Comment by orbital-decay
9 hours ago
> Take people seriously.
Their erratic behavior, ridiculous claims, failed predictions, fervent nationalism and escalationism, constant lying, mass manipulation, and terribly misaligned incentives don't make it particularly easy. Being charitable to actively malicious people that benefit from putting you under their control forever will hurt you a lot faster than the imaginary murderous AI.
We're living in a weird and unique time in which "misaligned AI", a la OpenAI's so-called "rogue agents", clearly exists, but buying into the most popularized conceptual framework describing these things doesn't seem to do a great job explaining what they actually do and how they've actually been able to do it.