Will computers eventually outsmart people?

[matt-browne-sfw]

Computers become faster every year. Computers can store more and more data every year. Eventually all mp3 files that exist somewhere on our planet will fit on a memory chip that we carry inside our wristwatches. Okay, that's all hardware related. For computers to become smart you need software. Well, you need very good software. How can it be created? Will reverse engineering of the human brain help? Or will reasonable good software at some point make further improvements without our help? Evolve on its own? Should we allow evolutionary algorithms to develop into true artificial intelligence... Survival of the fittest?

I know many people shudder at the prospect, while others say it's not feasible anyway, so no need to worry. What do you think?

Here's an interesting quote from a very intelligent and well-known physicist:

"Some people say that computers can never show true intelligence, whatever that may be. But it seems to me that if very complicated chemical molecules can operate in humans to make them intelligent, then equally complicated electronic circuits can also make computers act in an intelligent way. And if they are intelligent, they can presumably design computers that have even greater complexity and intelligence." -- Stephen Hawking

 

[Fake Vencar]

Computers will only outsmart people if people let them, if indeed these 'complicated electronic circuits' will ever be formed. Computers are stupid; they can only do what ever you tell them to do, including what viruses do.

 

[dustinzgirl]

It is very possible to create a software system that learns and remembers the way a human brain does. Its actually a pretty simple idea, mapping the functions of the brain, minus the hormonal changes, and apply that to a software program. The only problem is that we don't yet have the technology capable of completing the vision. We will someday. Technically, computers are smarter than people now anyways, they can remember and complete functions we can not do without the computer...but, the computer needs human input, so the real trick is creating a computer software program that does not need human input.

Skynet, anyone?

 

[Fake Vencar]

Sounds like Data off Star Trek. Incredibly brain, no hormones. Until someones gives them some.

Skynet!!! No!!!!

 

[PTeppic]

I was "studying" neural networks and pattern recognition (admittedly by missing 80% of the classes!) almost 15 years ago and I suppose they've made some developments since then! From my random ponderings on the subject (running 3-4 miles 3-4 times a week gives a sci-fi addict plenty of time to think about things) the key is perhaps not how to make computers "learn" but to take the philosophical or biophysics step backwards to determine what is the intelligence we're trying to replicate. Assuming we want to.

As mentioned, computers are stupendously good at calculating, in both senses: they are fantastically fast at the most complex calculations, if they know what it is they're supposed to be doing. Though can ironically make fundamental mistakes due to errors in hardware or software and aren't necessarily any good at spotting it. But so are inexperienced humans. But is that intelligence? They're awesome at memory - but what about recollection? I guess indexing technologies are on the rise. But still not intelligence. Even "pure" learning isn't the key: as I understand it you pretty much have to programme neural nets very specifically about the problem domain before they can learn on their own... whereas the human brain can simply learn the whole thing: that there is a problem, determine the domain, the rules and how to succeed in them. It's that associativity that we're probably still some way from. But it will only be time.

 

[Pyan]

In one very specific way, a computer has already proved it was smarter than a human - when IBM's chess computer Deep Blue beat the then world champion Gary Kasparov 3.5-2.5 in 1997.

IBM Deep Blue - Wikipedia, the free encyclopedia

 

[The Ace]

Give them a scratchcard.

 

[Pyan]

LOL, Ace - that should do it!

 

[dustinzgirl]

Give them a scratchcard.

Agent Smith found that hilarious.

 

[tangaloomababe]

Gosh mine often outsmarts me already!!!!!!!!!

 

[iansales]

In one very specific way, a computer has already proved it was smarter than a human - when IBM's chess computer Deep Blue beat the then world champion Gary Kasparov 3.5-2.5 in 1997.

But that's not really smarter, though, is it? All the computer does is work through every possible combination of moves resulting from its opponent's move, and chooses the best one according to a rule-set written by... a human.

Computers can process data faster than humans, but where they have to laboriously work through every data item, the human brain can shortcut the process - either through recognising a pattern, using experience, or even from "intuition".

I don't think computers will ever be "smarter" than humans. If true AI is ever achieved - and we need to know a lot more than we do about the brain before computers can become more than just sophisticated adding machines that can mimic some of the characteristics of a human - if true AI is ever achieved, then it will result in something that bears so little resemblance to human intelligence that no comparison will be really possible.

 

[TK-421]

Anyone see 2001: A Space Odyssey????

