[godelski]

Yeah we find this type of optimization all over nature. Even radiation is important to create noise. We need it in machine learning. A noisy optimizer is critical for generalized learning. Too much noise and you learn nothing but no noise and you only memorize. So there's a balance

[ErroneousBosh]

I'm reminded a bit of a robotics professor I had at uni, a very long time ago in the 1990s. When he was at uni in the late 70s he and a couple of friends were working on a robot arm, which didn't work because it kept dropping stuff. They made a more precise linkage for the gripper, and if anything it was worse. They made ever more precise bushings for the pivots, and it just didn't help or made it worse.

Eventually after a conversation in the pub with one of his friends who was studying sports physiotherapy, they ripped out the 47th set of really precise little Teflon bushings and put in new ones made of medical silicone rubber tubing.

Now all the joints were a bit sticky and squashy and wobbly, and it picked everything up perfectly every time.