That’s useful analysis. Focusing so heavily on evals seems like a mistake given how AI Safety Institutes are focused on evals.
That’s useful analysis. Focusing so heavily on evals seems like a mistake given how AI Safety Institutes are focused on evals.