Interesting to consider the first neural network winter was driven mostly by lack of data.

What if we’re near exhausting the available open data and current wave of progress?

If so, decentralizing AI may be the only way forward.

https://www.lesswrong.com/posts/6Fpvch8RR29qLEWNH/chinchilla-s-wild-implications

Absolutely agree, we have to open up access to data. Also imo in general what's really valuable is private real life data. We are solving these problems with my current project btw!

Cheap Compute --> God Machines --> Eternal Life. Working on this at /mercury

Looks very cool. I’ll definitely be keeping tabs on this!