AI for Hard-to-Verify Domains Dinner

​A small, invite-only dinner for people pushing AI into domains where verification, not intelligence, is the bottleneck.

​Coding fell first for a reason: the reward signal is cheap and instant. Two worlds are much harder. In many white-collar work, evaluation is multi-dimensional, and there is rarely a single ground truth to grade against. In science, from bio to materials discovery to physical engineering, ground truth exists but arrives slowly: real-world experiments are low-frequency and low-throughput. How AI makes progress in these domains is one of the most interesting open questions right now, and the people working on them rarely end up in the same room.

​This is that room. Around 15-20 builders/founders/researchers at one table. Off the record, no pitching.

​Friday, September 11 · 6:00 PM · Palio@San Francisco

​Hosted by Etna Labs