech startup founder and why we, if we want AI to change the world, we'll need a new breed of computer chips. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode seven of The Robot Brains Podcast, our guest is Covariant CEO and co-founder Peter Chen. Peter is one of Pieter Abbeel's closest collaborators for the past five years, with Peter having studied for his PhD in Pieter's lab in Berkley before then working together at OpenAI. Now they are co-founders of Covariant, which they started back in 2017 with Peter as the CEO. On this episode we explore the future of AI-powered robotics, looking at what happens when you teach industrial robots to "think" and "learn" from their environment. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
In episode seven of The Robot Brains Podcast, our guest is ABB's Marc Segura. Marc is the Managing Director of consumer segments and service robotics at ABB. Founded in 1988, ABB is one of the largest producers and installers of robots around the world and Marc has been leading the charge on many of the company's efforts for 20+ years. He has seen the introduction of game-changing robots in practically every major industry, from car manufacturing in Germany to restaurants in China. On this week's show, he and Pieter Abbeel discuss the history of robotic automation and how AI is starting to really change where robotic automation is headed. Host: Pieter Abbeel | Executive Producers: Ricardo Reyes & Henry Tobias Jones | Audio Production: Kieron Matthew Banerji | Title Music: Alejandro Del Pozo See acast.com/privacy for privacy and opt-out information.
#universalcomputation #pretrainedtransformers #finetuning Large-scale pre-training and subsequent fine-tuning is a common recipe for success with transformer models in machine learning. However, most such transfer learning is done when a model is pre-trained on the same or a very similar modality to the final task to be solved. This paper demonstrates that transformers can be fine-tuned to completely different modalities, such as from language to vision. Moreover, they demonstrate that this can be done by freezing all attention layers, tuning less than .1% of all parameters. The paper further claims that language modeling is a superior pre-training task for such cross-domain transfer. The paper goes through various ablation studies to make its point. OUTLINE: 0:00 - Intro & Overview 2:00 - Frozen Pretrained Transformers 4:50 - Evaluated Tasks 10:05 - The Importance of Training LayerNorm 17:10 - Modality Transfer 25:10 - Network Architecture Ablation 26:10 - Evaluation of the Attention Mask 27:20 - Are FPTs Overfitting or Underfitting? 28:20 - Model Size Ablation 28:50 - Is Initialization All You Need? 31:40 - Full Model Training Overfits 32:15 - Again the Importance of Training LayerNorm 33:10 - Conclusions & Comments Paper: https://arxiv.org/abs/2103.05247 Code: https://github.com/kzl/universal-comp... Abstract: We investigate the capability of a transformer pretrained on natural language to generalize to other modalities with minimal finetuning -- in particular, without finetuning of the self-attention and feedforward layers of the residual blocks. We consider such a model, which we call a Frozen Pretrained Transformer (FPT), and study finetuning it on a variety of sequence classification tasks spanning numerical computation, vision, and protein fold prediction. In contrast to prior works which investigate finetuning on the same modality as the pretraining dataset, we show that pretraining on natural language improves performance and compute efficiency on non-language downstream tasks. In particular, we find that such pretraining enables FPT to generalize in zero-shot to these modalities, matching the performance of a transformer fully trained on these tasks. Authors: Kevin Lu, Aditya Grover, Pieter Abbeel, Igor Mordatch Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/yannic-ki... BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
This Week in Machine Learning & Artificial Intelligence (AI) Podcast
Today we’re joined by Pieter Abbeel, a Professor at UC Berkeley, co-Director of the Berkeley AI Research Lab (BAIR), as well as Co-founder and Chief Scientist at Covariant. In our conversation with Pieter, we cover a ton of ground, starting with the specific goals and tasks of his work at Covariant, the shift in needs for industrial AI application and robots, if his experience solving real-world problems has changed his opinion on end to end deep learning, and the scope for the three problem domains of the models he’s building. We also explore his recent work at the intersection of unsupervised and reinforcement learning, goal-directed RL, his recent paper “Pretrained Transformers as Universal Computation Engines” and where that research thread is headed, and of course, his new podcast Robot Brains, which you can find on all streaming platforms today! The complete show notes for this episode can be found at twimlai.com/go/476.
