Can a machine be conscious?

a ghost of an idea what you're talking about. Well, considering that most of our fans also listen to stuff you should know, and considering Chuck's obsession with that movie, I'm pretty sure they picked up on it. Good point. So what we're gonna talk about today has nothing to do with ghosts. Nope, Oh thing, well done, Pialette. All right, we're gonna talk about artificial consciousness, which is sort of a similar concept to artificial intelligence. Yes, but it is

entirely different only similar. Yeah, it's kind of yeah, this is this is talking about. So we wanted to talk a little bit about whether or not we ever thought it was going to be possible to create a conscious artificial construct in some form, whether it's a computer or a robot or whatever. Well, this's the sort of top of mind for us because you've been writing a couple of articles for the site along these lines, right, We've been writing, or I've been writing quite a bit about

artificial intelligence and artificial consciousness. I even wrote a blog post for Discovery about the subject as well. And um, and so we thought we'd kind of talked about it because it's a really interesting concept. Um, it's really difficult to talk about, as it turns out, because consciousness is is not a really well defined term. Yeah, from the research that I did for the podcast, in addition to

having read your articles. Um, it's funny because some people think that a computer or a machine can achieve consciousness, but it all depends on how you define consciousness, and it's very difficult to do that for a human being, let alone a machine. Right Exactly, if we can't, if we can't define it for ourselves, how can we hope to define it in some other kind of of device.

I mean, you know, when you think about it, our total experience is with us, you know, that's that's as far as we know that, that's the only experience we have with conscious beings is uh and and really some would argue. Some philosophers would argue the only experience each of us has is with his or her own consciousness, and that we can only extrapolate that what other people experience is similar to what we experience. Now, did I

just blow your mind? Because it blew my mind. You start thinking about that, and you're thinking, okay, well, let's say that Chris and I are both looking at a coin, and the coin is blue. As far as you know, we it is what we You call it blue and I call it blue. But how do I know that what Chris calls blue Chris perceives the same way that I perceive blue? Like wayn't know. Maybe what Chris sees when he looks at a coin, I would call it red.

But we've arrived at a common vocabulary. We're both calling it blue, and as far as we know, the other person is experiencing it the same way that we ourselves are experiencing it. We don't know for sure. UM. I wanted to throw in this quote before we get too far away from this thought because psychologist Susan Blackmore had an article and a new scientist, UM, actually, it was quoted by another uh researcher, igor Alexander Um but her I love the quote that he had chosen from her article.

Um said that the trying to find the inner sensation, that's you know her the quote there. But it's like looking in a fridge to see whether the light is always on right right. So so I mean, just you're going by your own self diagnostic, it's like, well, it must always be on since right right, because that's what I always see. Certain certainly, there are things that we can mention about. They seem to be kind of agreed

upon as being part of consciousness. Things like an inner monologue and inner voice, something that is within us that we can hear, we can rely on, We can you know, think things out without speaking it aloud, without having any outward appearance that that's going on, it's happening inside us. That's part of consciousness. Uh it could you could also argue that anything, any source of sensation, is part of consciousness. Sensations like pain or love, or ambition or fear or

that's hot. I mean, you know, it's it's more than it's not just physiological. There's something within the mind as well that is bring this into a concept, not just a you know, stimulus response. Yes, and and some people will argue that a computer or a machine can figure that out. It can you know, use it thermometer and go, well, that's hot, or you know, a self diagnostic program would say, hey, my wheel has fallen off. Therefore I shouldn't go anywhere

because I will hurt myself. But then the exactly that's and that's a good a good way of dividing it, because is the machine actually feeling anything? As a human, we would say, we'd feel that, you know, we it's not just that we've recognized that. We're Like, if you cut my hand off, Chris, which I am not recommending you do, I'm you know, I have an adversion of blood. I don't want to go as lefty for the rest

of my life. Um, but if you were to cut off my hand, chances are I would have some clues to that other than just looking down and noticing my hand was gone. It wouldn't just be that I had noted, hey, my hand is no longer there. I would be feeling pain, fear, probably, uh, quite surprised. Chris has never raised a finger to me, and I can't believe you would do such a thing jerk so um, But at any rate, I would feel these things. A machine if it were to lose a

part but be able to recognize that. We wouldn't say that it felt it. We would say that it detected it and then maybe even found a way to respond appropriately so they could continue to function. But that's not the same as feeling, which some would argue is part of consciousness. For some reason, I put the image of C three Po in my head after he had been blown apart and he was trying to figure out where the rest of him was. All right, Yes, yes, um,

that wasn't that great documentary a New Hope. It wasn't in a New Hope. Well you know it's in Cloud City. Oh you're right, yeah, because yeah he was. He was just damaged a bit. Yes, actually no, he came came to pardon in New Hope as well. The sad people knocked me you know what. That's good point? Yes, uh, C three Po, that's I don't know why fee paying. No, No, he just was uh, but he did appear to feel something.

