4 ms·
All evals we have are just far too easy! <1% difference is just noise/bad data We need to figure out how to measure intelligence that is greater than human.
by rkwasny 2y ago
All evals we have are just far too easy! <1% difference is just noise/bad data
We need to figure out how to measure intelligence that is greater than human.
- omneity 2y agoGive it problems most/all humans can't solve on their own, but that are easy to verify. Math problems being one of them, if only LLMs were good at pure math. Another possibility is graph problems. Haven't tested this much though.