Is current AI just automated statistics or there is something more to it? Posted on October 7, 2021June 24, 2022 by Anonymous is current AI just automated statistics or there is something more to it?
think of neural nets as a useful way to encode (and learn the encoding of) arbitrary functions between things.
Yeah, except the useful part.
neural nets aren't magic fairy dust that solve all your problems, but given the right tasks they are excellent tools
What would you consider a more useful way to encode arbitrary unknown, incomplete functions?
by collapsing a wavefunction directly to states
And if you don't have a quantum computer?
Then stop being poor.
Isn't that the world's fault for producing so few quantum computers?
How do I make the world be less poor?
hidden layers are ruled by complexity.
Not much more. It's kinds like just optimizing a function that can correlate a bunch of variables in order to classify whatever the thing with the variables is.
composing multiple layers of linear and non-linear functions is a notable
What makes you think human (or animal) brains are any different?
what makes you think they are? I don't know how human brain functions.
They are not Turing Complete Machines.
Human brains are turing complete.
Prove it, gay. Run me a Turing tape in your head.
Turings original paper was coming up with a way to encapsulate how humans perform math with a machine.
Turing machines are a philosophical thought experiment to mechanize human thought.
I can imagine tapes moving back and forth, for precisely I can perform with pen and paper all processes a turing machine can even if in slow at it compared to a modern computer.
You can too.
>I can perform with pen and paper
No paper allowed. Run me a Turing tape.
Why does the machine get a tape but I dont?
The machine is the head that reads in the symbols on the tape and slides back and forth. My mind is the head that reads in input from the tape (paper) and slides back and forth.
I don't see the difference.
>Why does the machine get a tape but I dont?
The machine has reliable RAM. The RAM is a part of that system. Are pen+paper parts of the system of your brain? Yes? Why, because you perceive it to be so? Fine: I build a robot that presses a button to turn on the PC. The robot is Turing-complete.
>with pen and paper
In a Turing Complete Machine, the program and the data are all on the same machine, the fact that you need additional tools outside of your brain (such as pen and paper) proves it is not Turing Complete by definition.
So a human + paper is turing complete but a human alone is not? What would we be then?
>human + paper
No because it is two different things, the memory space in a turing complete machine needs to be capable of holding both the program and the data simultaneously, just like regular twins with separate births can never be conjoined twins even though they have a little bit in common with conjoined twins when considered together.
you don't need infinite tape gay, you need unbounded tape (add memory if necessary), since if your algorithm needs infinite tape it will never stop. Also it is instruction set that can be turing complete, not only the system.
>you don't need infinite tape gay
Okay. Now run me a Turing tape in your head. (No paper.)
>it is instruction set that can be turing complete, not only the system.
The system in this case is the brain, and I'm still waiting for you to run a Turing tape on it. Maybe your brain just doesn't run Turing tapes. I can give you a program in C++ instead -- they're Turing-equivalent.
If you want, I can just imagine a tape in my minds eye and slide back and forth in that way.
>If you want, I can just imagine a tape in my minds eye and slide back and forth in that way.
Very well. Here's a Turing tape for fluid dynamics. Run it on your brain and tell me the results.
I don't need to run any tape in my brain, all I need to is to
turing machine in my mind, which is easy.
Go ahead. Simulate me a machine that calculates fluid dynamics.
If you taped together all your neurons and laid them end to end, they would encircle the universe 42 times.
What's an objectively observable difference between sufficiently complex automated statistics, and "something more"? It's "automated statistics" until it becomes complex enough that the "automated statistics" view is too low-level for any insight.