Demis Hassabis & Fei-Fei Li Are Up Beat On AI
Written by Mike James   
Wednesday, 17 May 2023

With so many doomsayers around at the moment it is almost a refreshing breeze to hear people talk about the positive side of AI. In a video Demis Hassabis and Fei-Fei Li talk about AI and Science.

Demis Hassabis is the cofounder of DeepMind - what we usually refer to as Google's DeepMind. It has done a lot of neural network implementation that made headlines - Alpha Go probably being the best known. At the time Go was a game thought to be well outside the reach of AI, but Alpha Go, a neural network trained using reinforcement learning, beat the best human player Lee Sedol in 2016. From here there were more game-playing ground-breakers, but more recently the company has developed neural networks that help with scientific and medical discoveries - AlphaFold being the best known a network for deducing 3D protein structure from empirical formulas. This fits in with Hassabis's claim that his company would "solve intelligence" and then use intelligence "to solve everything else".

Fei Fei Li isn't as well known for headline-making breakthroughs, but back in 2010 she developed ImageNet, a labeled dataset of photos that has been used to train the most successful vision networks. This may not sound like much, but the effort wasn't that small and to realize that such a thing might have a payoff wasn't obvious. After a spell with Google, she has spent time and energy on education and ethics.



Li and Hassabis shared the stage at Stanford for a presentation on Using AI to Accelerate Scientific Discovery in April 2023 and, as you might expect, the outlook is upbeat. After the usual introductions we get to the uses of AlphaFold which are many more than you might have imagined. On to ideas of how digital science is science speeded up - especially digital biology.

Later Hassabis talks about large language models, which are all the rage and suggests that they are not the solution to general AI - even if we make them bigger - but they are a component in the more complete solution.


And don't miss the discussion at the end - the most interesting part of the video.

The revolution caused by AI may be causing doubts and worries about the future but surely the increased speed of science is all to the good?


More Information

Related Articles

Can Google DeepMind Win At AI?

All You Wanted To Know About AI From DeepMind

Can DeepMind's Alpha Code Outperform Human Coders?

AlphaFold Reads The DNA

AlphaFold Solves Fundamental Biology Problem

AlphaFold DeepMind's Protein Structure Breakthrough

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.



Andrew Tanenbaum Gains ACM Award

Andrew Tanenbaum has been awarded the 2023 ACM System Software Award for MINIX the operating system he created for teaching purposes and which was an important influence on Linux.

Pg_lakehouse Makes PostgreSQL Quack

Pg_Lakehouse from ParadeDB is an extension that turns PostgreSQL into the analytical engine of DuckDB. Why is that useful? How do you use it?

More News


Last Updated ( Wednesday, 17 May 2023 )