AI Updates Council: A Weekly Recap of Major AI Changes and Themes

•

In less than 24 hours, we will convene the AI updates council, to discuss every MAJOR AI thing that happened in the past week. twitter.com/i/spaces/1OyJAVYddeMxb If you don't have the patience to read ALL OF TWITTER for the past week ,join us to discuss the major AI changes and themes 🧵

Covering the following: Open AI's Process Supervision paper, better Math understanding, potentially better factuality for future models? twitter.com/altryne/status/1663958551478534144

Sam is on a world tour, Open AI roadmap for the future - which was discussed and then taken down from HumanLoop Raza Habib blog: • website-754fwhahs-humanloopml.vercel.app/blog/open_ai_talk?utm_source=bensbites&utm_medium=referral&utm_campaign=openai-s-roadmap • Falcon has removed royalties, best performing fully open sourced model

UAE's Falcon 40B, the world's top ranked open-source AI model from the Technology Innovation Institute (TII) has waived royalties on its use for commercial and research purposes. #TII #LLM #FalconLLM #Tech #Innovation #AI #AbuDhabi #UAE t.co/Qr4D8NxROz twitter.com/TIIuae/status/1663911042559234051

And it's coming to #GPT4ALL as well as plugins twitter.com/andriy_mulyar/status/1664318409801191425

• StyleDrop from @natanielruizg and the team at Google - allows for a zero shot personalization and style transfer. twitter.com/natanielruizg/status/1664653573597065225

@Teknium1 (AKA the AI shepherd) released Nouns-Hermes, a fine tuned LLaMa that outperforms some incredible open source models (and it's available in GPT4All as well) twitter.com/Teknium1/status/1664844441742618629

@clattner_llvm gave a full deep dive on Mojo and the Modular ecosystem with @lexfridman What can we expect from Mojo? Also, we found out that Chris is incredibly based 👏 twitter.com/altryne/status/1665053168206770177

Training on synthetic images can lead to better results than on real ones. This together with distillation means that AI can basically self improve? twitter.com/dilipkay/status/1664580057073942529

Apple not using the word AI even once: We will cover "ONLY" the AI stuff, none of that Vision talk, there has been more than enough of that going around 😂 I will moderate this so if you're tired of the Vision discussion... gotchu twitter.com/altryne/status/1665776272734289920

5 Million tokens context windows? I mean at some point let's call it "infinite context windows" right? We'll discuss what this means, and what this unlocks. twitter.com/magicailabs/status/1666116935904292869

@Tim_Dettmers with another banger, reshuffling the field of inference with an crazy algorithmic way to optimize inference by 15% and significantly reduce model sizes twitter.com/Tim_Dettmers/status/1666076553665744896

And as always, we'll have a panel of experts, @transitive_bs @xenovacom @JunaidDawud and guest speakers join, discuss and answer questions. Come hang out with us if the time works for you, the spaces will be recorded if the time doesn't work! twitter.com/i/spaces/1OyJAVYddeMxb

If you can't make it, help by RTing and sending this thread to your friends, it really helps 🫡 oh and PS, we may send an @getairchat invite or two for some of the folks who come up and speak :) No promises, but we'll see. See you tomorrow!