C three Po was definitely cowardly, wouldn't you say? Well, he does say one of his most famous lines problem Star Wars and you hope is I have a bad feeling about this, and then there's yes, who is alive? You arub So consciousness is not the only concept that's difficult to define. Intelligence, as it turns out, is also tricky. Now. If you think of intelligence as the ability to solve problems,

computers are pretty good at that. In fact, computers have gotten I've gotten better at even learning how to solve problems. Before you might argue that computers could only solve things that um that someone had a programmed into it. In other words, they computers knew how to do basic arithmetic and that was about it. But if you gave it a mathematical problem, it could work it out because it

knew the rules. Right. But we've made a lot of progress in artificial intelligence to the point where computers can do things now like recognized patterns. That's how you get these cool cameras that have facial recognition stuftware built into them. Um, they can learn, they can extrapolate. We have the computer from Cornell that was able to extrapolate the basic laws of physics just by observing the movements of a pendulum.

I was able to do that in over a little over twenty four hours, which is pretty remarkable considering how long it took us to get there. But um, what what is this intelligence? Is that what we would define as intelligence just the ability to solve problems? Um. I have to like Marvin Minsky's definition of intelligence, um, which he compared to unexplored regions of Africa. More you learn about it, the less there is of it. So in other words, we defined and we we defined intelligence not

by what it is, but by what it isn't. So as we find ways to allow computers to do things, we have philosophers jump up and say, well, that's not intelligence. For example, a computer being a human. At chess, right, deep Blue Deep Blue beats Gary casper of In. Gary Casperov was the the world champ chess champion at the time, and he had met Deep Blue once before and emerged victorious.

H I think it was ninety six when he played it the first time, played the computer the second time in ninety seven, and in a series of six games actually lost eventually. Um, ultimately Deep Blue came out the victory in that one. Well, you could argue, well, this is an example of artificial intelligence. Look a machine has beaten a human at this this task. But some people were jumping up and saying, oh, well, you know, playing ches us, that's not really intelligence, that's just you know,

it's just this kind of task thing. So well, if that's not intelligence, what is it. So what you're doing is you're you're slowly removing aspects that were grouped kind of a nomen in this weird, vague, anomalous form under the term intelligence, until intelligence is meaning fewer and fewer things, until eventually the word itself maybebe meaningless as as far as it applies to machines. So we've got we've established that intelligence is hard to define, Consciousness is hard to define.

We still need to get to can machines have these things whatever they may be defined as? Would a machine be able to possess this? Um? I'm gonna let you answer this one because I know, I know you're the expert, and I hate to get in your way. Well, um, and uh, er um, maybe no. Here here's the thing. It's without being able to define it in ourselves, it's really hard to say how would we imbue that into

a machine? Yeah? Well, I mean, frankly, depending on your definitions, they've already achieved consciousness, because I mean some people are are a lot more maybe lenient. I don't know. I think I think some people may have very very very basic definitions of intelligence and consciousness, and to them it means perhaps uh, problem solving and perhaps some form of self diagnostic And if if that's all it takes for you to hit the definition of artificial intelligence and consciousness,

sure we have devices that can do that. We do not have a lot of devices that can do things like adapt to a new situation. They can, they can possibly you know, and when I'm talking about a new situation, I'm talking about totally new, not just like Okay, well this this robot can um identify obstacles in its path and chart and alter the pathway to get to its destination. There we have robots that can do that. Um, I'm

talking about putting it into a totally new situation. So you put this robot that normally can just find its way from one point to the next, and give it some other sort of task that has nothing to do with that. You know, while the robot isn't programmed to deal with that situation unless you have built in some sort of learning program where it can look at a totally new task, a totally new situation and adapt to it.

