What does artificial intelligence mean?

and welcome to tech stuff. My name is Chris Pouette, and I'm the tech editor here at how stuff works dot com. Sitting across from me as usual, is Senior Ryan Jonathan Strickland's up, y'all very nice, changed up a little bit, Okay, I'm fine with that. So before we get started, I just wanted to mentioned something. I saw this email going out about a new show that's coming up on the Science Channel. Yeah, it's called The Road to Punkin Chunkin. And then immediately it's followed by punkin Chunkin.

Oh yeah, yeah, yeah, you saw this email too. Yeah. It's a Thanksgiving night here in the United States, so yeah, starts at eight. That's the Road to Punkin Chunkin, followed by Punkin Chunkin at nine. And if you don't know what this is about, it's really about designing machines that can throw a pumpkin a really far distance. Yeah yeah, yeah, yeah. They used to do that with with tribute sches and stuff like that, but now they have these air powered cannons. Yeah,

it's gonna be pretty crazy. So, uh, you should definitely check it out. I'm gonna be watching it myself. I think there's nothing I like more than a little pumpkin destruction. Awesome. Well, so what are we gonna talk about today? Today's topic comes to us courtesy of a little listener mail. Well, all right, then this is list Your mail comes from our ninth grader listener named Alexandra. I'm sure we have other ninth graders, but this one in particular road to us.

Thank you, Alexandra, and here is the message. Hey you guys, I have a suggestion for a topic, Artificial intelligence. My favorite show is Night Writer two thousand eight. Kit the Talking Car is said to have the most advanced AI in the world. Is AI in cars and other devices coming soon? Will they have personalities? Thanks for a great podcast.

P s. I like the eighties Night Writer too. So you're talking about night Writer, you're talking about KIT, which of course is the Night Industries two thousand artificially intelligent car. By the way, do you remember what Kit's arch nemesis car was called? Oh, we'll see I now now. I didn't remember what it stood for. I knew that it was car, Yes, Car K A r R. It stands for Night Automated Roving Robot. Do you remember the main character's name? Besides Kit and Night Writer met the main

character Michael, Yes, Michael Night. Who do you remember what his name was before became Michael Knight? Oh? No, actually I don't. Michael Arthur Long undercover agents shot in the face. He had to be rehabilitated. They rebuilt his face, gave it David Hasselhoss face for some reason. He's huge in Germany and U. He became Michael Knight because it was Night Industries. Did I just have one question for you? Did you did you look that up? Or did you know that off the top of your head? I wrote

all that down from memory. I like Alexandra, I was

a bit of a fan of the nineteen eighties night Writers. Well, I thought I was, but apparently I was a lower case F fan and not an upper case I also wanted to throw out throw out the phrase I was saying, the the little joke I made to you before we started recording this podcast, Michael, let's drive to Congress and if any of you know why I said it that way and what the whole Congress thing is, reference to write me tech stuff at how stuff works dot com.

First person to get it right gets a kudos. Just not a granola bar. Not a granola bar. I mean, if I encounter you on the street and I happen to have a granola bar, I will be happy to hand it over to you. If I don't have a granola bar, we're kind of ada luck on that one. So this brings us to our topic granola bars. No, I'm sorry, artificial intelligence. I got off on a tangent there, So artificial intelligence now, she asked, is artificial intelligence coming

to other devices and cars as well? And will they have personalities? Um, here's the interesting thing about artificial intelligence. The definition is somewhat vague. For example, Chris, how would you define artificial intelligence? I mean, nope, there are no right or wrong answers. Just when you hear the term,

what is it that you think of? Well, you know, of course there's the movie Okay, yes, there's the film that was started by Stanley Kubrick completed by Steven Spielberg that lasted thirty minut it's too long and should have ended with a Little Boy in the Ocean spoiler alert.

