Q&A with Google Chief AI Scientist Jeff Dean about the evolution of Google Search, TPUs, coding agents, balancing model efficiency and performance, and more
Latent.Space
Related Coverage
Discussion
-
@jeffdean
Jeff Dean
on x
Thanks for having me on the @latentspacepod, @swyx and @FanaHOVA! I enjoyed our discussion! Site with summary: https://latent.space/... Video: https://www.youtube.com/...
-
@swyx
@swyx
on x
the jeff dean pod is doing very well, grateful for all the positive shoutouts from u guys 🙇♂️ some fun lore: we're trying to learn the youtube game so everything is a/b tested. allen did a slop @NanoBanana thumbnail and i made 2 human clear ones, higher effort, more relevant. [i…
-
@swyx
@swyx
on x
excited to follow up our Gemini Deep Think coverage of @yitayml and now @JeffDean, and in my other capacity, host @_philschmid and other fantastic @GoogleDeepMind colleagues at AIE Europe! my favorite part of the Jeff conversation was how he gamely handled everything I could [ima…
-
@latentspacepod
@latentspacepod
on x
“Benchmarks, particularly external ones that are publicly available, have their utility, but they often have a lifespan of utility... I like to think the best kinds of benchmarks are ones where the initial scores are like 10% to 20% or 30% and then you can sort of work on [video]
-
@latentspacepod
@latentspacepod
on x
From rewriting Google's search stack in the early 2000s to reviving sparse trillion-parameter models and co-designing TPUs with frontier ML research, Jeff Dean has quietly shaped nearly every layer of the modern AI stack. As Chief AI Scientist at Google and a driving force behind…
-
@dhtikna
@dhtikna
on x
Bruh, im only 23 minutes in and this podcast by @JeffDean is already OMEGA MOGGING the @DarioAmodei one. VERY Info dense and so interesting
-
@dhtikna
@dhtikna
on x
Deepseek mentioned [18:20], DeepseekOCR. One mentioned passingly in a larger point about how some modalities are more imp than others like vision which can encode both text and audio. So expected Jeff didnt address the whale specifically