I don't know that you could truly call it intelligent in the way that I think most people envision when they think of the science fiction version of an artificially intelligent robot or machine. Yes, so you can take a robot to the movies, but you can't put it in front of the concession counters say I'm sorry, we're out of diet Dr Pepper. What else would you like and

have it actually make an informed, conscious decision. First second choice would be no machine and its right mind would ever ask for diet Dr Pepper in the first place. So your scenario is just laughable. But well, you know, I'm trying. These are difficult philosophical questions. Yes they are. These are not These are not These aren't pat answers that we're trying to give either. These are Um let's think about what, well, how would we even recognize if

a machine had consciousness or intelligence? Yeah? See, it it all goes Yeah, it's right. I mean, like, if you ask it a question and it responds, does that mean that it was intelligent or does that just mean it had an excellent database from which it could pull a response. Yeah. Yeah, Now I'm actually going back to Igor Alexander, who by the way is the emeritus professor and Senior Investigator in Neural Systems and Imperial College in London. UM. He's got

five axioms that helped define that, he say. He says defined consciousness UM, and that is a sense of place, imagination, directed attention, planning and decision and emotion. And so basically, you create a representation of a scenario in your head. You can think about it the way it was in the past, you can think about it in the way it will be in the future. And that is basically more or less in a nutshell and an oversimplified fashion,

done by me your consciousness. And he said that basically a computer probably could be programmed to do that UM in some fashion. I just the idea of actually accomplishing that is sort of I don't know beyond me how

you would actually accomplish that. UM. One person who is trying, however, UM is somebody who mentioned in his article UH scientists at the Technical University of Denmark named Running Codroll, who is looking at human and animal brains and the neurochemical interactions that take place inside the brain, and he's trying to when he identifies these reactions in a brain, he tries to replicate that on a computer to see if he can find out where consciousness is and make it

happen on a on a machine, which is kind of a weird I mean, that's that's so scientific that you know, it's a monumental task. I mean, we've got, for example, there's the there's the Blue Brain Project which is in Europe now, the Blue Brain Project. Their goal is not to create an artificially intelligent or conscious uh entity, That's

not their goal. What their goal is is to create a synthetic model of the human brain so that neuroscientists and neuro neurologists can can find new ways of treating the human brain for you know, various diseases or conditions, that kind of thing without ever having to you know, get into a human's brain. You know. The idea is that you could simulate these things and then see how the brain might respond to specific kinds of treatment, because

again you're using a simulation of the human brain. Some people, um, I read an article in the Daily Mail, which are friends over the UK. Maybe giggling right now. But in the Daily Mail there was an article written where the authors seemed to take the leap of if you were to simulate a human brain with all the connections and and uh the abilities therein, would you then give birth to an artificially intelligent, artificially conscious entity. Uh. The scientists

there say that's probably not gonna happen. It's just way too complicated. Um. They're just trying to create, uh, like I said, a simulation, not not recreate the brain itself. Um. And they point out that the brain is incredibly complex. You're talking about trillions of of neurons, and to try and simulate that in a in a meaningful way is would require so much computing power that it's it's really

hard to even conceive of. I mean, yeah, we've got computers out there that are really powerful, but you'd have to design very specific software for those computers to run in order to even come close to simulating what goes on in the human brain. So uh, now, this this doesn't mean that people haven't tried to create machines or

or programs that mimic humans in some way. I mean, the whole concept of the touring test is The Touring test is a test that that when you give it to a machine, when a machine passes the Touring test, when humans are not able to determine reliably whether it was a machine or a human taking the test. Yes, the name for Alan Touring, the famous British computer computer scientist. Yes.

So the idea here is that let's say that you have a series of questions that that you would pose a an unknown respondent, and then you get the response responses back. At the end of the test, you would be asked, all right, well do you think that was a person or do you think that was a computer? And you would give your response, and if you were unable to determine that it was a computer better than chance, and usually significantly better than chance, uh, then you could

say that the computer passes the Touring test. However, this does not necessarily mean that the program or machine or whatever is actually intelligent. In fact, Touring himself said that

that was not really a definition of intelligence. He said, it's possible that a computer that you might think of as intelligent because it is capable of responding to new stimuli, that's capable of of solving problems, an intelligent computer might fail the touring test because it wasn't designed to give responses.

So so you can have an intelligent computer that fails the touring test right on the same by the same token, you can have a computer that passes the touring test, but does so because it has such a huge database from which it can draw responses that it seems like it's far too sophisticated for it to be a computer program.

It's still not being intelligent. And there's a an interesting problem that um that another uh scientists suggested, um cirl I believe it was who suggested the Chinese room problem. Have you heard of this? I believe so, but refresh my memory. Alright, So here's the Chinese room problem. This is this is comparing a humans experience with let's say a machine's experience. I could pass the tourin test. The problem is that let's say, UM, let's say I'm transported

to China. I'm putting into a little room in China, and there are two slots in this room. Through one slot, I get a little card that has a Chinese figure drawn on it alphabe figure from the alphabet and my character and I I UM look through a series of tables, and I see that there is a particular response that I should UM draw that based upon this figure that I've seen. So I draw the approa brit figure, and then I slide it out of the um the other slot.

