/sci/ - Science & Math

> if you gave a computer the ability to learn would it be able to learn

We already have computers that learn. But they have shit tons of programming going into them, to every little thing. I am literally narrowing this down to the ability to move, and connect senses to each other, and a positive and negative response. And asking if this could develop into something as intelligent as us.

Probably, but it would take a computer far more powerful than what's available today. Not something you could jam in the head of a human-sized robot.

You're basically asking a computer to mimic the human brain. With our current technology, no. In 100 years? Probably not. 500? Large maybe, depending on whether or not we've destroyed ourselves or are beginning to revert to a primitive society. If we're still pumping out math and science in 1000 years, I'm sure we'll have it by then.

>>3066726
Actually, now that I think of it, if we can manage to utilize quantum entanglement, that'll be the big thing.

A normal human adult brain probably has like 900-1000 terabytes of memory

it would be very hard to get a computer to learn in the way humans do, between birth and adult hood, just the way in which we process thought and covert it to memory and action changes many times. The way in which these changes happen within the brain is dependent on the individual person and the environment they grow up in. So in other words you would need to create a computer that can no only learn from its surroundings but re-program its self to work more efficiently within that environment and with its new knowledge, and keep continuing to build from there.

I don't know much about the technology to achieve this, but even if it existed, I am not sure that there would be anybody capable of writing a program complex enough to handle it, and even then I doubt you would be getting anything in the way of emotions or feeling like Chobits seem to have.

>>3066734
Technically the human brain has infinite storage, since when it runs out of synapses it just recycles them. Its capacity at any given time is finite, but theoretically, over an infinite period of time a brain could store, or have stored in the past, all possible information.

Also I doubt you could measure the capacity of the human brain in bytes, since we probably don't store data in binary. I don't think its known exactly how we actually store memories.

>>3066772
It is possible to write a program that can modify its own code, but it would be very difficult: http://en.wikipedia.org/wiki/Self-modifying_code

If you combine the difficulty of doing that intelligently with the difficulty of programming a learning AI in the first place... you have an immensity of a task.

>>3067189

>Technically the human brain has infinite storage, since when it runs out of synapses it just recycles them. Its capacity at any given time is finite, but theoretically, over an infinite period of time a brain could store, or have stored in the past, all possible information.

>Infinite information storage because you can delete information and replace it with new.

I'm sorry, what?

>>3066726

Blue Brain project says otherwise. Project head said that they'd have a working human brain with molecular-level precision by about 2020. Whether you think he's right or not, it's extremely pessimistic to say we would take a hundred years to emulate a human brain, yet alone 500.

Once we can model a human brain, we can reverse-engineer it.

2050 is the latest I'd put on an AGI being possible.