{"661886":{"#nid":"661886","#data":{"type":"event","title":"A Decade of Machine Learning Accelerators: Lessons Learned and Carbon Footprint","body":[{"value":"\u003Cp\u003EDavid Patterson, professor-emeritus at UC Berkeley and a Google distinguished professor, will give a lecture, \u003Cem\u003EA Decade of Machine Learning Accelerators: Lessons Learned and Carbon Footprints\u003C\/em\u003E,\u0026nbsp;on October 12. The talk will take place at 11 am with a question-and-answer session at 11:40, and it will be in rooms 1116-1118 of the Marcus Nanotechnology Building.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe success of deep neural networks (DNNs) from Machine Learning (ML) has inspired domain specific architectures (DSAs) for them. Google\u0026rsquo;s first generation DSA offered 50x improvement over conventional architectures for ML inference in 2015. Google next built the first production DSA supercomputer for the much harder problem of training. Subsequent generations greatly improved performance of both phases. We start with ten lessons learned from such efforts.\u0026nbsp;\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe rapid growth of DNNs rightfully raised concerns about their\u0026nbsp;carbon\u0026nbsp;footprint. The second part of the talk identifies the \u0026ldquo;4Ms\u0026rdquo; (Model, Machine, Mechanization, Map) that, if optimized, can reduce ML training energy by up to 100x and\u0026nbsp;carbon\u0026nbsp;emissions up to 1000x. By improving the 4Ms, ML held steady at \u0026lt;15% of Google\u0026rsquo;s total energy use despite it consuming ~75% of its floating point operations.\u0026nbsp; With continuing focus on the 4Ms, we can realize the amazing potential of ML to positively impact many fields in a sustainable way.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Ca href=\u0022https:\/\/en.wikipedia.org\/wiki\/David_Patterson_(computer_scientist)\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/en.wikipedia.org\/wiki\/David_Patterson_(computer_scientist)\u0022\u003EDavid Patterson\u003C\/a\u003E\u0026nbsp;is a UC Berkeley Pardee professor emeritus, a Google distinguished engineer, and the\u003Ca href=\u0022https:\/\/riscv.org\/\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/riscv.org\/\u0022\u003E\u0026nbsp;RISC-V International\u003C\/a\u003E\u0026nbsp;Vice-Chair. His most influential Berkeley projects likely were\u003Ca href=\u0022https:\/\/en.wikipedia.org\/wiki\/Reduced_instruction_set_computer\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/en.wikipedia.org\/wiki\/Reduced_instruction_set_computer\u0022\u003E\u0026nbsp;RISC\u003C\/a\u003E\u0026nbsp;and\u003Ca href=\u0022https:\/\/en.wikipedia.org\/wiki\/RAID\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/en.wikipedia.org\/wiki\/RAID\u0022\u003E\u0026nbsp;RAID\u003C\/a\u003E. His best-known book is\u003Ca href=\u0022https:\/\/www.amazon.com\/Computer-Architecture-Quantitative-Approach-Kaufmann\/dp\/0128119055\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/www.amazon.com\/Computer-Architecture-Quantitative-Approach-Kaufmann\/dp\/0128119055\u0022\u003E\u0026nbsp;\u003Cem\u003EComputer Architecture: A Quantitative Approach\u003C\/em\u003E\u003C\/a\u003E. He and his co-author John Hennessy shared the\u003Ca href=\u0022https:\/\/www.acm.org\/media-center\/2018\/march\/turing-award-2017\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/www.acm.org\/media-center\/2018\/march\/turing-award-2017\u0022\u003E\u0026nbsp;2017 ACM A.M Turing Award\u003C\/a\u003E\u0026nbsp;and the\u003Ca href=\u0022https:\/\/cacm.acm.org\/news\/257773-risc-chip-innovators-receive-2022-charles-stark-draper-prize-for-engineering\u0022 style=\u0022color: blue; text-decoration: underline;\u0022 target=\u0022_blank\u0022 title=\u0022https:\/\/cacm.acm.org\/news\/257773-risc-chip-innovators-receive-2022-charles-stark-draper-prize-for-engineering\u0022\u003E\u0026nbsp;2022 NAE Charles Stark Draper Prize for Engineering\u003C\/a\u003E. The Turing Award is often referred to as the \u0026ldquo;Nobel Prize of Computing\u0026rdquo; and the Draper Prize is considered a \u0026ldquo;Nobel Prize of Engineering.\u0026rdquo;\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"David Patterson, professor-emeritus at UC Berkeley and a Google distinguished professor, will give a lecture titled \u201cA Decade of Machine Learning Accelerators: Lessons Learned and Carbon Footprints,\u201d on October 12."}],"uid":"32045","created_gmt":"2022-10-06 15:20:09","changed_gmt":"2022-10-06 15:20:45","author":"Ben Snedeker","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2022-10-12T12:00:00-04:00","event_time_end":"2022-10-12T13:00:00-04:00","event_time_end_last":"2022-10-12T13:00:00-04:00","gmt_time_start":"2022-10-12 16:00:00","gmt_time_end":"2022-10-12 17:00:00","gmt_time_end_last":"2022-10-12 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"47223","name":"College of Computing"},{"id":"576481","name":"ML@GT"},{"id":"50877","name":"School of Computational Science and Engineering"},{"id":"50875","name":"School of Computer Science"},{"id":"50876","name":"School of Interactive Computing"}],"categories":[],"keywords":[],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1795","name":"Seminar\/Lecture\/Colloquium"}],"invited_audience":[{"id":"78761","name":"Faculty\/Staff"},{"id":"177814","name":"Postdoc"},{"id":"174045","name":"Graduate students"},{"id":"78751","name":"Undergraduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EProfessor Tom Conte\u003Cbr \/\u003E\r\n\u003Ca href=\u0022mailto:conte@gatech.edu\u0022\u003Econte@gatech.edu\u003C\/a\u003E\u003C\/p\u003E\r\n","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}