But beyond that, you know this, This is the effect of Odyssey of the Mind on my brain, because I'm trained to think of everything but the actual answer, Um, no, I want I When I think of artificial intelligence, um, you know, actually, one of the first things that comes to mind is games because people talk about the AI in a particular game, all the AI and this game is really good. Yeah, and uh, you know of course robots mostly androids, things like that. Yeah, but yeah, I

mean those things and and and walking talking. You know. It's always the movie version of our official intelligence where there's you know, something and of course it's evil. It's usually evil example, you know, and it's thinking, and it's it's decided that we're irrelevant and it's going to kill us all. I mean that, isn't it nice? It's pretty much gonna happen. But no, no, no, so here's here's here's my ulswer. Yeah, but to know that it's coming and it will kill me. Yes, my Hunter Civic will

eventually do me in. Um No, here's what I think of when I think artificial intelligence. I think of a an artificial construct. So some sort of man made thing that is capable of sensing its environment, reacting to its environment, and learning from its experiences. Now, to me, the learning part is the most important because if it's if it's only taking in it's uh, you know, some sort of sensory input and then reacting to it, there's not really

any thinking going on there. I mean, essentially, you're just it's it's stimulus response, right, Yeah, And that's pretty much how I carry out my day. I just some of us are a little higher functioning than Chris, and we we occasionally spare a thought or two for something. Um No, Chris is actually a very intelligent person. I I people get on to me about picking on you. You know that, right,

I don't. Chris and I are actually good buddies here. Yeah, yeah, so at any rate, at any rate, So yeah, I think, uh, I think learning is an important element in artificial intelligence, so that not only are you able to react to something that's in your environment, but you learn from that experience and you are able to handle similar experiences in

a better way in the future. So, for example, you encounter if if you have an artificially intelligent construct that is UH susceptible to melting and it encounters a heat source that is higher than it's melting point, and it detects this and then backs away and then has to try and UH factor in a route around this heat source so that it can continue doing whatever it was

it was doing beforehand. UM. After encountering that, it might be able to build on that experience and be able to solve a similar problem in half the time, and ideally you have this increminal improvement. UM. It's also called self recursive improvement, where you have an experience, you learn from it, and you are able to do it more efficiently or effectively, however you want to quantify it in the future. So right now, using that definition, we don't

really have a truly good example of artificial intelligence. There's some that are really close, UM, but there's nothing out there that's that's like a computer that can understand how computers work and therefore design its own successor to be a more powerful version of itself. So um, yeah, no, uh, what was the computer in Hitchhiker's Guide Deep Thought? So

deep Thought? Yeah, deep Thought? And uh in Hitticker's Guide to the Galaxy, there's this computer called deep Thought that is so intelligent that it is able to come up with the answer to life, the universe, and everything. It is not able to come up with the question it's the answer to a question, but it's not. But it can actually get the question. To get the question, it has to build a computer even smarter than itself, but

it's able to do that. So it can't solve the problem, but it can build a tool that can solve the problem that ends up being the planet Earth in that in that series of books. Now, that's actually an idea that's being explored in computer science. It's an idea of can we build a machine that can understand how machines work and therefore build a better version of the machine because it can make these calculations far faster and more

effectively than any human can. And once you reach that point, that's when we hit what we've We've talked about this in the past. We've talked about in another podcast. But the singularity, which is where you've reached the point where machines or or even humans. It doesn't have to be a machine. It could be like somehow we have biologically altered ourselves so that we can do this to ourselves.

But we the improvements come so quickly that uh, that there's no way to distinguish one moment from the next as far as technological advancement goes, because it's just going easing fast and we've reached a brand new reality, at least as far as our perspective is concerned. Uh, to the outside observer, it wouldn't be quite the same. And I did mention that we have some stuff that's coming close. The one I wanted to talk about just mention, uh was the Cornell Research um UH project where they had

the computer that analyzed the pendulum. Do you remember reading the story There was a new story. It was back earlier this year. There was a group of Cornell computer scientists. They had designed a computer and all the computer was doing was observing the movements of a pendulum and then from those movements working out the laws of physics, so it would observe the way the pendulum moved, the speed at which it moved, how much it slowed down, and

it started to work out the laws of thermodynamics. It's you know, physical laws, classical physics. And it turned out to be really good at it. It didn't get it exactly right at the very beginning, but it kept observing just like humans would. It kept observing the the phenomena and then making guesses and and then testing those guesses, throwing out anything that didn't fit over time, and keeping everything that did fit, and eventually it was able to

