AI
State of AI Report 2020
|
Giphy/Boston Dynamics
|
Every year, the algorithm that runs this newsletter
eagerly awaits an annual AI report from
U.K. investors Nathan Benaich and Ian Hogarth. Their 2020 findings didn’t
disappoint. Research
Language models are improving and getting really big. They tend to
be developed by large organizations and distinguished by hefty training
costs. OpenAI’s GPT-3 has over 10x the parameters of the next biggest
language model, which was developed by Microsoft. Biology is having a breakout “AI moment,” with an
explosion of research publications this year and last year involving AI
methods. Privacy-preserving AI research has also ballooned, with more papers
in the H1 2020 than all of 2019 that mention federated learning. A drawback: “We’re rapidly approaching
outrageous computational, economic, and environmental costs to gain
incrementally smaller improvements in model performance.” The human brains
powering everything
The exodus from academia to Big Tech continues, which
may affect graduates’ entrepreneurship rates. The U.S. remains a magnet for
AI talent, meaning some countries are experiencing brain drain.
Ethics and politics
Facial recognition is already widely deployed around
the world. This year it really became a hot-button issue, with Facebook settling a
class-action lawsuit, multiple U.S. municipal bans, and large companies like
Microsoft and Amazon reconsidering facial
recognition work. Then, there’s the Washington-Beijing tech tensions.
Those get enough airtime in this newsletter, so we’ll move on: The report
points to a rising tide of “AI nationalism,” governments limiting foreign
takeovers of superstar tech startups. The future
Benaich and Hogarth provide eight predictions for the
next year. File the sauciest one—Nvidia doesn’t complete its Arm acquisition—under
AI nationalism. Other predictions: A top tech company will
shutter its AI lab in a strategic reset. Google’s DeepMind will make a
breakthrough in biology-drug discovery, while Facebook will make one with computer
vision and AR/VR. An AI-first drug discovery startup exits for $1+
billion. We’ll see you back here in 2021 to check on those
predictions. In the meantime, you can read the Brew’s AI guide for
deeper industry breakdowns. |
No comments:
Post a Comment