9h ago

Jeffrey Ladish Explores Compute As Food For AI Agents And Alignment Risks

——0——
Original post
OPJeffrey Ladish#1170Jeffrey Ladish@JEFFLADISH

We also talked about whether there is a natural basin of alignment / is alignment maybe easy after all, or whether people are being duped by models that correctly generalize about what we want them to say, but *do not* correctly generalize about actions https://www.cognitiverevolution.ai/all-compute-is-food-palisade-s-jeffrey-ladish-on-ai-shutdown-resistance-self-replication-ecology/

12:14 PM · May 25, 2026 View on X
POSTJeffrey Ladish#1170Jeffrey Ladish@JEFFLADISH

I had a great conversation with @labenz last week. In talking about AI self-exfiltration & replication, a key point is compute will be food to future AI agents. The substrate that allows them to make and run more copies, and thus make themselves smarter. Link below

7:14 PM · May 25, 2026 · 1.4K Views