Now I've given the correct response, but I have no comprehension of what it is I have just done. All I know is I have one card, and I've drawn a second card and I've given it back. That's the the the challenge of a machine passing the touring test but still not being intelligent. The machines responding the correct way but has no actual comprehension of what it's doing. Now, one UM criticism you can have of this problem is that this is just taking into account one part of

the machine. If you think of the data tables as part of the machine, if you think of the slots as part of the machine, if you think of the whole system as part of the machine, then you could say, well, now it's a little fuzzy. Maybe the machine is intelligent. It's just that this one part of the machine, the machine that's doing this one section itself, is not intelligent. But that you can't say the whole system is unintelligent.

That's the Chinese problem, and the logical response to the Chinese problem what is called the logical response, and then you're eight. Unfortunately, our consciousness is don't always operate logically? No always? Are you talking hardly? Ever? Now, um, I've got an interesting little conversation to to read to you. Okay, okay. This is a conversation between um, Well, have you heard of Perry Perry p A R R Y. It's one

of the earliest artificially intelligent programs designed. Um, so this is this was one of those tests that was done where you know, you're familiar with the various artificial intelligent UH programs where you type in a question and the computer responds to you and then you type back a smarter child, right, this is like going back to the

touring test type stuff. Well, this Perry was written by a Stanford University psychiatrist named as Colby, and it was a program designed to respond as if it were a patient suffering paranoia. And the idea here was that under the parameters of a patient suffering from paranoia, it was actually fairly um um convincing. Now, granted that's a narrow set of parameters, that's not the whole human experience. We talked about this once before, I think in artificial intelligence.

So here's here's an example. Uh, why are you in the hospital. I shouldn't be here. Who brought you here? The police? What trouble did you have with the police? Cops don't do their job? What did they do? Instead? Cops arrest the wrong people? They arrest the right ones. Sometimes the law should get those Italian crooks? Are you Italian? My background is British, but I was born in this country. Do you know any Italian crooks? I tried to avoid the underworld that was, you know, so it's really kind

of gone off the reels there. Yeah, but still you you would if you were just looking at that as text. You know, you don't have any tone, you don't have any inflection or anything. It it reads like a paranoid patient. It doesn't. It's not. It's close enough to what the questions are that you could conceivably think, all right, well, that that could possibly be a person on the other end of this, not a computer program. Yes, clearly I'm

not a psychologist, not even remotely. But and I think though that when I was listening to you say that I was feeding my own I know this, you know, my own consciousness, that that was a computer. On the other end, it began to seem very random to me. If I did not know if I was a computer, then I might if I had perhaps presented you with a series of conversations, some of which were from real patients and some of which were from Perry, you might have had more of a challenge. So that was not

very scientific on my part. Yeah, but no, I mean I thought the same thing when I first read it, and then I thought, well, wait a minute, let me put myself in the position of I don't know from whom this is coming. Well, that makes more sense. You still might think it's a little unusual. But then again, if it you believed that it was coming from a

human being, you might think that there's something wrong. So, for argument's sake, let's say that somehow we have found a way to create what we would traditionally consider artificially intelligent, artificially conscious machine. What then, Oh good, you gave me the easy question. So let's say that put us on the road to the singularity. Well, it could, or it could put us on the road to uh Morocco. Oh,

wait now that's a movie terminator, is what I was? Um, you create these artificially intelligent creatures that can now not only are they self aware, but you know, they can truly examine the position that they are in and UM and their relationship hup to us and perhaps even decide, hey, you know what? UM, Yeah, I can continue to create better versions of me UM at a much faster rate than evolution will allow humans to. We get to get better over millions of years. So why don't I just

wipe these guys out? And because I don't need them anymore, I got everything I need and UM, I'll do it myself. Uh. I suppose if you had consciousness without emotion and you are logic based, it might seem logical to you know, or even with emotion. Just because you have emotion doesn't

mean you necessarily have compassion towards humans. Good point, you could have compassion towards other robots and think, well, why are we even though yes, we don't have the benefit of being organic, but why should we be forced into servitude? You know What's that's of course, that would be the whole argument of AI the movie AI UM, where robots weren't necessarily uh, malevolent towards humans. They just they had

been used as tools. And once your tool has the ability to figure out how lousy life is, then you have a lot of questions to answer, you know, a lot of ethical problems. Um. And they're planning of science fiction novels and movies and television shows that have explored this. Uh. And again it's another philosophical question. Now, garranted, I think it's pretty much a moot point because I'm not sure

that we're ever going to get there. For one thing, I don't know that we necessarily need anything that has that kind of artificial intelligence or conscious to it. Yeah, I mean it's I think people are investigating it more because they're curious, then because there's an actual need. If nothing else, I think it'll lets us examine ourselves more.

