Science Thoughts on Artificial Intelligence

[Linguofreak]

A machine does not feel. It has no sense organs and does not react to the environment the way life does. Machines just do what its software and plain sensors are supposted to do. Software and machines won't ever have consciousness because metall, plastics, rubber, silicon, ventilators, laser and electricity can't feel, think and, combined or not, won't become alive.

This all depends on your metaphysical assumptions.

If you believe (as I do) that there is a component to human conciousness which exists outside the physical universe (i.e. a soul or spirit), then you are most probably right, unless there is some metaphysical rule that says that every physical computational structure that meets certain requirements will develop a soul, or unless God (or some lesser being) sees fit to "assign" souls to machines. I think these conditions are most likely not met, and so I generally agree with you that machines can't feel, because I think that feeling involves more than just a physical processing of sensory information.

But if you believe that the human conciousness exists entirely in this universe, then feeling *does* break down to merely the physical processing of sensory information, and I don't see how you can say that a sufficiently complex machine could not emulate the human mind.

---------- Post added at 11:24 PM ---------- Previous post was at 11:21 PM ----------

Do carbon chains feel?

According to you, yes.

By "carbon chains," he is refering to all the wonderful molecules that make up the human brain and body, and form the computational infrastructure behind the human mind.

---------- Post added at 11:27 PM ---------- Previous post was at 11:24 PM ----------

If metal, plastics, rubber, and silicon cannot love or feel, how then do carbon chains do it? Without invoking magic.

I'm not sure about Moonwalker, but my answer is that carbon chains don't do it at all without invoking "magic," but I have no problems at all with invoking "magic".

 

[Hielor]

I'm not sure about Moonwalker, but my answer is that carbon chains don't do it at all without invoking "magic," but I have no problems at all with invoking "magic".

There's nothing wrong with invoking magic. I don't accept that as a valid scientific argument, but it needs to be declared--either you're arguing from a scientific perspective, in which case you cannot invoke metaphysics, or you believe in metaphysics, in which case you should not be arguing that scientifically.

I don't have a problem with people believing in metaphysics. I have a problem with people passing off metaphysics as science, or failing to make the distinction.

 

[Kaito]

metal, plastics, rubber, silicon, ventilators, laser and electricity can't feel, think and, combined or not, won't become alive.

How sad humans have to be be to think machines could be something like humans. Why not start with a plain dog or fish? Even this doesn't work.

The human brain is made up of protein and fat and liquid. To an outside observer, protein and fat "combined or not, wont become alive", and yet here we are. Just thought I'd point that out

Dog of fish? While my dog may be a clumsy, she's smart. We tell her to come here, and she comes. She can learn. Dogs mainly dont communicate verbally, but with body language. Just learn the body language and you can communicate JUST AS EASLIY with dogs as with humans. And you say dogs aren't like humans. Bah humbug.

Also: I do feel that REAL AI will eventually surface. I dont know how or when, but I feel that it will. Computers are doubling in power every 18 months, and eventually we will get to that point.
People didn't think we could fly: we did
People didn't think we could break the sound barrier: we did
People didn't think we could go to the moon: we did

 

[Spicer]

The real question is that if a computer can replicate thinking, feeling,etc to the point that a human cannot tell the difference...what is the difference to a human?

 

[jedidia]

I see that we're getting into the typical "strong AI vs. weak AI" debate.
Strong AI being the hypothesis that emulating a behaviour on the basis of zeroes and ones is essentially the same as bringing that behaviour about based on biochemical reactions.

That debate is as old as the idea of AI itself, and I think we won't find an answer to that here.

However, I want to point out that the human body is a lot more than "too weak a power supply". The biochemistry of our body influences our behaviour a big deal (I'd say at least 1/3rd). This can get higher or lower, depending on a humans "character", which AGAIN is partly based on his biochemistry, and the rest on how his brain developed.

