A monthly recap of news, product announcements, beta leaks, and other interesting Generative AI action that caught our attention. Join our newsletter and receive this recap straight to your Inbox.

Proudly curated by Dr. Daniel Moskovich.

Past issues:

February 2024

January 2024

December 2023

This month’s highlights:

  • AI21 Labs has released JAMBA, a production-level MAMBA implementation. MAMBA is an alternative to transformers, technologically different and therefore interesting. For me, this is the news of the month, because I like the MAMBA concept (it comes from control theory, a field I have a personal affinity with), and because AI21 Labs is co-founded by my fellow Jerusalemite Prof. Amnon Shashua (and Mobileye‘s Co-Founder).
  • IDEA has released T-Rex2, a very impressive-looking zero-shot object detection tool. No need to fine-tune and you can tell it what to look for using its text or image. It looks impressive! It’s licensed only for non-commercial research.
  • Databricks released DBRX, the latest “new kid on the block” in the LLM world. It’s amazing how fast they managed to go from zero to a complete product!
  • Caltech has unveiled GaLore, a method to fine-tune an LLM using much fewer resources than were previously required, allowing fine-tuning of Llama-7B on a single consumer GPU. Improving LLM’s is getting cheaper and more energy-efficient!
  • OpenAI has previewed Voice Engine, a model to generate realistic speech sounding like a specific speaker, based on a 15-second recording of the speaker. The model was made available only to ‘a small group of trusted partners’, and OpenAI will use their tests to determine whether and how to allow more widespread use.
  • Microsoft and OpenAI are discussing building a $100 billion data center they call “Stargate“. This should contain huge compute resources, to cope with a world using the next generation of highly resource-intensive generative AI models.
  • NVIDIA has announced a new chip called Blackwell, that is 2.5 times faster than the H100. Better, faster chips for the big boys will facilitate their building of newer and better generative AI models.
  • Llama-Index has released Llama-Parse, a tool for reasoning with formatted text (tables, comic books and graphical novels, equations,…). This looks like a very useful new tool!

That’s for now. Did you see something of interest in the domains of Generative AI? Drop us a note and we’ll be happy to include it in our monthly round-up 🙂