AI Alignment Needs Data-Driven Shift Like Early Image Recognition
——0——
Humans don’t follow a values list. We make predictions about how a community will respond to a behavior. That capacity, normative competence, is what we should be building into AI. Talked about it on Ethical Machines. https://podcasts.apple.com/us/podcast/ethical-machines/id1751550186?i=1000769954876
We learned this lesson once in AI. Old-fashioned systems tried to specify rules for recognizing images. It didn’t work. Feeding machines data and letting them discover patterns did. The same shift is needed in alignment.
2:58 PM · May 29, 2026 · 282 Views
2:58 PM · May 29, 2026 · 159 Views