SLAC Public Lecture | How scientists are building the AI-powered laboratory

[JeffKeryk]

Starts at 7:00 PM Pacific Time.
Here is the overview in case you are interested:

"How do we find the next super material? How do we discover the next breakthrough drug? How do we unlock the mystery of dark matter and the structure of the universe? Scientists develop theories about these questions and test those theories with two powerful tools: laboratory experiments and computational models. But these tests can be costly: Each experiment can use hours or days of person-power on specialized equipment, and simulations can take days or weeks on a supercomputer. Access to these resources is limited; every measurement needs to count. This talk explores how scientists are using artificial intelligence to accelerate discovery, not only to analyze data more effectively, but to actively guide choices about the best next measurements to take to meet scientific goals. Along the way, I will explain what is unique about using AI for scientific research, share examples from my own work, and offer a glimpse of where AI-powered science is headed."

 

[JeffKeryk]

Enjoyed the lecture. The speaker spoke a little too rapidly and could be hard to follow. The lecture spoke of AI's benefit to speed up Scientific experimentation by 10 to 100 times by Modeling.

Some of the best stuff came out of the questions afterward. A school teacher asked about the difference between LLMs and other types of AI. I found this a great question, as it seems to me AI is this big nebulous thing people hate, are afraid of, think is a bubble and basically have no concrete idea about.

I may say this badly, but here's my SWAG:
LLMs are language based associations based on text. I think most people think this is AI.
Math AI is more number computational based, like Linear Regression models for example.

Of course there is much more, but Language and Math AI are different, and there are other types of non-language AI.
Scientists at SLAC and Stanford are excited about AI's promise to speed up discovery, etc.