3 ms·
I'm no AI hypester. But I have made some observations over my "N" decades on this planet so far. That a lot of humans are either: 1. dumb 2. ignorant 3. cra
by syngrog66 3y ago
I'm no AI hypester. But I have made some observations over my "N" decades on this planet so far. That a lot of humans are either:
1. dumb
2. ignorant
3. crazy
4. hateful
5. violent
6. biased
7. indoctrinated
8. flaky
9. dishonest
10. fabulists
11. bored
12. lazy
13. conspiring
And usually an ever-churning mix of many of them, all at once.
Therefore, to make a human-equiv AI its not necessarily "incorrect" to give them those traits too. NOT saying that would be useful, or a desirable ideal to strive for. But it WOULD make an AI act more like an authentic human!
I walk the talk too.
How so? For years now I've been writing a little story serial (DSPR) featuring an AI character who has all those qualities. Sometimes all evidenced in the same single conversation. And lately have been building a game featuring the same kind of insights, and shenanigans. (Slartboz -- see my bio if curious for more details.)
- barking_biscuit 3y agoAn absolutely huge part of it for humans is that as the complexity of the thing being reasoned about grows, the accuracy of the mental models declines. That and we also train billions of individual models on a narrow slice of reality, and leverage a distributed consensus protocol through low-bandwidth "conversations" to update the weights. I think you could probably emulate this with conversational agents each with their own LoRA, that way they wind up with differing opinions.
- syngrog66 3y agogreat insights! agreed