As for comparing the human brain to software, I think that, on the basis of what we know about it's working, we can say pretty certainly that a Human brain is hardware... Allthough it FULLFILLS the purposes we presently use software to achive.
You remember the first computers before microelectronics got far enough? they had to switch conections with relays, which actually did a part of what today software is doing. The brain is similar, though indefinitely more developed: It is a piece of hardware that can make connections and disconnections freely and very fast, but it is a PHYSICAL process, not a digital one. This is why a human HAS to learn things, and why he can forgett: He has to adapt his whole hardware to the circumstances (and he needs between 18 to 25 years to do it properly. Women are faster than men, because their biochemistry favours the cause. Which again shows how big the influnece of the body on the brain is.)
And again, the workings of the human brain is to a major part controled by our bodies chemistry. You have a problem with a certain hormone (e.g. Serotonin), you'r brain won't work correctly allthough there isn't per se something wrong with it (like your computer won't work with the wrong power supply, allthough he would be in perfect shape).
You feed your body strong chemical substances (drugs) your brain goes temporarily havock allthough there isn't per se anything wrong with it (allthough it might get damaged in the process).

Now suppose we HAVE the computing power to emulate all this, and a behaviour can be achieved that isn't discernable from human behaviour anymore. According to Strong AI, there's no difference. According to weak AI, as long as the basis of the behaviour is only zeroes and ones, it stays an emulation and has no feelings involved, and hence cannot be called "aware". Like I said I won't debate that point too far.

To achieve a "machine" that is conscious in the eyes of weak AI, the whole data has to be hardware-based. This could work by assembling a biocemical (read organic) machine, which would in it's essence not be a Robot, but a Replicant, and it's "hello there, Philip K. Dick!".

Or it might be achievable through nano machines, which sounds even more scifi.

 

[Jarvitä]

As I said: metall, plastics, rubber, silicon, ventilators, laser, electricity and so on... (like your cogs and levers as well) can't love, can't feel, can't think and can't develope consciousness. Computers are no organisms. They are no creatures. They are indeed pretty death and absolutely non-amazing compared to what the nature offers. No matter how perfectly you simulate a human brain, it will remain death and do nothing more than just convey the impression of being similar to a real human brain.

Wrong. What makes interconnected neurons superior to transistors/vacuum tubes/cogs? It's still just a finitely complex computing machine, working with finitely complex algorithms, there's nothing special about "life", it's basic organic chemistry.

Or, to indulge your fantasy, let's say we had a computer powerful enough to simulate the brain on the atomic level - of course, with the accompanying scanning technology that allowed us to perfectly replicate every brain function. Is the computer simulation (now no doubt fully "conscious" and "self-aware") still just "dead silicon"? Or, let's say we had the technology to replace an individual neuron with an identically functioning chip of identical size and shape - does a human become a "dead machine" once you replace a single neuron? Where do you put the boundary?

Your argument is flawed, neurons aren't anything special compared to integrated circuits. They aren't even terribly optimised.

To achieve a "machine" that is conscious in the eyes of weak AI, the whole data has to be hardware-based. This could work by assembling a biocemical (read organic) machine, which would in it's essence not be a Robot, but a Replicant, and it's "hello there, Philip K. Dick!".

Any software is hardware based, whether it's as a set of (relay switch/neuron) connections or a set of ones and zeros on a storage medium. Flawed thinking again.

 

[Urwumpe]

Even worse - what makes Neurons strong is the fact that they are badly optimized. And they are painfully slow. In the time for one fast reacting neutron to react on a chemical impulse from another Neuron, Orbiter rendered a frame.

What makes our brain special is just one thing: We can't yet build computers which work like it. No central timing, today computer designers go great distances to make sure the timing works. When software designers work with synchronization, our brain does not care about it. It just splits the task into smaller and smaller subtasks and does not wait until these subtasks are done before already starting the next task.

That is also the reason why our brain only uses a small number of neurons at a time - only a few neurons are needed at a time, the others just idle waiting for the next task. Or are specialized for some tasks, which are not always needed. The brain can afford having neurons do nothing for a long time.

