Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

.Jeff Dean 's full talk at NeurIPS 24 on Machine Learning for Systems and how Google thinks about these things

wh

17,732 subscribers

54,154 views • 1 year ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

5 Comments

LLMSalaryman1 year ago

@JeffDean Hats off to the researchers with the balls to name an unrelated lstm algo “llama” in 2024

dreaming android1 year ago

@JeffDean Do you have youtube link?

FSDwhisper1 year ago

@JeffDean Nice

Vibhu Pandey1 year ago

@JeffDean @dwarkesh_sp

Rekha@031 year ago

@JeffDean Great talk! What are the key takeaways?

Related Videos

4 questions with Jeff Dean (Chief Scientist, Google and one of the authors of the famous "cat paper" about unsupervised learning and neural networks in 2012) in under 2 minutes at #GoogleIO:

4 questions with Jeff Dean (Chief Scientist, Google and one of the authors of the famous "cat paper" about unsupervised learning and neural networks in 2012) in under 2 minutes at #GoogleIO:

Google

80,369 views • 3 months ago

In Episode 4 of Advanced Insights, Amin Vahdat, Engineering Fellow and VP of Machine Learning, Systems and Cloud AI at Google, and Mark Papermaster discuss the 5th Epoch of Distributed Computing, Moore’s Law, and more. Watch on YouTube:

In Episode 4 of Advanced Insights, Amin Vahdat, Engineering Fellow and VP of Machine Learning, Systems and Cloud AI at Google, and Mark Papermaster discuss the 5th Epoch of Distributed Computing, Moore’s Law, and more. Watch on YouTube:

AMD

13,406 views • 2 years ago

Google just released a free 1-hour lecture from Chief Scientist Jeff Dean on full AI Engineering. How to go from one LLM to coordinating 100 AI agents: 01:45 - Build an LLM from scratch 17:22 - Learn how to actually use AI models 30:03 - Master prompt engineering 52:35 - Coordinate 100 agents as one human Most people are still learning how to prompt one model. Jeff Dean is showing the full stack: LLMs → Prompts → Agents → Agent Teams → Graphs Prompting is the old workflow. Coordinating agent systems is the new one. This is 27 years of Google AI compressed into one hour. Bookmark and watch it today Then read the full Graph Engineering guide below ↓

Google just released a free 1-hour lecture from Chief Scientist Jeff Dean on full AI Engineering. How to go from one LLM to coordinating 100 AI agents: 01:45 - Build an LLM from scratch 17:22 - Learn how to actually use AI models 30:03 - Master prompt engineering 52:35 - Coordinate 100 agents as one human Most people are still learning how to prompt one model. Jeff Dean is showing the full stack: LLMs → Prompts → Agents → Agent Teams → Graphs Prompting is the old workflow. Coordinating agent systems is the new one. This is 27 years of Google AI compressed into one hour. Bookmark and watch it today Then read the full Graph Engineering guide below ↓

Lunar

102,011 views • 21 days ago

We’re hiring for machine learning, embedded systems, firmware, electrical, and mechanical engineering roles at Allen Control Systems systems. We’re based in Austin, Texas. We shoot down drones. Enjoy the weekend!

We’re hiring for machine learning, embedded systems, firmware, electrical, and mechanical engineering roles at Allen Control Systems systems. We’re based in Austin, Texas. We shoot down drones. Enjoy the weekend!

Steven Simoni

116,807 views • 1 year ago

Former Google Chief Scientist Jeff Dean was almost in tears while talking about his decision to leave Google after 27 years to launch an independent startup "i have incredible fondness for my time at google... amazing colleagues" 🥲

Former Google Chief Scientist Jeff Dean was almost in tears while talking about his decision to leave Google after 27 years to launch an independent startup "i have incredible fondness for my time at google... amazing colleagues" 🥲

Haider.

135,313 views • 9 days ago

THIS is the future of AI and creativity: Collaboration with the machine. Google collaborated with Lupe Fiasco to create tools to amplify how he works, writes and thinks. If you're a creative thinking about AI at all - these 6 minutes could change your entire creative future.

THIS is the future of AI and creativity: Collaboration with the machine. Google collaborated with Lupe Fiasco to create tools to amplify how he works, writes and thinks. If you're a creative thinking about AI at all - these 6 minutes could change your entire creative future.

Jeremy Somers 𝕏 NotContent

62,051 views • 3 years ago

Wondering what Agentic AI means for developers? In this GTC session, Bill Dally (Chief Scientist and SVP of Research, NVIDIA) and Jeff Dean (Chief Scientist, Google DeepMind and Google Research) discuss how agents can act like a senior mentor in your stack to rapidly deploy experiments and more. Advancing to AI's Next Frontier – Bill Dally and Jeff Dean #NVIDIAGTC

Wondering what Agentic AI means for developers? In this GTC session, Bill Dally (Chief Scientist and SVP of Research, NVIDIA) and Jeff Dean (Chief Scientist, Google DeepMind and Google Research) discuss how agents can act like a senior mentor in your stack to rapidly deploy experiments and more. Advancing to AI's Next Frontier – Bill Dally and Jeff Dean #NVIDIAGTC

