Friday Links: Indian AI, Millions of new Materials Apple's AI Cloud in your Pocket

AI is spreading out with many organizations pushing on with R&D, that's a long term positive

Friday Links: Indian AI, Millions of new Materials Apple's AI Cloud in your Pocket

A hugely busy week with lots of funding announcements (picking just one here). Here are the links for the week:

  • Indian AI startup Sarvam scores $41M funding. National champions for AI foundational research are a thing (Mistral and Aleph Aleph are similar). While a lot of money needs to go into each effort, I think it’s a positive sign that investment is happening for diverse cultures, languages, and approaches to foundation models. There’s still a lot to learn, and assuming an Anglo-centric world is the wrong thing to do. I would add that in each country, there needs to be a solid public R&D effort supported alongside the startups. Some of the challenges are going to be much longer term than a startup can focus on.
  • Databricks launches new tools for building high-quality RAG apps. Databricks is going all in on its position as the data store for large organizations. It’s natural that they’ll try to provide the platform and toolsets to build data-augmented AI apps. It puts them in a strong position even versus the big three cloud platforms that they run on.
  • Apple releases an M-Chip optimized model framework. Apple is the stealth cloud player. They have been quiet on the AI front, but a framework like this shows at least part of the gameplan: AI on your local device, not on the public cloud. It’s long been Apple’s view that they don’t need to run an AWS scale cloud — they are already shipping plenty of compute to user’s desktops and pockets. You can bet that there’s be an AI privacy angle to the marketing when it comes.
  • AI Discovery of new materials. One of the Wow bits of news of the week was Google Deepmind’s paper in nature describing the use of AI to discover millions of potential new crystal structures. Using AI to explore the space of possibilities like this is a huge win in many areas, from materials science to design and health.
  • Where are we with AGI, defining levels? Michael Spence and Abhinav Upadhyay have a wide-ranging piece on opinions on the level of AI v’s AGI which includes a look at a recent Google Deepmind paper that tries to focus debate on levels of AI. Some opinions on that in a future post!

Some exciting personal news too. Kin Lane and I banded together to launch a new version of the APIS.IO search service this week. See the launch post and we’d love your feedback. Thank you to the APIDays conference for hosting us!

Wishing you a wonderful weekend.

Midjourney: A steampunk iPhone with a large computer chip shown on the screen

Thank you for reading SteampunkAI. This post is public so feel free to share it.