Can computers become self aware?
 

What would it take? How much processing power and speed? Would it be considered "giving life"? When could this take place? Would they turn on us? Would we be stupid enough to try it?

I've never slapped my monitor or slammed down on the keyboard so I would be safe.

It already knows everything about us and our world. It just doesn't care. Yet.

in some ways, a computer is already "self aware"

it has states that are goal-directed (ie., go to other states, or stay in the same one)

the real game in town is artificial neural networks that can learn..we have built ANN with about 5 million neurons so far. can learn not to run a robot into obstacles, etc. a human has about 100 Billion better-organized neurons so we have a long way to go.

one of my teachers had an idea to simulate evolution. take one thousand computers with the programming of a cockroach, and set up a random mutation and very simple problems for them to solve. every generation might take a day. those that show progress are allowed to survive. those that don't start over.

the idea was that over time "intelligence" would arise on it's own.

Each neuron in the human brain makes ~1000 synapses with other neurons. Simulating one synapse in an neural network requires 4 bytes of memory. A 5 million neuron network requires about 400,000 gigabytes. To simulate the entire human brain would require about 8 million gigabytes.

Of course, there's also the issue of whether it's even possible to accurately model biological self-awareness. For a computer to be truly self aware, it would need to cease merely following instructions and start to make decisions and develop ideas of its own accord.

Yes. It's already happened with biological computers. I imagine we'll see honest-to-goodness AI in a completely synthetic computer in our lifetimes.

It may already exist, but you didn't hear this from me. Remember that military stuff runs about 30-40 years ahead of "cutting edge" commercial-grade technology... Further in some areas...

No, we don't have the processing power yet. Even computers that "learn" do so because of very limited instructions.

One idea I was playing with for a while was writing predictive models that self-adjusted--that is, learned from their own results and adjusted their own calculations to produce better results. Even when this was in the bubbling around in my head phase, I quickly realized the processing power it would take to look at and evaluate new variables in real-time. (And I have access to a lot of big iron.)

I remember hearing that science fiction writers in general tend to underestimate hardware advances and overestimate software advances. I consider intelligence and awareness to be software functions.

That sounds like any normal human baby.

Skynet became self aware at 2:14 am EDT August 29, 1997

So a compter might say, "I process, therefore I am."

lol. Thanks, John Conner

By mathematical proof, its impossible for a computer to tell you when it will be done with any given calculation. Progress bars? A rough estimation at best.

Im betting the math will get in the way of the sci-fi dream of iRobot-style AI.

Sci-Fi writers are responsible for most of my unhappiness. Growing up, I was made to believe that by 2002, we would be in space conducting scientific expeditions and engaging in intergalactic trade, all the while blasting enemies into submission with our super-cool weapons and meeting super-hot alien chicks who want human men very badly.

Flash to 2009. I am in a cubicle, under fluorescent lighting, running sensitivity analyses.

>:(

people are becoming less aware...

i actually yelled a young lady this morning - blocking my path while walking slowly down a staircase checking her PDA while i had seconds to catch my train..

I imagine it won't happen untill we write the software that is capable of it. A computer without the software capable of taking it somewhere is just an expensive heater. :)

So I think we're a long way off from this.

No,

A bunch of scientist and engineers at MIT studied this for a long time several years ago.

Long story short computers are only machines that can only do what we tell (program) them to do.

Computers are very good for processing faster and remembering more (memory storage)

They will continue to get faster and have more memory storage but that is all.

Yeah me too.

Except my job is boring.

That sucks.

42?

and where is my flying car?

Evolution :)

It seems that humans have evolved a new skill lately (and a strange evolution, since it's cropped up spontaneously in many people, not just a newly born generation).

They can tell exactly, and unconsciously, how to stand in a hallway/stairwell/foyer such that they can block the entire thing for everyone else, even if it's 10 feet wide.

It's an amazing new skill, that will be of significant use to the human race.

Quit yer b!tchin'. At least you HAVE a job. Plenty of people would love to have it if you don't want it.

Deep Thought.

you're no Jack Handy

Nor am I Douglas Adams

Common sense tells us that the very pursuit of automated killing machines that can dominate the best manned weapons of war will result in automated killing machines designed to dominate the best manned weapons of war. The US military is pursuing such unmanned/automated weapons systems right now.

Of all the things i see as the most likely cause of the destruction of mankind, our own machines are #1 at the top of the list.

http://www.darpa.mil/j-ucas/X-47/gal...res/ucav_n.jpg

Magpies are self aware, how man neurons are in their brain?

In the strict sense of the term, machines do not evolve.

They do not produce copies of themselve; they cannot mutate.

In their current form, machines do only exactly what we tell them to. Sometimes we fail to understand our own instructions.

Machines are good at mimicing human behavior that requires discreet rules: performing calculations, remembering things. They are pretty bad at doing things that require learning or judgement: walking with two legs, driving vehicles in traffic, carrying on a believable conversation.

Deciding who to bomb, with what munitions, and when.

The US military is working diligently at teaching them.

1) I was talking about stupid people...
2) Machines can make copies of themselves, and they can mutate. Whether those mutations will move towards a higher level of complexity, who knows.
3) I was talking about stupid people ;)

A little trick for the oblivious hallway blockers:

Get uncomfortably close and start to read what they are texting/typing. They'll start paying attention to their surroundings.

Really?

I've got several computers and I've never seen little ones running around the house...

Things written in binary seem to be even more sensitive than DNA. A single misplaced byte general causes everything to stop working. I've never seen a group of corrupted bytes ever produce results that don't also end in failure, let alone produce some useful adaptation.

You give them too much credit! Sometimes, they just seem to stop (on that exact magical point) and shutdown for a bit...just long enough to mess up everyone else.

I think they're downloading instructions from the Mother Computer.

Programing not teaching.

Building and programing a robot to find and pick up dog poo in the yard or vacuum your living room is no different than building a flying robot programed to deliver a payload.

The warhead doesn't care if it explodes on the launchpad destroying it's maker or lands on target.

Science fiction is fun... but it is still fiction.

Are we talking about computers now, or machines? Obviously a computer cannot replicate, but it can control a machine, which can. Robots building robots building robots. You will get glitches, and some of them might be environmentally stronger.

The problem is that you'd need a massive amount of energy and resources, and TIME to run through enough iterations to get any changes.

I mean, it's not a complexity issue, virii and RNA replicate, and they are simpler than your calculator (fewer parts, fewer interactions, fewer instructions).

They're teaching them to decide when it is appropriate to attack, with what tactics, and with what weapons. These "UCAV's" (and other military programs being developed) are fully automated autonomous hunter killers. Once unleashed they no longer need any human input to perform their mission from start to finish(or at least, that's the plan)

A hell of a lot of things that were once science fiction are now very much science fact.

all the thing needs is some piece of code to understand the concept of "self", combined with enough room and encouragement to improve itself...

As cpu's are now moving more and more from serial processing to parallel, it could be there anytime. or not..

why do I have a feeling that this question will not be resolved on this bbs?

Forbidden
You don't have permission to access /2001/tma1/wav/disconne.wav on this server.