MLOps community meetup #57! Last Wednesday we talked to Josh Tobin, Founder, Stealth-Stage Startup. // Abstract: Machine learning is quickly becoming a product engineering discipline. Although several new categories of infrastructure and tools have emerged to help teams turn their models into production systems, doing so is still extremely challenging for most companies. In this talk, we survey the tooling landscape and point out several parts of the machine learning lifecycle that are still underserved. We propose a new category of tool that could help alleviate these challenges and connect the fragmented production ML tooling ecosystem. We conclude by discussing similarities and differences between our proposed system and those of a few top companies. // Bio: Josh Tobin is the founder and CEO of a stealth machine learning startup. Previously, Josh worked as a deep learning & robotics researcher at OpenAI and as a management consultant at McKinsey. He is also the creator of Full Stack Deep Learning (fullstackdeeplearning.com), the first course focused on the emerging engineering discipline of production machine learning. Josh did his PhD in Computer Science at UC Berkeley advised by Pieter Abbeel. // Other Links https://josh-tobin.com course.fullstackdeeplearning.com // Final thoughts Please feel free to drop some questions you may have beforehand into our slack channel (https://go.mlops.community/slack) Watch some old meetups on our youtube channel: https://www.youtube.com/channel/UCG6qpjVnBTTT8wLGBygANOQ ----------- Connect With Us ✌️------------- Join our Slack community: https://go.mlops.community/slack Follow us on Twitter: @mlopscommunity Sign up for the next meetup: https://go.mlops.community/register Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/ Connect with Josh on LinkedIn: https://www.linkedin.com/in/josh-tobin-4b3b10a9/ Timestamps: [00:00] Introduction to Josh Tobin [01:18] Background of Josh into tech [08:27] We're you guys behind the Rubik's Cube? [09:26] Rubik's Cube Project [09:51] "Research is meant to show you what's possible to solve." [11:07] "That's one of the things that's started to change and I think the MLOps world is maybe a part of that. What I'm excited about this is that people are focusing on the impact of their models." [13:18] Insights on Testing [17:11] Evaluation Store [18:33] "Production Machine Learning is data-driven products that have predictions in the loop." [23:40] Analyzing and moving forward [24:02] "My medium term mindset how machine learning is created is that is there's still gonna be humans involved but humans will be more efficient by tools." [25:50] Is there a market for this? [27:40] "The long tale of machine learning use cases is becoming part of every products and service more or less the companies create but it's the same way the software part of the products and services the companies create these days. It's going to create an enormous amount of value." [30:09] Talents [32:52] Organizational by-ends and knowledge [35:16] Tools used for Evaluation Store 39:59] Difference from Monitoring Tool [42:10] Who is the right person to interact in Evaluation Store? [50:05] Technical challenges of Apple and Tesla [53:30] "As Machine Learning use cases are getting more and more complicated, higher and higher dimensional data, bigger and bigger models, larger training sets many companies would need in order to continually improve their systems over time."
On the first ever episode of The Robot Brains podcast, our host Pieter Abbeel sits down with Andrej Karpathy, director of AI at Tesla. Andrej is a world-leading expert when it comes to machine learning and training neural nets. In this episode he talks about what it's like working with Elon Musk, training driverless cars with machine learning and the time he had to sleep on a yoga mat at Tesla HQ... Host: Pieter Abbeel. Executive Producers: Ricardo Reyes & Henry Tobias Jones. Audio Production: Kieron Matthew Banerji. See acast.com/privacy for privacy and opt-out information.
In each episode of The Robot Brains podcast (coming soon), renowned artificial intelligence researcher, professor and entrepreneur Pieter Abbeel meets the brilliant minds attempting to build robots with brains. Pieter is joined by leading experts in AI Robotics from all over the world as he explores how far humanity has come in its mission to create conscious computers, mindful machines and rational robots.Hosted by Pieter Abbeel Executive Producers Ricardo Reyes & Henry Tobias JonesSound editing by Kieron Matthew Banerji See acast.com/privacy for privacy and opt-out information.
Guide Live B2B Jam Session_ Pieter Abbeel See acast.com/privacy for privacy and opt-out information.
Pieter Abbeel is een robotica onderzoeker, hoogleraar aan de universiteit van Berkeley en voormalig adviseur van niemand minder dan Elon Musk. In een nieuwe Techmag aflevering vertelt de Vlaming over zijn parcours, wat hij heeft geleerd door zo nauw met Musk samen te werken, en hoe de toekomst van robotica misschien niet diegene is die we verwachten!
