AI Tidbits

Share this post

AI Roundup 11/17 -> 11/24/2022

aitidbits.substack.com

AI Roundup 11/17 -> 11/24/2022

Last week's most interesting AI news in <2 minutes

Sahar Mor
Nov 24, 2022
Share this post

AI Roundup 11/17 -> 11/24/2022

aitidbits.substack.com
  1. Meta unveils CICERO - an AI surpassing human-level performance at diplomacy and negotiation (Meta AI)

  1. Stability AI releases Stable Diffusion 2.0 which supports new inpainting capabilities, higher resolution image generation, and a Depth-to-Image diffusion model (Company blog)

  1. A group of researchers from Weizmann Institute of Science presents SinFusion - a diffusion model trained on a single image or a video capable of generating video extrapolations and variations (Paper's website)

  1. Researchers present MinD-Vis - a diffusion model that reconstructs highly plausible images from brain recordings (Paper's website)

  1. Sebastian Ruder publishes a comprehensive outlook on the state of multilingual AI across NLP, computer vision, and speech (Blog)

  1. Meta AI and researchers from several universities release TART - a multi-task retrieval system that can follow human-written instructions to find the best documents for a given query, outperforming models up to three times larger (Twitter)

  1. Multi, a new generative AI startup, releases MultiFlow to easily create, deploy, and iterate on generative AI workflows (Twitter)

  2. Google is rumored to be working on Pitchfork - an AI model built for "teaching code to write and rewrite itself" by learning programming styles and then writing new code based on those learnings (Paywall - Business Insider)

  3. The Stanford AI Center for Research on Foundation Models (CRFM) announces Holistic Evaluation of Language Models (HELM) a new project to "serve as a map for the world of language models" as they grow in popularity (VentureBeat)

  1. Researchers from MIT introduce a new multi-task benchmark for evaluating text-to-image models (Paper)

  1. Researchers from The Chinese University of Hong Kong propose Uni-Perceiver v2 - a generalist model outperforming all existing generalist models in both versatility and performance (Paper)

  1. Nvidia releases Magic3D - a novel text-to-3D model to power its metaverse builder platform Omniverse (Ars Technica)

  1. OpenAI's startup fund leads a $5M round in Harvey, an AI startup developing a natural language platform that answers legal questions and performs legal research (TechCrunch)

Last week’s AI Tidbits

Thanks for reading AI Tidbits! Stay up to date with the latest in AI in <2 minutes a week.

Share this post

AI Roundup 11/17 -> 11/24/2022

aitidbits.substack.com
Comments
TopNewCommunity

No posts

Ready for more?

© 2023 Sahar Mor
Privacy ∙ Terms ∙ Collection notice
Start WritingGet the app
Substack is the home for great writing