[space848]

Feels like people take these new models way too seriously.

Yes, GTP-3 is a very significant leap. But how is this different than someone fine tuning any recent, open and free AI model over e.g. all emails and chats from a late friend or even fiancée?

If doing so gives anyone value, how is it wrong?

I'm pretty sure this is the same alarmist thinking as TV will mess up the kids, or video games causing violent tendencies. Initially, we'll all be blown away by ever improving AI models. Then, optimal models for imitating patterns over a given text set will be a commodity, if it's not already. And it'll be everywhere, and we won't think much of it.