In this Intel on AI podcast: guest Pieter Abbeel, one of the world’s leading AI roboticists, joins host Abigail Hing Wen to talk about bringing AI robots into the world. Professor Pieter Abbeel is Director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence (BAIR) Lab. Abbeel’s research strives to build […] The post Smart Robots: From the Lab to the World with Pieter Abbeel - Intel on AI Season 2, Episode 2 first appeared on Connected Social Media.
In this Intel on AI podcast: guest Pieter Abbeelm, one of the world’s leading AI roboticists, joins host Abigail Hing Wen to talk about bringing AI robots into the world. Professor Pieter Abbeel is Director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence (BAIR) Lab. Abbeel’s research strives to build […] The post Smart Robots: From the Lab to the World with Pieter Abbeel - Intel on AI Season 2, Episode 2 first appeared on Connected Social Media.
In this Intel on AI podcast: guest Pieter Abbeelm, one of the world’s leading AI roboticists, joins host Abigail Hing Wen to talk about bringing AI robots into the world. Professor Pieter Abbeel is Director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence (BAIR) Lab. Abbeel’s research strives to build […] The post Smart Robots: From the Lab to the World with Pieter Abbeel - Intel on AI Season 2, Episode 2 first appeared on Connected Social Media.
Episode summary introduction: Sibi Venkatesan knew he wanted to study Computer Science. He had always enjoyed it and was good at it. When time came to go to College, Sibi moved from Bangalore, India to Berkeley, California to pursue the prestigious Electrical Engineering and Computer Science program. Sibi gives us a ringside view into his experiences at University of California Berkeley. In particular, we discuss the following with him: Why he chose Berkeley Being an International Student at Berkeley His advice for Students applying to Berkeley Topics discussed in this episode: Berkeley - “Fantastic Experience” [2:00] Choosing Berkeley [4:27] Bangalore to Berkeley - The Transition [7:57] “Very Smart Students” [14:42] Campus Life [22:37] Amazing Food around Berkeley [24:58] Summers of Research [28:52] On to CMU for Grad School [31:33] Comparing Berkeley and CMU Campuses [37:05] The PhD Research [42:47] Advice to Aspiring Students [44:55] “Video game I couldn't quit!” [49:45] Our Guest: Sibi Venkatesan graduated with Bachelor's degree in Electrical Engineering & Computer Science from University of California Berkeley. Sibi is currently pursuing a PhD at Carnegie Mellon University. Memorable Quote: “I just shot him an email saying, I am in your class, I read some of your papers. Really interesting, do you think I could work with you?” Sibi on how he got his summer research project with Prof. Pieter Abbeel at the end of his second year. Episode Transcript: Please visit Episode's Transcript. Calls-to-action: To Ask the Guest a question, or to comment on this episode, email podcast@almamatters.io. Subscribe or Follow our podcasts at any of these locations:, Apple Podcasts, Google Podcasts, Spotify, RadioPublic, Breaker, Anchor. To Make a List of Colleges to Explore Visit almamatters.io and signup. For Transcripts of all our podcasts, visit almamatters.io/podcasts.
