Comment by bretpiatt

From what I'm experiencing, we have recursive self improvement that can find local maxima, I'm not seeing really strong evidence of unguided RSI that finds the true optimal.