But imagine an economically designed CPU, which had many small cores for just letting 20 programs run in parallel if needed, even if you normally just have 2. Or which let the two programs run in parallel on all cores in a sequence, to react on new data, while the old data is still processed. Our dogmas can't deal with it yet - and we don't want to. If CPUs would work like human brains, we would have to imagine being bothered with porn textures on the walls, while trying to play Quake and don't get hit when MS Word distracts us.

Our brain is used to be constantly distracted. A brain which can perfectly focus all the time, is not healthy.

 

[jedidia]

Any software is hardware based, whether it's as a set of (relay switch/neuron) connections or a set of ones and zeros on a storage medium. Flawed thinking again.

I merely pointed out the definitions of the two sides of a debate that has been going on for decades, between people more qualified than us (well, than me anyways). If you will so, those are not my arguments , and yours are not yours, but have been brought up from each side over and over again.

supporters of strong or weak AI usually don't even come to terms wheather the debate is philosophical or technical, much less they reach a consensus on the actual topic.

 

[Moonwalker]

The human brain is made up of protein and fat and liquid. To an outside observer, protein and fat "combined or not, wont become alive", and yet here we are. Just thought I'd point that out

Protein combined with fat and other organic matter becomes pretty alive. And more important, nobody has an explanation why and how we are what we are. We are just able to look at protein, neurons and fat. But this does not answer anything.

Science and technology allows us to fly, to break the sound barrier and to land on the moon. But it does not yet offer answers required to replicate human beings including something like soul, consciousness, love and hate which are the key-features of humans beings. Some people try to sound smart, they make sound everything easy and simple while they're not able to answer anything at all. Where do ideas and concepts come from? Sure, obviously out of the brain. But how do they get into it? How do they emerge and why? As long was we can't answer such questions, we won't build anything that is barely close to human beings.

And one thing is for sure anyway: machines/computers are no human beings and they won't become so. They are no beings at all. Just death matter. Replicate human beings won't work the computational way and death matter way at all.

 

[Urwumpe]

Protein combined with fat and other organic matter becomes pretty alive. And more important, nobody has an explanation why and how we are what we are. We are just able to look at protein, neurons and fat. But this does not answer anything.

That is wrong. We do already know how proteins and other chemical processes work. Research focuses on predicting the behavior of a protein if you only know the DNA or RNA sequence. Which both is extremely complex compared to just tell what a known protein does - RNA can act as catalyst for many chemical reactions, DNA is not always expressed the same way.

By using the car analogy: We already know that the car is propelled by air and gasoline and that it can be controlled by pedals and steering wheel. We already know how a combustion engine works and how to build one. But we don't know how the engine in this car looks like, we just have many different examples from engines from simpler cars and now try to find out more by experimenting with them.

I would say, in pure insanity: In 10 years from now, we will be able to tell how an animal will look like by just knowing a complete genome.

But we still won't be able to produce the animal by machines.

 

[Eagle]

But we still won't be able to produce the animal by machines.

That's partially because DNA doesn't contain all the required information. There's mitochondrial genomes and some specific proteins that aren't transcribed from DNA.

I think true AI will come when we build something that can learn new patterns and actions from observation and make creative connections between patterns to solve new problems. The computer brain will need some level of self modifying code to help adapt its ability recognize patterns and make decisions, but I don't think the whole procedure needs to be modifiable.

 

[Kaito]

But this does not answer anything.

Science and technology allows us to fly, to break the sound barrier and to land on the moon.

The reason I stated the bit about the brain chemistry is because I just wanted to point out that our brain is nothing "special" in terms of its makeup. So it could possibly be the silicon and metal and plastic could eventually become like our human brain.

The bit about flying and the sound barrier was just meant to show that people didn't believe we could do such outlandish things, yet we somehow accomplished them.