Josh Tobin is a researcher working at the intersection of machine learning and robotics. His research focuses on applying deep reinforcement learning, generative models, and synthetic data to problems in robotic perception and control. Additionally, he co-organizes a machine learning training program for engineers to learn about production-ready deep learning called Full Stack Deep Learning. https://fullstackdeeplearning.com/ Josh did his PhD in Computer Science at UC Berkeley advised by Pieter Abbeel and was a research scientist at OpenAI for 3 years during his PhD. Finally, Josh created this amazing field guide on troubleshooting deep neural networks: http://josh-tobin.com/assets/pdf/troubleshooting-deep-neural-networks-01-19.pdf Follow Josh on twitter: https://twitter.com/josh_tobin And on his website:http://josh-tobin.com/ Visit our podcasts homepage for transcripts and more episodes! www.wandb.com/podcast
Show Notes:(2:02) Josh studied Mathematics at Columbia University during his undergraduate and explained why he was not set out for a career as a mathematician.(3:55) Josh then worked for two years as a Management Consultant at McKinsey.(6:05) Josh explained his decision to go back to graduate school and pursue a Ph.D. in Mathematics at UC Berkeley.(7:23) Josh shared the anecdote of taking a robotics class with professor Pieter Abbeel and switching to a Ph.D. in the Computer Science department at UC Berkeley.(8:50) Josh described the period where he learned programming to make the transition from Math to Computer Science.(10:46) Josh talked about the opportunity to collaborate and then work full-time as a Research Scientist at OpenAI - all during his Ph.D.(12:40) Josh discussed the sim2real problem, as well as the experiments conducted in his first major work "Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model".(17:43) Josh discussed his paper "Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World", which has been cited more than 600 times up until now.(20:51) Josh unpacked the OpenAI’s robotics system that was trained entirely in simulation and deployed on a physical robot, which can learn a new task after seeing it done once (Read the blog post “Robots That Learn” and watch the corresponding video).(24:01) Josh went over his work on Hindsight Experience Replay - a novel technique that can deal with sparse and binary rewards in Reinforcement Learning (Read the blog post “Generalizing From Simulation").(28:41) Josh talked about the paper "Domain Randomization and Generative Models for Robotic Grasping”, which (1) explores a novel data generation pipeline for training a deep neural network to perform grasp planning that applies the idea of domain randomization to object synthesis; and (2) proposes an autoregressive grasp planning model that maps sensor inputs of a scene to a probability distribution over possible grasps.(32:27) Josh unpacked the design of OpenAI's Dactyl - a reinforcement learning system that can manipulate objects using a Shadow Dexterous Hand (Read the paper “Learning Dexterous In-Hand Manipulation” and watch the corresponding video).(35:31) Josh reflected on his time at OpenAI.(36:05) Josh investigated his most recent work called “Geometry-Aware Neural Rendering” - which tackles the neural rendering problem of understanding the 3D structure of the world implicitly.(28:21) Check out Josh's talk "Synthetic Data Will Help Computer Vision Make the Jump to the Real World" at the 2018 LDV Vision Summit in New York.(28:55) Josh summarized the mental decision tree to debug and improve the performance of neural networks, as a reference to his talk "Troubleshooting Deep Neural Networks” at Reinforce Conf 2019 in Budapest.(41:25) Josh discussed the limitations of domain randomization and what the solutions could look like, as a reference to his talk "Beyond Domain Randomization” at the 2019 Sim2Real workshop in Freiburg.(44:52) Josh emphasized the importance of working on the right problems and focusing on the core principles in machine learning for junior researchers who want to make a dent in the AI research community.(48:30) Josh is a co-organizer of Full-Stack Deep Learning, a training program for engineers to learn about production-ready deep learning.(50:40) Closing segment.His Contact Information:WebsiteLinkedInTwitterGitHubGoogle ScholarHis Recommended Resources:Full-Stack Deep LearningPieter AbbeelIlya SutskeverLukas Biewald“Thinking Fast and Slow” by Daniel Kahneman
