A monthly recap of some news, 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:

December 2023

This month’s highlights:

  • NVIDIA has created a custom Generative AI kernel withXQA. A factor of around 2.5 speeds up generation on a GPU. This will become available soon on NVIDIA GPUs, beginning with Optimum.
  • Mistral has confirmed a leak of a model whose performance approaches that of GPT-4. This is the most powerful LLM ever leaked, and it seems it’s here to stay!
  • J.P. Morgan is showing impressive performance on querying documents, by considering documents as spatial data, not just strings of text. This sounds very much up my alley! Unfortunately, the details are a trade secret.
  • Small language models (LLMs that can run on devices such as smartphones) are developing rapidly, and lots of big companies are showing interest and signing deals, with Google seemingly particularly interested in this sector. We’re skeptical in the short term but look forward to being proven wrong. In any event, if it can be in a phone, it can be in a car’s infotainment system too.
  • There is plenty of progress in generating 3D models. They are looking better and more realistic. A theoretical survey is here. Some corporate signings are here.

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 🙂