AI Top-of-Mind for Dec 15
Top-of-mind, on the open-source front, Mistral out of France has released their new model based on the same approach as GPT-4, and early analysis shows it to be as powerful as many of the closed models. As reportedin Medium, it could lead to a vibrant open-source development community, much like Stable Diffusion in the image space. The company is also receiving a bunch of interest from the investor side as well.
And on that same thought, I mentioned I’d dive deeper into some of the image creation models. Starting with Stable Diffusion, one of the best explanations comes from Jay Alammar entitled ‘The Illustrated Stale Diffusion.’ The image below is from the article, and if you want to dig deeper, he lists a set of great references. Onkar Mishra has also published a good overview.
And no surprise that different models result in different outcomes from the same prompts. Jim Clyde Monge in ‘Generative AI’ looks at this in detail, with one example below.
Lastly, a wake-up call to the financial sector, with ‘CNN’ reporting that regulators are warning for the first time the risks of AI. From the article:
The Financial Stability Oversight Council, a team of leading regulators across the US government, formally classified AI on Thursday as an “emerging vulnerability.”