 

[iansales]

Er, that's not real, you know.

 

[tangaloomababe]

Give it time!!!!!!!

 

[iansales]

You might not have noticed but 2001 was 6 years ago

 

[matt-browne-sfw]

In one very specific way, a computer has already proved it was smarter than a human - when IBM's chess computer Deep Blue beat the then world champion Gary Kasparov 3.5-2.5 in 1997.

IBM Deep Blue - Wikipedia, the free encyclopedia

With Blue Brain IBM has become even more ambitious

Blue Brain - Wikipedia, the free encyclopedia

A computer simulation of the entire human brain:

The initial goal of the project is the simulation of a neocortical column, which can be considered the smallest functional unit of the neocortex (the part of the brain thought to be responsible for higher functions such as conscious thought). Such a column is about 2 mm tall, has a diameter of 0.5 mm and contains about 60,000 neurons in humans; the simulation will focus on rat neocortical columns, which are very similar in structure but contain only 10,000 neurons (and 10^8 synapses).

 

[ice.monkey]

You would really have to define what 'outsmart' is. Because, as has been pointed out, the speed that a computer can calculate is beyond that of a normal person and so with any math a computer could already be said to 'outsmart' people.

But really, a computer would have to be able to, of its own causality, create something beyond that which has been created by people in order for it to truly 'outsmart' people. In which case it would have to be aware and also aware of being aware. It would have to be capable of completely independent thought.

So you simulate the human brain. You'd have to map every neural network to do that and that hasn't been done. If you read all the present data on the brain you'll see how much is just theory and guesses. It's believed that this part of the brain is responsible for this and that part responsible for that, but if you ask 'how is it responsible?' You won't get a definitive answer.

So for ibm to simulate the human brain when no-one actually knows how it works is just p***ing in the wind and a complete waste of money.

As far as I'm aware only one "animal's" neural network has actually been mapped - that's the C. elegans worm that has a total of 302 nuerons. And this mapping shows which nueron is resposible for what bodily function. And nowhere is a nueron mapped as holding memory, or being aware, or capable of problem solving - or the synapses either. And so it was concluded that C. elegans has no memory or intelligence. Until some bright spark stuck the worms in the classic maze experiment and proved that it does have memory and resolves the problem of where to go for regular food.

So, as yet, science can't even explain how a worm with 302 nuerons works mentally. How the hell can it even come close to mapping the human brain?

Which goes to show how far away from the 'holy grail' of true AI and an aware computer we are if the line of simulating the brain is pursued. Backwards engineering it is too far off to be contemplated now. Other avenues will have to be pursued. Personally I think we're still more likely to destroy the planet before we create a computer that will 'outsmart' people.

 

[Dave]

I listened to a programme on BBC Radio 4 yesterday that was discussing Intelligent Agents. These might be a robot or an embedded real time software system that observes and acts upon an environment. They are capable of perception and action and can learn from experience, so they could be considered intelligent if they interact with their environment in a manner that would normally be regarded as intelligent if that interaction were carried out by a human being. Software based Intelligent Agents have been used to buy and sell on stock markets, and to work out the best time to bid on eBay to win the auction at the lowest price (this isn't actually at the very, very last moment apparently, because as that is what everyone tries to do, that strategy fails to work.)

But they aren't really intelligent, even though they can learn to do better with experience, they are still only as clever as the person who wrote them.

 

[manephelien]

Indeed. Computers are able to simulate various parts of human (and I'd argue mammalian) intelligence, but it'll take a while before computers are able to make a more or less coherent whole of it. Computers are good at the kinds of tasks that require left-brain intelligence, i.e. numbers and linear reasoning. They're just about useless at taking a lot of disparate and apparently unconnected facts and make a working hypothesis of it, i.e. right-brain functions which are more parallel in nature. If we ever get quantum processing to work on a large scale, that may change. Computers are just about useless at understanding language, for example. That's why there'll be work for translators like me for a long time to come.

 

[Ursa major]

...so they could be considered intelligent if they interact with their environment in a manner that would normally be regarded as intelligent if that interaction were carried out by a human being. Software based Intelligent Agents have been used to buy and sell on stock markets....

So they can exhibit herd instincts, can they?

Given that a company's perceived value, based on its share price, can vary widely (by one or more percentage points) each day, yet on the Friday end close to its Monday starting price, I wouldn't want this to be used as a definitive measure of human intelligence. Sheep, maybe....