I think humans actually are doing the same thing. The world is procedurally generated, and so are our creations :)
We also clearly act in pursuit of reward functions. You could go so far as to say that is the definition of “normal” behavior for humans.
Learning about how ML works is actually humbling to me, specifically because it reveals how much our skills reduce to pattern recognition.
Humans seem much better at learning than machines, ie. we can “figure out” the Victorian aesthetic after a mere handful of examples. What’s not clear is how much we learn this blank-slate as children versus how much encoded / genetic knowledge and memory we have (to what degree is the brain pre-trained?). But I assume we’ll close that gap.
So what differences remain to label something as truly intelligent?
We have a strong Theory of Mind (ability to reason about what others are thinking and feeling). I think that for us to recognize a machine as genuinely sapient we would expect it to have theory of mind about itself and us. Many animals seem to have this characteristic.
It seems unique to humans (so far) is that we have intrinsic motivation as well, and our behaviors are the combination of external and internal motivations. This is where true creativity happens - when one person makes something totally new, not directly derived from prior work. Be that a new musical composition or mathematical theorem. as a species we seek out novelty, experimentation, and “progress.”
Now, in our various AI and ML techniques we can and do simulate this with the injection of randomness, and in “genetic programming” we go to lengths to try and emulate that evolutionary model. So maybe there will come a point where the AI is generalized and then it starts having internal motivations that aren’t transparent to us.
I don’t know where the line is, but theory of mind and intrinsic motivation seem important, so perhaps there’s something there.
We also clearly act in pursuit of reward functions. You could go so far as to say that is the definition of “normal” behavior for humans.
Learning about how ML works is actually humbling to me, specifically because it reveals how much our skills reduce to pattern recognition.
Humans seem much better at learning than machines, ie. we can “figure out” the Victorian aesthetic after a mere handful of examples. What’s not clear is how much we learn this blank-slate as children versus how much encoded / genetic knowledge and memory we have (to what degree is the brain pre-trained?). But I assume we’ll close that gap.
So what differences remain to label something as truly intelligent?
We have a strong Theory of Mind (ability to reason about what others are thinking and feeling). I think that for us to recognize a machine as genuinely sapient we would expect it to have theory of mind about itself and us. Many animals seem to have this characteristic.
It seems unique to humans (so far) is that we have intrinsic motivation as well, and our behaviors are the combination of external and internal motivations. This is where true creativity happens - when one person makes something totally new, not directly derived from prior work. Be that a new musical composition or mathematical theorem. as a species we seek out novelty, experimentation, and “progress.”
Now, in our various AI and ML techniques we can and do simulate this with the injection of randomness, and in “genetic programming” we go to lengths to try and emulate that evolutionary model. So maybe there will come a point where the AI is generalized and then it starts having internal motivations that aren’t transparent to us.
I don’t know where the line is, but theory of mind and intrinsic motivation seem important, so perhaps there’s something there.