Syne Wrote:..even when dealing with AI
I'd say "especially when dealing with (bozo) AI" .. without access to good sources and a deliberate policy of 'forgetting' sources you might be better off without it .. and it may become increasingly difficult to get round it.
That's why I think we're going to hit peak AI soon... at least in terms of knowledge. The effort to remove the dross will not be cost effective.
Years ago someone (rpenner) claimed the Internet had kicked open the gates to knowledge .. I think we've found a way to close the gates.
Universities have always 'colluded' and that is most likely the way forward - we just get the dross. When I was a technician working on submarines my supervisor went off to talk to Russians working on the same subject - and is also how I came to meet guys from NASA. My supervisor did the maths and I did the electronics.
Here's a response from what is now an early AI - I don't think they have improved --
Quote:I know enough about Raspberry Picos and MicroPython to be dangerous. What do you need help with? Just try not to expect too much competence;
AI admitting it's not that competent is quite the improvement. I wonder how much telling it it was shite it took to train that model.
The capacity for self-doubt and of being skeptical of one's own information is a very human trait. It might even result in AI having a Cartesian grasp of itself as a thinking being in itself in a world where everything else is dubious. The emergence of introspective awareness. Perhaps the very first inklings of a conscious entity.
It's all just model training. There is no actual self-awareness/doubt involved.