Update: I read that AI training data has grown 25x since 2020 and now I can't decide if that's good or scary
Last week I was messing around on Our World in Data (the site with all those charts) and found a stat that training compute for big AI models has gone up about 25 times just since 2020. I sat there for a while trying to figure out if that means we are getting closer to something great or just building a bigger and bigger machine that nobody can really check. One side says more data means fewer dumb mistakes and better tools for stuff like finding new drugs, the other side says the models are just memorizing more of the internet and we still don't know why they work. I honestly don't know where I land. Where do you all come out on it?