NVIDIA GTC

40,623 views • 2 months ago

What's next for the future of robotics? Tom and Yuval — Google DeepMind Research Scientists (and long-time friends & collaborators) — shared their paths to Google and their insights on what lies ahead. Learn more about Google's work in artificial intelligence and machine learning ➡️

What's next for the future of robotics? Tom and Yuval — Google DeepMind Research Scientists (and long-time friends & collaborators) — shared their paths to Google and their insights on what lies ahead. Learn more about Google's work in artificial intelligence and machine learning ➡️

Life at Google

39,603 views • 4 months ago

Jeff Dean explaining Google built a ChatGPT-style chatbot roughly a year before OpenAI released theirs. Tens of thousands of Google employees were already using it and loved it. Leadership still killed any public launch because: • “What if people stop searching on Google?” • “What if it says inaccurate or ‘bad’ things?” Google invented the Transformer architecture, and still threw away the first-mover advantage for nothing. Innovator's Dilemma at its finest.

Jeff Dean explaining Google built a ChatGPT-style chatbot roughly a year before OpenAI released theirs. Tens of thousands of Google employees were already using it and loved it. Leadership still killed any public launch because: • “What if people stop searching on Google?” • “What if it says inaccurate or ‘bad’ things?” Google invented the Transformer architecture, and still threw away the first-mover advantage for nothing. Innovator's Dilemma at its finest.

Linas Beliūnas

391,126 views • 23 days ago

We don’t talk enough about compatibility. Everyone obsesses over size, performance, and “how to last longer” but barely anyone talks about fit. Two bodies, two nervous systems, learning each other. Get the book now:

Sensitive content

We don’t talk enough about compatibility. Everyone obsesses over size, performance, and “how to last longer” but barely anyone talks about fit. Two bodies, two nervous systems, learning each other. Get the book now:

Caitlin V

72,289 views • 4 months ago

Mark and Hugh talk about Remigration and how we achieve it. “We have to prepare for backlash, millions won’t just leave…. The systems that brought these people here have to be dismantled.”

Mark and Hugh talk about Remigration and how we achieve it. “We have to prepare for backlash, millions won’t just leave…. The systems that brought these people here have to be dismantled.”

Hugh Anthony

19,020 views • 3 months ago

New ways to personalize learning? Yes, please! 🙌 Google Gemini and @NotebookLM are giving K-12 education a huge boost. See how these four schools are using Google AI to create personalized learning experiences for every student. Learn more:

New ways to personalize learning? Yes, please! 🙌 Google Gemini and @NotebookLM are giving K-12 education a huge boost. See how these four schools are using Google AI to create personalized learning experiences for every student. Learn more:

Google for Education

18,569 views • 1 year ago

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Jürgen Schmidhuber

232,164 views • 5 months ago

Ex-Google Jeff Dean just dropped a 1-hour lecture on full AI engineering: from LLMs to Graphs 1:45 - LLM from scratch 17:22 - How to use AI models 30:03 - Prompt engineering 52:35 - One human coordinating 100 agents 27 years of building AI at Google, compressed into one hour. Watch it, then read the full guide on Graph engineering below.

Ex-Google Jeff Dean just dropped a 1-hour lecture on full AI engineering: from LLMs to Graphs 1:45 - LLM from scratch 17:22 - How to use AI models 30:03 - Prompt engineering 52:35 - One human coordinating 100 agents 27 years of building AI at Google, compressed into one hour. Watch it, then read the full guide on Graph engineering below.

Mahax

89,485 views • 13 days ago

Microsoft Chief Scientific Officer Eric Horvitz on what stood out at NeurIPS 2025, what’s next, and why these moments matter for the future of AI research.

Microsoft Chief Scientific Officer Eric Horvitz on what stood out at NeurIPS 2025, what’s next, and why these moments matter for the future of AI research.

Microsoft Research

86,914 views • 8 months ago

👀 Google (👓 actually) snuck in a demo for a pair of AR glasses (codenamed Project Iris). They didn't show us much or talk about it at all, but they look to be in regular glasses form factor. I can't wait to have these things on my face.

👀 Google (👓 actually) snuck in a demo for a pair of AR glasses (codenamed Project Iris). They didn't show us much or talk about it at all, but they look to be in regular glasses form factor. I can't wait to have these things on my face.

Sheel Mohnot

39,447 views • 2 years ago

Security is a must for the development of AI and machine learning — and vice versa Dawn Song, Founder of Oasis Labs, explains how AI models can benefit security research View the full interview — where Dawn shares her journey into AI, Machine Learning, and web3 — below:

Security is a must for the development of AI and machine learning — and vice versa Dawn Song, Founder of Oasis Labs, explains how AI models can benefit security research View the full interview — where Dawn shares her journey into AI, Machine Learning, and web3 — below:

Oasis Labs

33,169 views • 3 years ago

New Unsupervised Learning with Joshua Meier @ NeurIPS (Chai Discovery) on how AI x drug discovery is starting to inflect and: - The 3 waves of AI biotech - How AI models think differently than chemists - Building "Photoshop for molecules" - Deciding what to open source

New Unsupervised Learning with Joshua Meier @ NeurIPS (Chai Discovery) on how AI x drug discovery is starting to inflect and: - The 3 waves of AI biotech - How AI models think differently than chemists - Building "Photoshop for molecules" - Deciding what to open source

Jacob Effron

12,351 views • 1 year ago