Posts
Comments
As stated the statement "Randomness didn't buy me any accuracy, it was a way of trading accuracy for development time." is a complete non-sequitur since nobody actually believes that randomness is a means of increased accuracy in the long term it is used as a performance booster. It merely means you can get the answer faster with less knowledge which we do all the time since there are a plethora of systems that we can't understand.
Well Eliezer seems to be stuck in the mud here the only solutions to a problem that he can accept are ones that fit into bayesian statistics and a logical syllogism. But he seems to be blithely unaware of the vast scope of possible approaches to solving any given problem. Not to mention the post seems to be a straw man since I would find it hard to imagine that any real scientist or engineer would ever claim that randomness is categorically better then an engineered solution. But I guess thats because I associate and work with real engineers and scientists so what would I know. Not to mention the statement that a randomized algorithm can perform better is true in a limited extent. The statement about the randomized algorithm being better is a specific statement about that algorithm and I would imagine the people making it would agree that one could engineer a better solution. I can't imagine any real engineer making the claim that the randomized algorithm will always perform better then the engineered one.