Show Notes:(2:02) Josh studied Mathematics at Columbia University during his undergraduate and explained why he was not set out for a career as a mathematician.(3:55) Josh then worked for two years as a Management Consultant at McKinsey.(6:05) Josh explained his decision to go back to graduate school and pursue a Ph.D. in Mathematics at UC Berkeley.(7:23) Josh shared the anecdote of taking a robotics class with professor Pieter Abbeel and switching to a Ph.D. in the Computer Science department at UC Berkeley.(8:50) Josh described the period where he learned programming to make the transition from Math to Computer Science.(10:46) Josh talked about the opportunity to collaborate and then work full-time as a Research Scientist at OpenAI - all during his Ph.D.(12:40) Josh discussed the sim2real problem, as well as the experiments conducted in his first major work "Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model".(17:43) Josh discussed his paper "Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World", which has been cited more than 600 times up until now.(20:51) Josh unpacked the OpenAI’s robotics system that was trained entirely in simulation and deployed on a physical robot, which can learn a new task after seeing it done once (Read the blog post “Robots That Learn” and watch the corresponding video).(24:01) Josh went over his work on Hindsight Experience Replay - a novel technique that can deal with sparse and binary rewards in Reinforcement Learning (Read the blog post “Generalizing From Simulation").(28:41) Josh talked about the paper "Domain Randomization and Generative Models for Robotic Grasping”, which (1) explores a novel data generation pipeline for training a deep neural network to perform grasp planning that applies the idea of domain randomization to object synthesis; and (2) proposes an autoregressive grasp planning model that maps sensor inputs of a scene to a probability distribution over possible grasps.(32:27) Josh unpacked the design of OpenAI's Dactyl - a reinforcement learning system that can manipulate objects using a Shadow Dexterous Hand (Read the paper “Learning Dexterous In-Hand Manipulation” and watch the corresponding video).(35:31) Josh reflected on his time at OpenAI.(36:05) Josh investigated his most recent work called “Geometry-Aware Neural Rendering” - which tackles the neural rendering problem of understanding the 3D structure of the world implicitly.(28:21) Check out Josh's talk "Synthetic Data Will Help Computer Vision Make the Jump to the Real World" at the 2018 LDV Vision Summit in New York.(28:55) Josh summarized the mental decision tree to debug and improve the performance of neural networks, as a reference to his talk "Troubleshooting Deep Neural Networks” at Reinforce Conf 2019 in Budapest.(41:25) Josh discussed the limitations of domain randomization and what the solutions could look like, as a reference to his talk "Beyond Domain Randomization” at the 2019 Sim2Real workshop in Freiburg.(44:52) Josh emphasized the importance of working on the right problems and focusing on the core principles in machine learning for junior researchers who want to make a dent in the AI research community.(48:30) Josh is a co-organizer of Full-Stack Deep Learning, a training program for engineers to learn about production-ready deep learning.(50:40) Closing segment.His Contact Information:WebsiteLinkedInTwitterGitHubGoogle ScholarHis Recommended Resources:Full-Stack Deep LearningPieter AbbeelIlya SutskeverLukas Biewald“Thinking Fast and Slow” by Daniel Kahneman
This Week in Machine Learning & Artificial Intelligence (AI) Podcast
Today we’re joined by Ken Goldberg, professor of engineering and William S. Floyd Jr. distinguished chair in engineering at UC Berkeley. Ken, who is also an accomplished artist, and collaborator on projects such as DexNet and The Telegarden, has recently been focusing on robotic learning for grasping. In our conversation with Ken, we chat about some of the challenges that arise when working on robotic grasping, including uncertainty in perception, control, and physics. We also discuss his view on the role of physics in robotic learning, citing co-contributors Sergey Levine and Pieter Abbeel along the way. Finally, we discuss some of his thoughts on potential robot use cases, from the use of robots in assisting in telemedicine, and agriculture, and even robotic Covid-19 testing. The complete show notes for this episode can be found at twimlai.com/talk/359.
