BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Problem with Deep Learning in Science (and how to fix it). - M
 iles Cranmer\, DAMTP
DTSTART:20231108T140000Z
DTEND:20231108T150000Z
UID:TALK208039@talks.cam.ac.uk
CONTACT:Amanda Stagg
DESCRIPTION:Would Kepler have discovered his laws if machine learning had 
 been around in 1609? Or would he have been satisfied with the accuracy of 
 some black box regression model\, leaving Newton without the inspiration t
 o find the law of gravitation? In this talk I will present a quick introdu
 ction to machine learning and two major issues facing its application in t
 he sciences: (1) a lack of interpretability\, and (2) no physical priors. 
 I will then present new potential solutions to each of these. For (1)\, I 
 will introduce a method for translating a neural network into a symbolic m
 odel\, using symbolic regression techniques such as PySR (github.com/Miles
 Cranmer/PySR). Then\, for (2)\, I will discuss the idea behind “foundati
 on models\,” which are large\, general models pre-trained on vast amount
 s of data\, endowing them with strong general priors – such as ChatGPT a
 nd Stable Diffusion. I will present “Polymathic AI\,” a new research c
 ollaboration (polymathic-ai.org) which aims to build an analogous foundati
 on model for scientific data\, and describe our recent work.
LOCATION:MR2
END:VEVENT
END:VCALENDAR