Ed H. Chi

Research VP @ GoogleDeepMind. ACM Fellow.

[Massachusetts Institute of Technology (MIT) @MIT]

We’ve lost an icon. Margaret Hamilton, a computer science pioneer best known for her work at MIT leading development of the flight software for the Apollo missions to the moon, has died at 90. Read about her legacy via MIT News: https://t.co/ODkbdIsoHw

[Sundar Pichai @sundarpichai]

Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon!

It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback.

Here’s a look at the benchmarks:

[Ed H. Chi @edchi]

@latentjasper Sorry to hear this...

[Ed H. Chi @edchi]

@YaoQin_UCSB @ai_ucsb @ucsantabarbara Congrats!

[Ed H. Chi @edchi]

Collecting my fanboi photo with Field Medalist Terrance Tao.

#aasf at @Stanford

[Ed H. Chi @edchi]

@OriolVinyalsML @quocleix I once did a photography show at that exact cafe, btw.

[Ed H. Chi @edchi]

@Stanford @JeffDean @xdh It's crazy to think this was just a little more than 3 years ago.

[Ed H. Chi @edchi]

At the #AASF forum @Stanford, in introducing @JeffDean, @xdh said "it is _not_ the size of the ship, but rather it is the ocean that you are navigating in that really matters." (My rephrasing)

Beautiful metaphor, as Jeff has certainly navigated amazing oceans.

Miss you, Jeff!

[Ed H. Chi @edchi]

@Stanford @JeffDean @xdh @quocleix @HengTze @sundarpichai Meanwhile, @quocleix and I, and the lamda team was thinking about creativity, and @denny_zhou and I were thinking further more about reasoning and Chain-of-thought implications.

That was an exciting time for launching large reasoning machines.

[Ed H. Chi @edchi]

@Stanford @JeffDean @xdh I was heads down and deep in the Lamda project with @quocleix and @HengTze and call into action by @sundarpichai to help launch it in 100 days, which became the Bard experiment.

We rapidly scaled down the lamda model from 128B to 35B, but Jeff was already thinking about Gemini.

[Ed H. Chi @edchi]

Berkeley RDI's AI Summit has become one of the best events in the bay area for AI. I flew back from Taiwan just in time to participate. Was fun to to discuss Frontier problems with Igor, Richard, and Dogus.

[Dawn Song @dawnsongtweets]

11/15 Frontier Research
What research breakthroughs will define the next generation of AI systems and models?

Join Richard Socher @RichardSocher, Co-Founder and CEO @Recursive_SI and @youdotcom; Ed Chi @edchi, VP of Research @GoogleDeepMind; Ekin Dogus Cubuk @ekindogus, Co-Founder @periodiclabs; Krithik Puthalath, CEO @ZyphraAI; and Andi Peng @TheAndiPenguin, Co-Founder @humansand, and Igor Babuschkin @ibab, Co-Founder and CEO @river_ai_inc, on new frontiers of AI.
👇 Help shape the conversation in this session: What questions would you like the speakers to tackle? What topics would you like to see covered? Reply below to help guide the discussion!

[Sundar Pichai @sundarpichai]

Very happy to support this on behalf of Google. We have long benefited from open source, are big contributors to open source and in fact have consistently made open weights models with Gemma available from @GoogleDeepMind @demishassabis . Onwards!

[Ed H. Chi @edchi]

A lot of people don't seem to realize the following:
Large model (with CoT baked-in) + harness = neural symbolic system.

[Ed H. Chi @edchi]

@growwithsanch Currently, the available evidences suggest that agent frameworks can only improve creativity performance up to a threshold before it appears to plateau. I would surmise that both are necessary in the future.

[Ed H. Chi @edchi]

Increasingly, I'm focusing my research on the creativity question: how can reasoning and chain-of-thought (schemata) might lead to "creativity" in large models.

Answer is more complicated than I expected---better memory mechanisms, planning, connecting novel knowledge.

[Ed H. Chi @edchi]

@YiTayML @qx_dong and it's awesome that she is so eager to discover new things! Great hire!

[Ed H. Chi @edchi]

@RISignal Yes, spot on!

[Ed H. Chi @edchi]

We need more research of reasoning under uncertainty in LLMs. We cannot get to AGI without it. Our progress toward AGI is in trouble because we're over invested in deductive reasoning.

[Ed H. Chi @edchi]

@LichangChen2 In specific domains, current frontier models long surpassed human intelligence. We don't call them AGI-capable already is because the models still make simple reasoning mistakes that are often unreasonable.

[Ed H. Chi @edchi]

@JohnYue122333 Agree. Many problems are under-specified and ambiguous, and there are not enough environmental constraints known to fully specify which solution is the best.