I have not seen one where an agent behaving catastrophically badly can't be substituted for behaving stupidly and failing
-
"stupid" is many different things, *which is why* it's important to distinguish them instead of pretending they're the same
my argument is that none of the thought experiments I've seen have raised issues of "stupidity" that is uniquely catastrophic
-
-
critique is that invoking catastrophe in the thought experiments I've seen is unnecessary to capture and explore the failure mode
-
surely the stakes are not totally irrelevant
- 4 more replies
New conversation -
-
-
I'd argue that the stupidity of misalignment is intrinsically linked with the impossibility of defining metrics.
-
We can't define what we want either - and we only seem to have compatible goals because our capabilities are limited
End of conversation
New conversation -
-
-
What about an AI that is super competent at finding and wireheading people because you screwed up giving it values
-
like there is nothing you can do to stop it or convince it otherwise
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.