{"id":889512,"date":"2022-10-24T09:00:24","date_gmt":"2022-10-24T16:00:24","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-video&p=889512"},"modified":"2022-10-25T06:31:23","modified_gmt":"2022-10-25T13:31:23","slug":"research-talk-differentially-private-fine-tuning-of-large-language-models","status":"publish","type":"msr-video","link":"https:\/\/www.microsoft.com\/en-us\/research\/video\/research-talk-differentially-private-fine-tuning-of-large-language-models\/","title":{"rendered":"Research talk: Differentially private fine-tuning of large language models"},"content":{"rendered":"

We have come a long way in terms of protecting privacy when training ML models, particularly with large language models. We recently demonstrated that using differentially private stochastic gradient descent (DP-SGD) to fine-tune very large language models, such as GPT-3, is not only feasible but shows very promising results with respect to the privacy-utility tradeoff. In this talk, we highlight the challenges we have overcome over the past year and the opportunities our research enables for a range of product applications.<\/p>\n

Learn more about the 2022 Microsoft Research Summit ><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

We have come a long way in terms of protecting privacy when training ML models, particularly with large language models. We recently demonstrated that using differentially private stochastic gradient descent (DP-SGD) to fine-tune very large language models, such as GPT-3, is not only feasible but shows very promising results with respect to the privacy-utility tradeoff. […]<\/p>\n","protected":false},"featured_media":889515,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"research-area":[13556,13558],"msr-video-type":[266193,266181],"msr-locale":[268875],"msr-impact-theme":[261676],"msr-pillar":[],"class_list":["post-889512","msr-video","type-msr-video","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-research-area-security-privacy-cryptography","msr-video-type-building-user-trust-through-privacy-identity-and-responsible-ai","msr-video-type-microsoft-research-summit-2022","msr-locale-en_us"],"msr_download_urls":"","msr_external_url":"https:\/\/youtu.be\/Qe8WsNiL6RA","msr_secondary_video_url":"","msr_video_file":"","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/889512"}],"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\/889512\/revisions"}],"predecessor-version":[{"id":889518,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/889512\/revisions\/889518"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/889515"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=889512"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=889512"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=889512"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=889512"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=889512"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=889512"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}