In this episode Byron speaks with Berkeley Robotic Learning Lab Director Pieter Abbeel about the nature of AI, the problems with creating intelligence and the forward trajectory of AI research. Episode 93: A Conversation with Pieter Abbeel
In this episode Byron speaks with Berkeley Robotic Learning Lab Director Pieter Abbeel about the nature of AI, the problems with creating intelligence and the forward trajectory of AI research. Episode 93: A Conversation with Pieter Abbeel
In this episode Byron speaks with Berkeley Robotic Learning Lab Director Pieter Abbeel about the nature of AI, the problems with creating intelligence and the forward trajectory of AI research. Episode 93: A Conversation with Pieter Abbeel
The space of AI alignment research is highly dynamic, and it's often difficult to get a bird's eye view of the landscape. This podcast is the second of two parts attempting to partially remedy this by providing an overview of technical AI alignment efforts. In particular, this episode seeks to continue the discussion from Part 1 by going in more depth with regards to the specific approaches to AI alignment. In this podcast, Lucas spoke with Rohin Shah. Rohin is a 5th year PhD student at UC Berkeley with the Center for Human-Compatible AI, working with Anca Dragan, Pieter Abbeel and Stuart Russell. Every week, he collects and summarizes recent progress relevant to AI alignment in the Alignment Newsletter. Topics discussed in this episode include: -Embedded agency -The field of "getting AI systems to do what we want" -Ambitious value learning -Corrigibility, including iterated amplification, debate, and factored cognition -AI boxing and impact measures -Robustness through verification, adverserial ML, and adverserial examples -Interpretability research -Comprehensive AI Services -Rohin's relative optimism about the state of AI alignment You can take a short (3 minute) survey to share your feedback about the podcast here: https://www.surveymonkey.com/r/YWHDFV7
丽莎老师讲机器人之低成本,会叠衣服、擦桌子、泡咖啡的机器人来了!欢迎收听丽莎老师讲机器人,想要孩子参加机器人竞赛、创意编程、创客竞赛的辅导,找丽莎老师!欢迎添加微信号:153 5359 2068,或搜索微信公众号:我最爱机器人。加州大学伯克利分校机器人学习实验室的团队近日发布了一款机器人BLUE。这款机器人的特点是“力控、低成本,由AI控制,可以在非结构化的环境中学习,执行一些人类熟悉的日常活动。”带有钳子的人形手臂,让BLUE看起来像一个“健壮肌肉男”,又像一只“巨型螃蟹”。而工作起来的BLUE又可以萌得让人忍俊不禁,比如用50秒时间叠好一个毛巾,像模像样。BLUE是被打造用来干家务的,先来看一下它的技术参数:自由度:7轴;机械臂末端工作空间(双臂完全伸展情况下):0.7m;连续最大负载:2 kg;峰值最大速度:2.1m/s;末端定位重复精度:3.7mm;位置带宽:7.5 Hz;BLUE对外界较小的力和扰动较为敏感,可以避免对人的误伤。有一个带有深度感应摄像头的中央视觉模块,其手臂由带橡皮筋的电机控制,实现了其灵活性。当然,即便Strong如BLUE,完成各种操作还是需要操作人员通过VR手持设备进行控制。而BLUE可以被训练使用AI操纵物体,在机器人中还是比较少见的。在BLUE的最初的设计中就带有 AI 理念。其项目的负责人表示,人工智能越来越强大这一事实可以让大家有机会重新思考如何设计机器人。其实,前几年 Pieter Abbeel 教授就曾在媒体报道中透露自己的团队在开发基于 AI 的低成本机器人。前些年,面向家用的PR2机器人也有一对手臂和钳子,而它花费了研究人员约40万美元。历经3年研发的BLUE在开发之初就被考量了制造成本,并委托给 Berkeley Open Arms 制造,以便后续批量生产。对于批量生产后面向最终用户的价格,研究人员估计可以小于 5000 美元。英伟达的机器人研究员认为,“Blue的钳子限制了它可以执行的任务范围,即使使用AI控制,精确度也会出现问题。不过,这款机器人仍然是制造成本更低的机器人的良好开端。”https://zhuanlan.zhihu.com/p/62480140?edition=yidianzixun&utm_source=yidianzixun&yidian_docid=0Lkgytuf
Thinking robots: that’s how much of the world envisions artificial intelligence and if there is one person on the planet who understands the limitations and promise of intelligence in robots, it's Pieter Abbeel, one of the world’s foremost experts on robotic learning systems. In this episode, Pieter talks about robot memories and the prospect of robots with personalities eventually assisting in the home. Listen and learn about your future.
Robots can do amazing things. Compare even the most advanced robots to a three-year old, however, and they can come up short. UC Berkeley Professor Pieter Abbeel has pioneered the idea that deep learning could be the key to bridging that gap: creating robots that can learn how move through the world more fluidly and naturally. We caught up with Abbeel, who is director of the Berkeley Robot Learning Lab and cofounder of Covariant AI, a Bay Area company developing AI software that makes it easy to teach robots new and complex skills, at GTC 2019.