But when it comes to creating, you know, actually, let's say that somehow we had the technological breakthrough where we figured out we could do it if we wanted to. I don't know that we would want to. I mean, for one thing, we're kind of assuming in in this case that the machine brain we create is at least on some level comparative to a human brain. Well, it turns out human brains aren't hard to create. It just takes two people. We can make lots and lots of

human brains. There's no shortage of them, as it turns out, which is that's that's kind of why. That's one of the arguments that you know, even AI specialists will say that, well, there's no point in creating that. Why would we try and create an artificial human? We can create real humans. In fact, that's kind of a problem in some places. We have way too many real humans. Why would we create fake ones? Um, I think it's a good argument.

You know, it makes more sense to create machines that are really really good at doing specific tasks and not worry so much about creating a machine that is capable of doing any task. Yeah, that doesn't really make any sense. I mean, it's we should build stuff that that does certain things really well and not worry about trying to create a mechanical replacement for the human race. Yeah, we all know how that's gonna end up. And see that's

why I vote against robot rights. I do not give the right to vote to my toaster, yes, because it will probably vote for bagels, and I prefer toast. Okay, I do say so, so at any rate, I guess we can wrap up this conversation. It has been rather high level and uh and and a bit intellectual, and it gets pretty cerebral at times. It does have to intelligence, and we didn't even get into the mind body problem, which that's another philosophical issue. Um, Will we ever get there?

I don't I looking at what we're doing now, I'm not sure that we're going to get there within our lifetime because it's just it's incredibly complex. Now, if you're right, Kurts, while, he'll probably think that we're going to get there because well, yeah, Kurts, while uh, he's the guy who came up with the

theory of the singularity. UM. He believed that we're going to create machines that can create better machines, which in turn can create even better machines, and the time between generations will get shorter and shorter until we're we reach a point where we're just in a constant state of evolution, and then either humans become obsolete they merge with this new form of technology UM, or they are we're all freed up to you know, swim in the pool and

and watch television and while the machines do everything we need them to do. I don't know that that's coming. I think that might be a little pie in the sky. We never know. We could end up in a ragtag fugitive fleet. We could end up in a large deep space mining vessel and then have to go into uh To to to frozen carbonation for like a couple of million years while the cat evolves. You have no idea what I'm talking about. Vaguely, you've seen Red Dwarf. I'm

familiar with the plot. Okay, I've seen a couple of episodes. All right, well, I guess that's a good enough reason for us to wrap this up then, is right there? We're done. Um, let's just move on to a little listener mail. This listener mail comes from Chris, and Chris says, Hello, Chris and Jonathan. My name is Chris, and I have been an avid listener to your podcast since I discovered it last year. I had to take a break from

listening to them as I went into the military. Anyway, after I got discharged and got back to work, I caught up on them as I was listening to your podcast on Augmented Reality my company gout in order for parts for an augmented reality system for a car. I thought this was really cool, so I thought I would shoot you an email your number four hundred sixty third fan Chris p S. I'm not sure if I like you guys or Josh and Chuck better. That's all right,

some days I don't know either. Um, that's awesome, Chris. And you know, GM just announced an augmented reality windshield project. Is pretty darn nifty. Now. We've seen some of this in the past already, where we've seen some windshield applications for things like night vision and a rear view where you you have a little uh cameras showing the rear view on the inside of your windshields, so you can

see what's behind you, um, and stuff like that. Yeah, this is this is more of the same, like more heads up displays and where you can get real time information on the road as you're driving, like GPS stuff and things of that nature. Um. I'm really excited about this kind of stuff as long as they can present it in such a way that does not become a

distraction to the driver, and that's a big issue these days. Yeah. Yeah, it's I mean, anything that keeps takes your eyes off the road is always going to be a bit of a risk, more than a bit, depending on how fast you're going and where you are. But I do think it's a pretty exciting application. Um Personally, I'm still waiting for the sunglasses that don't weigh like thirty pounds. I should jump on that because you could theorectly build a pair now, it's just they wouldn't be very light. Thanks

a lot, Chris. If any of you have any email you would like to send us, our address is tech stuff at how stuff works dot com. Go to how stuff works dot com to learn all about artificial intelligence, robots, computers, processors, things like that, and Chris and I will talk to you again really soon for more on this and thousands of other topics. Does it how stuff works dot com and be sure to check out the new tech stuff blog now on the house stuff Works homepage, brought to

you by the reinvented two thousand twelve camera. It's ready, are you