"1) Human minds are universal computers. That means, it's possible for them to do any computation. That means it's possible for them to learn anything."
A property of universal computers that people often overlook: they can compute anything given infinite time and storage.
We know that we can compute anything (because at least we can build and use universal computers). Does it follow that we can therefore learn anything? Can we learn the entire code for Mac OS X, for example? What's the difference between learning and computation
( ... )
@Speed and learning: OK, so can we at least say that for anything that one person understands, another person can also understand? That our brains don't differ so much that the speed or capacity is hugely different between people. If so, all of my conclusions above are fine (intact).
Also, some things, like the entire code for Mac OS X, might not be interesting or useful to learn. Usually memorising sets of facts is neither -- we can just look them up. So the important thing here is that we can learn anything interesting. (Related: Deutsch's Law.)
This is kinda the point to my original post. Deep theories don't use up much space, because they apply to and explain lots of stuff, instead of having a different theory for each thing.
"If I conjecture a new solution to an as-yet-unsolved problem, is that learning? [...] If so, then why is it good to think about learning in that way?" Yes, it is the same process. It's good to think about learning in that way because it's a fundamental fact about how acquiring any knowledge works, and it'
( ... )
Speed and learning: OK, so can we at least say that for anything that one person understands, another person can also understand? That our brains don't differ so much that the speed or capacity is hugely different between people. If so, all of my conclusions above are fine (intact). I'm not sure. By 'understand' do you mean 'immediately,' or do you mean 'eventually?'
Some people think that there are things that adults can learn that children can't. They're wrong; here's a simple algorithm that any child can follow to learn adult-idea X:
Become an adult.
Learn idea X.
It takes some time, but eventually the child can learn it.
This algorithm is true, but isn't very helpful in understanding how parenting should work.
This is kinda the point to my original post. Deep theories don't use up much space, because they apply to and explain lots of stuff, instead of having a different theory for each thing. Which is fine, if you've got deep theories. Making theories deep while still keeping them accurate is hard, though
( ... )
What reason do you have to think that animals can learn?
There exist people who lack the face-recognition hardware but who have learnt how to recognise faces. So this would suggest that the people who haven't learnt still have the ability to (as hard as it may be), they just haven't yet.
If you give me any examples of empirical tests that contradict my claims, I will change my position and blog about it. (AFAIK all my claims are consistent with reality/observations/tests.)
One corrects errors in one's rendering of reality by conjecturing and refuting. If the rate at which one corrects errors in one's rendering (by conjecturing and refuting) is slower than the changes occurring in the environment that one is rendering, one's model of reality will deteriorate over time
( ... )
Comments 7
A property of universal computers that people often overlook: they can compute anything given infinite time and storage.
We know that we can compute anything (because at least we can build and use universal computers). Does it follow that we can therefore learn anything? Can we learn the entire code for Mac OS X, for example? What's the difference between learning and computation ( ... )
Reply
Also, some things, like the entire code for Mac OS X, might not be interesting or useful to learn. Usually memorising sets of facts is neither -- we can just look them up. So the important thing here is that we can learn anything interesting. (Related: Deutsch's Law.)
This is kinda the point to my original post. Deep theories don't use up much space, because they apply to and explain lots of stuff, instead of having a different theory for each thing.
"If I conjecture a new solution to an as-yet-unsolved problem, is that learning? [...] If so, then why is it good to think about learning in that way?"
Yes, it is the same process. It's good to think about learning in that way because it's a fundamental fact about how acquiring any knowledge works, and it' ( ... )
Reply
I'm not sure. By 'understand' do you mean 'immediately,' or do you mean 'eventually?'
Some people think that there are things that adults can learn that children can't. They're wrong; here's a simple algorithm that any child can follow to learn adult-idea X:
- Become an adult.
- Learn idea X.
It takes some time, but eventually the child can learn it.This algorithm is true, but isn't very helpful in understanding how parenting should work.
This is kinda the point to my original post. Deep theories don't use up much space, because they apply to and explain lots of stuff, instead of having a different theory for each thing.
Which is fine, if you've got deep theories. Making theories deep while still keeping them accurate is hard, though ( ... )
Reply
Reply
There exist people who lack the face-recognition hardware but who have learnt how to recognise faces. So this would suggest that the people who haven't learnt still have the ability to (as hard as it may be), they just haven't yet.
If you give me any examples of empirical tests that contradict my claims, I will change my position and blog about it. (AFAIK all my claims are consistent with reality/observations/tests.)
Reply
Reply
Reply
Leave a comment