{"id":929718,"date":"2023-02-20T11:00:50","date_gmt":"2023-02-20T19:00:50","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-video&#038;p=929718"},"modified":"2023-03-22T11:13:02","modified_gmt":"2023-03-22T18:13:02","slug":"physics-of-ai","status":"publish","type":"msr-video","link":"https:\/\/www.microsoft.com\/en-us\/research\/video\/physics-of-ai\/","title":{"rendered":"Physics of AI"},"content":{"rendered":"<p>We propose an approach to the science of deep learning that roughly follows what physicists do to understand reality: (1) explore phenomena through controlled experiments, and (2) build theories based on toy mathematical models and non-fully- rigorous mathematical reasoning. I illustrate (1) with the LEGO study (LEGO stands for Learning Equality and Group Operations), where we observe how transformers learn to solve simple linear systems of equations. I will also briefly illustrate (2) with an analysis of the emergence of threshold units when training a two-layers neural network to solve a simple sparse coding problem. The latter analysis connects to the recently discovered Edge of Stability phenomenon.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>We propose an approach to the science of deep learning that roughly follows what physicists do to understand reality: (1) explore phenomena through controlled experiments, and (2) build theories based on toy mathematical models and non-fully- rigorous mathematical reasoning. I illustrate (1) with the LEGO study (LEGO stands for Learning Equality and Group Operations), where [&hellip;]<\/p>\n","protected":false},"featured_media":929721,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"research-area":[13556],"msr-video-type":[],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-929718","msr-video","type-msr-video","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us"],"msr_download_urls":"","msr_external_url":"https:\/\/youtu.be\/XLNmgviQHPA","msr_secondary_video_url":"","msr_video_file":"","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/929718"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-video"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/929718\/revisions"}],"predecessor-version":[{"id":929727,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/929718\/revisions\/929727"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/929721"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=929718"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=929718"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=929718"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=929718"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=929718"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=929718"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}