I wonder how much they can innovate on the chip design frontier by using internal models. They’re certainly incentivized to work on this but what’s funny is Nvidia’s investment will probably be paying for a part of this.

In the official openai podcast, they mention using current models to assist in chip design, but didn't go into detail.

Just recycle the headlines from 5 years go:

https://www.nature.com/articles/s41586-021-03544-w.epdf