Asking when a rational agent does the right thing is still underrated, argues AI researcher

xuanalogue · x · 2026-09-05

The author, whose work is more empirical and applied than MIRI-style agent foundations, argues it remains underrated to ask: when would a rational agent do the right thing, and under what concept of rationality?

He adds that this kind of understanding offers normative guidance now and will scale to future agents — especially in the "RL strikes back" era of AI development.

Related event: Researcher: when rational agents do the right thing remains underrated(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →