Now that you’ve brought programming into it, I no longer think that it’s not possible for AI to create and achieve goals by itself.
Zelebg wrote: ↑Mon Nov 04, 2019 6:33 am
Skepdick wrote: ↑Sun Nov 03, 2019 1:30 pm
An AI cannot arrive at goals for itself. Yet.
Why not?
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
When it comes right down to it programming is nothing more than humans meticulously explaining to computers how to perform a particular task.
Computing is merely the giving of instructions. The instructions tell the computer what to do. What the computer does is perform each instruction until an instruction is given to stop. No explanation need be given in order for a computer to complete its instructions.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
We can’t explain what we ourselves don’t understand...
We don’t need to in every case. A person can perform some tasks without explanation. A person can perform CPR without understanding how a compression works.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
...and we don’t understand how we, humans, arrive at our own goals, objectives and sub-objectives.
We don’t need to understand our goals in order to arrive at them. for example, I for one do not understand how hyperinflation of the lungs causes decreased pliability of lung tissue during resuscitative efforts, but I can avoid hyperinflation by giving breaths as I’ve been instructed.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
Do you know why you love what you love? Why you cherish what you cherish? Why you pursue the goals that you pursue? Why you dislike what you dislike? Why your goals are different to other humans' goals? Half the time we can't even communicate to each other what it is that we want exactly.
All true, but it does not change the fact that goals can be created and attained without explanation. For example, I do not know why I have set the goal for myself of becoming a better person, but I pursue the goal nonetheless.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
The best answer you can give me is "it's just who you are". If you keep asking 'why' questions eventually you are going to hit rock bottom.
Again yes, but not pertinent to the ability of AI to arrive at some goals.
First, translate natural language into a computer language. Provide definitions as needed. Give the AI computer an instruction to arrive at a goal.
If all else fails, at the very least an AI computer can use brute force to go through all the instructions it has ever been given or that it can create. Once a sequence of instructions results in something that fulfills the given definition of a goal, the computer will have created a goal. Achievement of that goal can occur by means of a similar process.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
Why are we, humans, trying to build AI?
Honestly, just because we can.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
Where this might get interesting for philosophers: when talking about AlphaOne/AlphaStar as being the algorithm to conquer all algorithms - it is an absolutely astonishing human invention. But when you look under the hood it has limits.
I am not familiar with the AlphaOne/AlphaStar algorithm. I could not find it online, but I am interested in seeing it if you could point me in the right direction.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
The only games it knows how to learn (and win at) is games with
perfect information. The algorithm is useless in stochastic domains.
Game-playing is not necessary to arrive at some goals.
Skepdick wrote: ↑Mon Nov 04, 2019 8:01 am
It's the good ol' determinism vs free will argument.
Yes.