For reasons related to Godel’s incompleteness theorems and mathematically proven minimum difficulties for certain algorithms, I believe there is an upper limit on how intelligent an agent can be.
We already know the upper limit on intelligence. It takes one bit of evidence to narrow down the possibilities by a factor of two.
I believe that human hardware can—in principle—be as intelligent as it is possible to be.
By “human hardware” do you mean an actual human brain, in the shape it normally forms, or just anything made out of neurons? If you mean the former, this is obviously false. We have a limited memory and thus a limited intelligence. If you mean the latter, we already know neurons are Turing complete, though you could still build a more efficient computer that does it faster and with less energy.
Do you mean that a human brain could, in principle, come very close to the upper limit of effective intelligence? That is, you might not be able to memorize 10^50 digits, but you could still answer any question you’d reasonably come across just as well?
Also, are you talking about just training a normal human, or something where their neurons have to just happen to be wired exactly right?
Also, there’s the question of how to measure intelligence. Is it just how likely we are to set off a utilitronium shockwave, and how accurately it follows our goals?
I’m not sure what you mean.
We already know the upper limit on intelligence. It takes one bit of evidence to narrow down the possibilities by a factor of two.
By “human hardware” do you mean an actual human brain, in the shape it normally forms, or just anything made out of neurons? If you mean the former, this is obviously false. We have a limited memory and thus a limited intelligence. If you mean the latter, we already know neurons are Turing complete, though you could still build a more efficient computer that does it faster and with less energy.
Do you mean that a human brain could, in principle, come very close to the upper limit of effective intelligence? That is, you might not be able to memorize 10^50 digits, but you could still answer any question you’d reasonably come across just as well?
Also, are you talking about just training a normal human, or something where their neurons have to just happen to be wired exactly right?
Also, there’s the question of how to measure intelligence. Is it just how likely we are to set off a utilitronium shockwave, and how accurately it follows our goals?