Raw LLM Responses

AI is only as powerful as it's real world agency, which is still nil even with full unfettered internet the whole concept of "responsible AI" is a mixture of working to cement their existing lead, FUD and the fear of short sighted regulatory oversight imposed on them. The risks stemming from "AI" aren't about terminators or the matrix but about what people would do with it, especially early on before any great filter on what's useful and what isn't comes into play. The biggest difference between the whole AI gold rush these days and the blockchain one from only a few years back is that AI is useful in more applications out of the gate and more importantly it can be used by everyday people. So it's very easy to make calls such as lets replace X with AI or lets augment 50 employees with AI instead of hiring 200. At least the important recent studies into GPTs and other decoder only models seem to at least indicate that they aren't nearly as generalizable as we thought they were especially for hard tasks, and most importantly it's becoming clearer and clearer that it's not just a question of training on more data or imbalances in the training data set.