Not necessarily. For example, you might learn that value adaptively based on observed behavior of the system (and therefore respond to dynamic changes), or learn different settings of that value for different machines on which your system is being run.
@jeffdean
-
Fan recognition for researcher Rubenia Borge’s research and teaching
By
–
I'm a fan of @Rubenia_Borge! We've exchanged many messages, and I've followed her progress on her research, her excitement of teaching, and her all around good nature.
-
AI Translation and Captioning Features Unite Global Communication
By
–
Great to see how well integrated features like automatic translation and captioning can expand how people meet and work together effectively (or just casually chat). As the number of languages increases, we're all brought a little bit closer together. https://t.co/Zh5tljbcYw
— Jeff Dean (@JeffDean) 1 septembre 2023Great to see how well integrated features like automatic translation and captioning can expand how people meet and work together effectively (or just casually chat). As the number of languages increases, we're all brought a little bit closer together.
-
Learning Heuristics: ML Systems Symbiosis Approach
By
–
I agree with @finbarrtimbers
: the potential to replace hand-written heuristics with ones that are learned is vast! I gave a talk about this at the 2018 MLSys: "Systems and Machine Learning Symbiosis" (ML for systems part starts at about 17:30). https://
youtu.be/Nj6uxDki6-0?si
=_vHLwDoc4hFuqJ2N
… -
ML Replaces Hand-Written Heuristics in Compilers
By
–
There's a huge opportunity to use ML to replace hand-written heuristics in compilers. Want to try? Check out this @kaggle contest organized by Ashley Chow, Bryan Perozzi, HCL-Jevster, inversion, Mangpo Phothilimthana, and Sami Abu-El-Haija!
-
Google Cloud Unveils MultiSlice Innovation with Nisha Johnson
By
–
The great work here is from @nisha_m_johnson and so many other colleagues in @googlecloud and elsewhere across Google who developed MultiSlice!
-
Google Cloud MultiSlice Scales ML Training Across TPU Pods
By
–
The @googlecloud MultiSlice system (see blog post in tweet below) is an external version of some internal software+hardware we've been using for a while to scale some of our largest ML training runs across multiple TPUv4 and TPUv5e pods. Scaling made easy!
-
Google TPUv4 System and Reconfigurable Optical Network Explained
By
–
A nice writeup about the TPUv4 system based on a talk my colleagues Norm Jouppi and Andy Swing delivered at @hotchipsorg 2023 on Tuesday. Learn more about the reconfigurable optical network!
-
US AI Systems Break One Hour Per Day Performance Barrier
By
–
Meanwhile breaking the one hour per day barrier looks possible for the US. I'd say more but we have to eat and run.