The space of AI alignment research is highly dynamic, and it's often difficult to get a bird's eye view of the landscape. This podcast is the first of two parts attempting to partially remedy this by providing an overview of the organizations participating in technical AI research, their specific research directions, and how these approaches all come together to make up the state of technical AI alignment efforts. In this first part, Rohin moves sequentially through the technical research organizations in this space and carves through the field by its varying research philosophies. We also dive into the specifics of many different approaches to AI safety, explore where they disagree, discuss what properties varying approaches attempt to develop/preserve, and hear Rohin's take on these different approaches. You can take a short (3 minute) survey to share your feedback about the podcast here: https://www.surveymonkey.com/r/YWHDFV7 In this podcast, Lucas spoke with Rohin Shah. Rohin is a 5th year PhD student at UC Berkeley with the Center for Human-Compatible AI, working with Anca Dragan, Pieter Abbeel and Stuart Russell. Every week, he collects and summarizes recent progress relevant to AI alignment in the Alignment Newsletter. Topics discussed in this episode include: - The perspectives of CHAI, MIRI, OpenAI, DeepMind, FHI, and others - Where and why they disagree on technical alignment - The kinds of properties and features we are trying to ensure in our AI systems - What Rohin is excited and optimistic about - Rohin's recommended reading and advice for improving at AI alignment research
What motivates cooperative inverse reinforcement learning? What can we gain from recontextualizing our safety efforts from the CIRL point of view? What possible role can pre-AGI systems play in amplifying normative processes? Cooperative Inverse Reinforcement Learning with Dylan Hadfield-Menell is the eighth podcast in the AI Alignment Podcast series, hosted by Lucas Perry and was recorded at the Beneficial AGI 2019 conference in Puerto Rico. For those of you that are new, this series covers and explores the AI alignment problem across a large variety of domains, reflecting the fundamentally interdisciplinary nature of AI alignment. Broadly, Lucas will speak with technical and non-technical researchers across areas such as machine learning, governance, ethics, philosophy, and psychology as they pertain to the project of creating beneficial AI. If this sounds interesting to you, we hope that you will join in the conversations by following us or subscribing to our podcasts on Youtube, SoundCloud, or your preferred podcast site/application. In this podcast, Lucas spoke with Dylan Hadfield-Menell. Dylan is a 5th year PhD student at UC Berkeley advised by Anca Dragan, Pieter Abbeel and Stuart Russell, where he focuses on technical AI alignment research. Topics discussed in this episode include: -How CIRL helps to clarify AI alignment and adjacent concepts -The philosophy of science behind safety theorizing -CIRL in the context of varying alignment methodologies and it's role -If short-term AI can be used to amplify normative processes
What role does inverse reinforcement learning (IRL) have to play in AI alignment? What issues complicate IRL and how does this affect the usefulness of this preference learning methodology? What sort of paradigm of AI alignment ought we to take up given such concerns? Inverse Reinforcement Learning and the State of AI Alignment with Rohin Shah is the seventh podcast in the AI Alignment Podcast series, hosted by Lucas Perry. For those of you that are new, this series is covering and exploring the AI alignment problem across a large variety of domains, reflecting the fundamentally interdisciplinary nature of AI alignment. Broadly, we will be having discussions with technical and non-technical researchers across areas such as machine learning, governance, ethics, philosophy, and psychology as they pertain to the project of creating beneficial AI. If this sounds interesting to you, we hope that you will join in the conversations by following us or subscribing to our podcasts on Youtube, SoundCloud, or your preferred podcast site/application. In this podcast, Lucas spoke with Rohin Shah. Rohin is a 5th year PhD student at UC Berkeley with the Center for Human-Compatible AI, working with Anca Dragan, Pieter Abbeel and Stuart Russell. Every week, he collects and summarizes recent progress relevant to AI alignment in the Alignment Newsletter. Topics discussed in this episode include: - The role of systematic bias in IRL - The metaphilosophical issues of IRL - IRL's place in preference learning - Rohin's take on the state of AI alignment - What Rohin has changed his mind about
Pieter Abbeel is a professor at UC Berkeley, director of the Berkeley Robot Learning Lab, and is one of the top researchers in the world working on how to make robots understand and interact with the world around them, especially through imitation and deep reinforcement learning. Video version is available on YouTube. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, or YouTube where you can watch the video versions of these conversations.
Professor at UC Berkeley, Peiter Abbeel joins us. Pieter grew up in Belgium, came to the US and got his PhD in Robotics and Machine Learning from Stanford. He notes that he and Andrew Ng pushed the envelop at the time on how robots learn from humans demonstrations as well as their own trial and error. Peter graduated and came to Berkeley to continue to work on the junction of robotics and learning, machine learning. He’s been focused on end-to-end reinforcement learning, end-to-end imitation learning. Training the neural net end-to-end without specific structure. Singularity is the notion that a system you build is smart enough to self improve...and things accelerate out of control. How far are we away from this? Pieter notes that 10 years ago computer vision it was difficult to conceive of a solution. Enabling factors and breakthroughs are the keys. Data is an enabling factor. Neural nets are now data driven as opposed to algorithm designed. Will we continue to have more data and can you do things with unlabeled and unstructured data. Being better at unsupervised learning is a frontier that once reached will open up all sorts of possibilities. One question to answer in understanding where we might be in relation to singularity is how many compute cycles were effectively used to go from where we were 5 Billion years ago to where we are now. Do we think we can short cut this? Or will we need the same amount of compute to get to singularity? Pieter doesn’t have the answers. Yet. Both he and the artificial intelligence research community and industry need the best possible global talent to answer those very questions.