The problem with human intelligence is that we know how to rebel. If your 14 year old rebels, you take away his computer and TV, and he subsides. If a robot with 5x human strength rebels...you're SOL. Now, what I think will be the practical "AI" (without the risk of rebellion) will be a robot that does EXACTLY what its told. We would send a robot to "school" to learn basic things to help humans (I'm thinking iRobot atm) such as open doors, make coffee, answer the phone. They would learn this by trial and error. After x amount of days in school, they could be ready to be sent to an owner where they would be put to use, but also learn new things, specific things about that owner (such as "Open da door" = "open the door" = "open door" = "Open the door please"). The nice thing is that once 1 robot goes to school, the info could be "uploaded" to all other robots. I think THAT will be the extent of commercial use of "AI"

 

[Andy44]

Some humans are very submisive and do what they're told, even though they are intelligent and self-aware. Lack of testosterone or some other hormone, maybe.

What would be the AI equivalent of that? The human brain is thought to be an independent computer, but the effect that hormones and self-image has on it are contrary to that idea. The mind and body are integrated.

An AI must have such a connection to its environment as humans do through a body. And the way to control the behavior of an AI without it reverting to a mindless fatory robot is to find the AI equivalent of hormones.

 

[Ghostrider]

I keep reading about how we could emulate human intelligence on a machine, but apart from it feasibility, would it be desirable? Our current computers can only do computational errors due to hardware or software defects, crash a lot and lose our precious pr0n collection. Do we really want to make a machine able to do stupid judgement errors and, most humanlike of all, defend those mistakes or put a spin on them?

And let's assume we do make such a machine. How do we determine its intelligence is humanlike? Someone mentioned the Turing Test but the Turing Test is fundamentally flawed. Can you tell from my posts that I'm not some kind of machine? Can you tell if I'm one or more persons hiding behind the same nickname?

What about self-awareness? That can't be measured because it's subjective and it's not even a constant. Are you self-aware when you're sleeping? Maybe you are when you're dreaming but how can you tell someone is self-aware or not? And it's already hard when considering humans, a machine would be a whole different pair of socks.

And in the end, what is our definition of intelligence? Who is more intelligent, the boy genius who can slice through college at 15 or the 15-year old slacker who doesn't like math but can save his own and everybody else's bacon in an emergency situation because he can adapt to the environment faster than anyone else? And let's not even consider intuition which is yet another headache.

 

[Andy44]

I keep reading about how we could emulate human intelligence on a machine, but apart from it feasibility, would it be desirable?

Well, someone wants it. I would settle for a machine that's smart enough to clean my house, do my laundry, and drive me around when I'm drunk. It doesn't need to be able to talk like a human, just do what I want.

I mean, come on, it's 2009 for heaven's sake? Where's my damn robot already? Not to mention my nuclear-powered flying car.

And let's assume we do make such a machine. How do we determine its intelligence is humanlike? Someone mentioned the Turing Test but the Turing Test is fundamentally flawed. Can you tell from my posts that I'm not some kind of machine? Can you tell if I'm one or more persons hiding behind the same nickname?

I'm artificially intelligent. That is, I sometimes act like I'm intelligent, but I'm really just faking it.

 

[Ghostrider]

By the way, I'm a friend of Sarah Connor. Could I see her please?

 

[RisingFury]

But is that already intelligence - can the AI do, what the programmer didn't think about?

See, that's the thing...

In UT3, it was not expected that the bots would start exploiting map bugs. They started behaving very similar to they way players with "no honor" behave in game.

We could argue that any program that has the ability to learn and base it's decisions on the knowledge has a measure of intelligence, weather or not it can be predicted. If you increase the the complexity of algorithms and knowledge gathering ability, there will be a point where you reach human standards. And you know where that leads to......

 

[Andy44]

Speaking of AI, here's a grunge-music-era trip through memory lane:

 

[Eagle]

An AI must have such a connection to its environment as humans do through a body. And the way to control the behavior of an AI without it reverting to a mindless fatory robot is to find the AI equivalent of hormones.

Which can lead to an interesting security flaw if somebody manages to hack their pleasure circuits....

 

[RisingFury]

Speaking of AI, here's a grunge-music-era trip through memory lane:

YouTube - Dr. Sbaitso

There's are several bots like this one that you can add to your MSN and talk to. But they don't make for a very interesting conversation because they don't even remember the subject of last sentence.