Tbh, I just needed some text before the image. But I have negative sentiment for both Epoch and OpenPhil. From my perspective, creating benchmarks to measure things adjacent to RSI is likely net negative, and teetering on the edge of gain of function. Such measures so often become a target. And it should not come as surprise, I think, that key Epoch people just went off and activity started working on this stuff.
As in, you think frontier math is (strongly) net negative for this reason? Or you think Epoch was going to do work along these lines that you think is net negative?
I mean, not super impressed with their relationship with OpenAI re: frontier math. The org has a bad smell to me, but I won’t claim to know the whole of what they’ve done.
Tbh, I just needed some text before the image. But I have negative sentiment for both Epoch and OpenPhil. From my perspective, creating benchmarks to measure things adjacent to RSI is likely net negative, and teetering on the edge of gain of function. Such measures so often become a target. And it should not come as surprise, I think, that key Epoch people just went off and activity started working on this stuff.
As in, you think frontier math is (strongly) net negative for this reason? Or you think Epoch was going to do work along these lines that you think is net negative?
I mean, not super impressed with their relationship with OpenAI re: frontier math. The org has a bad smell to me, but I won’t claim to know the whole of what they’ve done.