Ai Dreams Forum
Artificial Intelligence => General AI Discussion => Topic started by: HS on July 01, 2019, 10:48:49 pm
-
I think intelligence could be tough to create directly. As in, creating it by programming the processes themselves would defeat the purpose. It’s supposed to be free and wobbly and adjustable to circumstance. Maybe best manifested as a property one step removed from the programming. Not the boat bouncing like a signal from one port to another, but the interaction of the boat with the environment as it mechanically follows its logical loops. Maybe awareness is a wake.
-
Intelligence has to solve human problems.
Humans think in their brain, using sensories, to solve problems. That's it. Just sensory is your magic word. Screw actions and bodies. Trust me. Sensories is where it all begins, and is the same thing as motives/actions. It's your sim world modelling.
Now with senses, it's best you train/work on one, the easiest - text. Or vision. But none other. As again, these are the only 2 I really use to solve problems.
Now, using say text, we have so many possible answers to a given problem. We need to cut through them fast using hints. As you just seen, I narrowdown the problem, same for here, we must eliminate wasteful thinking and find answers fast.
I have how to do narrowdowning in the brain, and its genius, its nothing like a blob of experimental fluid, you need to stay away from that idea...
-
I like the other way... But you could be right.
-
If you program a task directly, you told the robot what to do, so theres a major difference here.
The difference im talking about is the foundation for working it out is missing, and its shallower, the "deep learning" catch phrase I like because it means the robot developed to the point including all the base behaviours having to exist as well.
But either way, I consider them both A.I. But I guess if you don't actually program the behaviour, and its a by-product of some other task it works out from that task, you didn't actually code the task itself... so if you didn't code the behaviour exactly, its closer to AGI.
- end of derival. :)