Build AI Systems for Discernment, Not Approval - Angel Ortmann Lee, Duolingo

Summary

The main theme is understanding human-in-the-loop AI and its implications for trust and cognitive surrender. The talk references examples like GPS navigation and AI search summaries, culminating in the observed phenomenon of cognitive surrender. The practical takeaway is to build AI systems for discernment, not mere approval, acknowledging that the human in the loop may not always be actively thinking.