To put the "scale" narrative into perspective... The brain runs on 15 watts, at 8-35 hertz. And while we have ~90B neurons, usually only ~1B are active at any given time. The brain is very slow and does a lot with very little.

We (Barbara Engelhardt, Naila Murray and I) are proud to announce the creation of a Journal-to-Conference track, in collaboration with JMLR and conferences NeurIPS 2022, ICLR 2023 and ICML 2023! neurips.cc/public/Journal… iclr.cc/public/Journal… icml.cc/public/Journal…

Starting a new position as Principal Researcher at Microsoft Research (MSR) at Microsoft! lnkd.in/gUQwmKYs

AutoMoE: Neural Architecture Search for Efficient Sparsely Activated Transformers abs: arxiv.org/abs/2210.07535

We are hiring research interns to work in Efficient Large-scale AI at Microsoft Research for summer 2023! Looking for candidates interested in building adaptive, modular and efficient NLU/NLG models at scale. The internship will be…lnkd.in/gn3ddFsm lnkd.in/gDBreieR

We kick-off the near year with a talk by Subhabrata Mukherjee about ‘AutoMoE: Neural Architecture Search for Efficient Sparsely Activated Transformers’ on Thursday. Note that, this time the AutoML seminar will be at 6pm CET / 9am PDT. More details at automl-seminars.github.io

Do you want to make your transformers more efficient? Check out Subhabrata Mukherjee talk on ‘AutoMoE: Neural Architecture Search for Efficient Sparsely Activated Transformers’: youtu.be/sSusEYtL-YM Don’t forget to subscribe!

Bothered by the expensive runs on Auto-GPT and LangChain agents? Check out our recent work, ReWOO, that eliminates token redundancy in prevailing 'thought-action-observation' paradigm, achieving better task completion with 5x less token usage at inference. #LLMs #AIGA #NLP

Source: Conversation with Bing Chat about Orca - 6/23/2023   We asked Bing about our recent language model Orca, what makes its learning process unique and the promises it holds for the future. This is its response (references in…lnkd.in/g6tXh8kc lnkd.in/gTPRXtjy

SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inference Obtains 2-5x inference speedups with negligible regression across a variety of tasks arxiv.org/abs//2307.02628

We are truly excited to find Eric Topol summarizing #Polaris in his report. Read about our LLM constellation work for real-time patient-AI voice conversations in erictopol.substack.com/p/a-big-week-i… Preprint: arxiv.org/abs/2403.13313 #GenerativeAI #healthcare hippocraticai

Hippocratic is Augmenting Repetitive Healthcare Tasks with AI linkedin.com/pulse/hippocra…

We are super excited to have Subhabrata Mukherjee, Chief Scientific Officer & Co-founder at hippocraticai to join us next Tuesday! He will share how to build a safety-focused healthcare AI agent focusing on non-diagnostic patient-facing use-case!

We hosted our annual AI Pioneers Summit this week to celebrate the technical leaders at the forefront of deploying AI and LLMs in production today! 🚀 Over 170 notable product and engineering execs from 136 companies gathered to celebrate this year’s award winners: ***Expanding