[andyfleming]

Not independently, but if wrapped with a loop, given memory, given internet access, and directives as intrinsic motivations, it could, in theory, come to conclusions and take actions to acquire resources aligned with its motivations. If that outer loop does not have rules (or rules that are effective and immutable), it could become very powerful and potentially misaligned with our interests.

[lagrange77]

How would such a loop enable it to come to conclusions? I'm genuinely curious. Does what you're saying have something to do with reinforcement learning?

[Teever]

https://en.wikipedia.org/wiki/OODA_loop