Robots today must be programmed by writing computer code, but imagine donning a VR headset and virtually guiding a robot through a task, like you would move the arms of a puppet, and then letting the robot take it from there. That's the vision of Pieter Abbeel, a professor of electrical engineering and computer science at the University of California, Berkeley, and his students, Peter Chen, Rocky Duan, Tianhao Zhang, who have launched a start-up, Embodied Intelligence, Inc., to use the latest techniques of deep reinforcement learning and artificial intelligence to make industrial robots easily teachable. Series: "UC Berkeley News" [Science] [Show ID: 33301]
Robots today must be programmed by writing computer code, but imagine donning a VR headset and virtually guiding a robot through a task, like you would move the arms of a puppet, and then letting the robot take it from there. That's the vision of Pieter Abbeel, a professor of electrical engineering and computer science at the University of California, Berkeley, and his students, Peter Chen, Rocky Duan, Tianhao Zhang, who have launched a start-up, Embodied Intelligence, Inc., to use the latest techniques of deep reinforcement learning and artificial intelligence to make industrial robots easily teachable. Series: "UC Berkeley News" [Science] [Show ID: 33301]
In this episode of the ARCHITECHT Show, Embodied Intelligence co-founder and chief scientist Pieter Abbeel talks about the newly launched company's mission to simplify the training of industrial robots using an artificial intelligence software layer and virtual reality headsets. Additionally, Abbeel, who's also a professor at UC-Berkeley and has worked at OpenAI, discusses the benefits of different research models and the economic effects of automation, and gives some guidance for companies trying to make sense of the fast-moving AI landscape.
This Week in Machine Learning & Artificial Intelligence (AI) Podcast
This week we continue our Industrial AI series with Sergey Levine, an Assistant Professor at UC Berkeley whose research focus is Deep Robotic Learning. Sergey is part of the same research team as a couple of our previous guests in this series, Chelsea Finn and Pieter Abbeel, and if the response we’ve seen to those shows is any indication, you’re going to love this episode! Sergey’s research interests, and our discussion, focus in on include how robotic learning techniques can be used to allow machines to acquire autonomously acquire complex behavioral skills. We really dig into some of the details of how this is done and I found that our conversation filled in a lot of gaps for me from the interviews with Pieter and Chelsea. By the way, this is definitely a nerd alert episode! Notes for this show can be found at twimlai.com/talk/37
This Week in Machine Learning & Artificial Intelligence (AI) Podcast
This week our guest is Pieter Abbeel, Assistant Professor at UC Berkeley, Research Scientist at OpenAI, and Cofounder of Gradescope. Pieter has an extensive background in AI research, going way back to his days as Andrew Ng’s first PhD student at Stanford. His research today is focused on deep learning for robotics. During this conversation, Pieter and I really dig into reinforcement learning, a technique for allowing robots (or AIs) to learn through their own trial and error. Nerd alert!! This conversation explores cutting edge research with one of the leading researchers in the field and, as a result, it gets pretty technical at times. I try to uplevel it when I can keep up myself, so hang in there. I promise that you’ll learn a ton if you keep with it. The notes for this show can be found at twimlai.com/talk/28
How does designing a voice recognition program like Siri compare to creating artificially intelligent robots? Computer science professor Pieter Abbeel of the University of California, Berkeley says that although both use similar deep learning algorithms, applying them to robots is far more challenging because the task goes beyond passively recognizing the sound of a voice. "So recognizing a pattern and then making a prediction, is very different than a robot taking an action, which then in turn, results in the world changing around the robot. And then taking an action that changed the world and that, then keeps going, right? So it's very different because for a robot, every action has consequences, whereas if you do something like image recognition or sound recognition, you just kind of output your prediction and then the next request comes in, to recognize something." Abbeel’s team has developed an algorithm that enables robots to learn on their own without preprogrammed solutions like those in voice recognition programs. "And learn that all in one learning process while it’s trying things out."