Google Research in 2021 and Beyond

(T) As he did since 2017, Jeff Dean, the tech lead of the Google Research and Health Teams, and probably one of the most famous Silicon Valley engineers, shared some of the key achievements of his team for 2021 in a recent blog post:

But this year, instead of listening the key achievements of Google Research one after another one, Mr. Dean organized them in themes that I pretentiously tried to summarize with a few keywords (meaning you should read the full blog post or at least the sections that interest you):

  • Trend 1: More Capable, General-Purpose ML Models = {larger and sparse transformers ~ billions to trillions parameters, transfomers for everything, improved GANs outputs, multiple modal models (inputs can mix languages, images, speeches videos), long-term dream = pathway = general-purpose models with multiple modalities of data generalized to solve thousands or millions of tasks}
  • Trend 2: Continued Efficiency Improvements for ML = { Meta learning algorithms, TPUv4, NAS, sparsity & mixture of experts}
  • Trend 3: ML Is Becoming More Personally and Communally Beneficial = {TF processor on Pixel, improvements in cameras & pictures and speech recognition for many user interactions}
  • Trend 4: Growing Benefits of ML in Science, Health and Sustainability = { Computer systems = ML to design chips; Health = ML in genomics, medical imagery, daily health & well being; Climate Crisis = ML for reducing carbon emissions, ML for stable plasmas for nuclear fusion, preventing wildfires and floods}
  • Trend 5: Deeper and Broader Understanding of ML = {responsible AI, open data sets, contributing research to the community}

Reference: Google Brain Research in 2020

Note: The picture above is a fishing boat from Half-Moon Bay coming back from fishing crabs.

Copyright © 2005-2022 by Serge-Paul Carrasco. All rights reserved.
Contact Us: asvinsider at gmail dot com.