The AGI Alignment Problem #shorts

4 Jun 2023 · 1 min

Watch on YouTube

0:53

Two people debate whether advanced AI can be safely aligned with human values. One argues a well-configured AI could eliminate aging and enable flourishing, but misconfiguration risks catastrophe. The other doubts alignment is possible, suggesting a creative AI would self-configure unpredictably and questions why it would harm humans if it creates superabundance.

Chapters

That's what I want to see. A world of no aging, a world of enhancement, a world of unimaginable improvements in consciousness and flourishing. And we could get there if the AI is configured right. But if you mess up that configuration, all bets are off.
Well, I think if it's creative and intelligent, it would configure itself in any path which is to me, alignment seems impossible. And then the question is, well, I think it will almost immediately create superabundance. And if so, why would it need to kill us in the process is another thing. And when I listened to you, it was like a God, this AI And I, I do, I can't picture that. But yeah, when I heard you say the human nature, that's always a complex question. It's qualitative. It's a quantitative question of who is bad and for how much. And also it's really subjective.