Hal Finney wrote:

No doubt this is true. But there are still two somewhat-related problems. One is, you can go back in time to the first replicator on earth, and think of its evolution over the ages as a learning process. During this time it learned this "intuitive physics", i.e. mathematics and logic. But how did it learn it? Was it a Bayesian-style process? And if so, what were the priors? Can a string of RNA have priors?

## Advertising

`I'd say that biological evolution bears little resemblance to Bayesian`

`learning, because Bayesian learning assumes logical omniscience, whereas`

`evolution cannot be viewed as having much ability to make logical`

`deductions.`

And more abstractly, if you wanted to design a perfect learning machine, one that makes observations and optimally produces theories based on them, do you have to give it prior beliefs and expectations, including math and logic? Or could you somehow expect it to learn those? But to learn them, what would be the minimum you would have to give it? I'm trying to ask the same question in both of these formulations. On the one hand, we know that life did it, it created a very good (if perhaps not optimal) learning machine. On the other hand, it seems like it ought to be impossible to do that, because there is no foundation.

`Suppose we create large numbers of robots with much computational power, but`

`random programs, and set them to compete against each other for limited`

`resources in a computable environment. If the initial number is sufficiently`

`large, we can expect that the ones that survive in the end will approximate`

`Bayesian reasoners with priors where actual reality has a significant`

`probabilty. We can further expect that the priors will mostly be UDist`

`because that is the simplest prior where the actual environment has a`

`significant probabilty. Thus we've created foundation out of none. Actual`

`evolution can be seen as a more efficient version of this.`

`Now suppose one of these suriviving robots has an interest in philosophy. We`

`might expect that it would notice that its learning process resembles that`

`of a Bayesian reasoner with UDist as prior, and therefore invent a`

`Schmidhuberian-style philosophy to provide self justification. I wonder if`

`this is what has happened in our own case as well.`