It’s interesting to see the collision of academic mindsets with the proliferation of AI. I largely agree with you: if something uses public data it should be offered as a public good.
It makes sense to me why it happened, though. Growing up, I always had the expectation that the stuff I put on the Internet would be there forever and basically used in ways that I wouldn’t expect. I didn’t feel the need to tell people what they could and could not learn from my posts and whatever tiny amount of learning I impart on a person feels perfectly analogous to whatever tiny amount of weight updates I’ll provide to a model.
Academia felt very similar to me. We put out papers because we found the subjects interesting. We didn’t make money from them (public funding) and people would build things on what we learned. It felt like a good model. Everyone shares the knowledge.
My biggest beef with AI isn’t that they trained on the Internet, it’s that they trained on the Internet and kept the result for themselves.
It took me longer to cook the noodles and brussel sprouts last night than it did to prepare the cranberry bread I made this afternoon. Cooking feels like coordinating the landing of a half dozen aircraft on a single runway. “My onions are browning too fast and I don’t have my other reduction ready!”. Baking feels like, “I can calculate the trajectory of this artillery shell, then set it and fire it off when I’m good and ready. I’ll check on the crater in 30 minutes.”