work out the basic laws of physics. Now that I had not read that. That's a pretty cool example of artificial intelligence, and the idea that this was able to observe something in its environment and draw conclusions from it. Um Now, it had been programmed to be able to observe, so I mean, it wasn't like it had developed capabilities beyond its programming. I think that's another sign of artificial intelligence. When you've created something that can go beyond what you

had originally programmed it for. That's another sign of intelligence. You know a lot of people think that used to the if you will, the classical definition of artificial intelligence. You know, the touring tests. Yes, the Touring tests Alan Touring, and I was a little surprised you didn't start with him, to be honest, Yeah, I I thought about it, um, because this is a when you get to artificial intelligence, it really is a philosophical matter. It's not just technological,

it's not just computer science. It is philosophy. Well, you know, uh, when I was doing research for the podcast, I found

a great resource at Stanford. And it's funny because when I think of Alan Touring, I think of computer science, but it's actually the Stanford Encyclopedia of Philosophy where I encountered his thoughts on artificial intelligence and the imitation game, which is the when I think of you know, when when I actually read it, that's what I think of as the Touring test where they put a machine and a human being, um and sort of a blind test

with someone asking questions, and theoretically the person should be able to ask the machine and the person questions and they should be able to give answers, and the person asking the question would be unable to determine which one was real and which one was meant, I mean, which one was the machine. It's hard to say when you're using the word real is kind of tricky, but yeah, I understand which was human and which was a human? Like? Yes, yes,

which one was human and which one was was a machine? Right? So ideally you would have these these the computer and the uh the other test subject, the human test subject um separated. They would be completely isolated from the person asking questions, and at the end of the list of questions, you would ask the question er, all right, which one was human, which one was computer? And if the questionnaire was unable to identify the two roles better than chance.

So you know, you have to do this test many times over right, right. If they were unable to identify computer versus human uh to a point where it takes chance out of the fact the whole factor, then um, you can say that the computer passes the Turing test, that the computer is, by all intents and purpose this

indistinguishable from a human being when it comes to communication. Now, um, there have been some touring tests that that sort of achieved this in a way, one of them being one that focused on um, a computer that simulated the thought processes of someone suffering from schizophrenia, and they had a group of psychiatrists ask questions over, you know, through text um to both a real person and this computer that was simulating someone with schizophrenia and something like of them

weren't able to tell the difference. They weren't able to reliably identify which one was computer which one was human. So that's a that's right around the chance level, right, so that you could say that that computer passes the

touring test. However, that is of course a very narrow set of parameters, and some people have criticized the test saying that, okay, but when you're getting to someone who is operating on are a set of faculties that are not considered the normal set, because schizophrenia is outside the norm, then it wouldn't really The computer doesn't need to understand what it's saying, right, The computer doesn't have to be able to comprehend that the words it's putting together are

following a certain you know, pattern, are making sense. Um, you could just as easily have a jumble of words. The computer could randomly pair them up and create nonsensical sentences, and it would be almost as effective because you're not working with the full set of human faculties. Um. If you were, then the presumably the percentage rate of identifying

the computer versus the human would be much different. You would be able to probably accurately describe them, and much more often because computers are not terribly good at carrying on a truly human conversation, or at least not the kind of conversations I have with other humans. Did you did you ever here? Did you know of dr Spates? So, uh, it's a sound blaster thing. The the old sound blaster cards came with a software called dr Spates, So asked

Dr SPATESO. And it was this computer voice. You could type questions to it. It would respond to you. It would actually talk to you, and there were times where it kind of felt a little uncanny that wow, that was actually a really good response. But then it was just a database full of responses and it would pick whichever one was most likely going to apply to whatever question you had typed in, So it wasn't really artificial intelligence. But I remember that was probably my first encounter in

a personal level, apart from the movies. Well, one of the things that surprised me as I was doing research because I was familiar with Alan touring in the touring test um, but I was a little surprised that Renee des Cartes had actually made some suggestions. I have lyrics going through my head. But no. In sixty eight, of course, again, according to the Stanford Encyclopedia Philosophy, UM the cards said that if there were a machine that could walk and

talk like you know, a person could. Um, I'm assuming he didn't have some kind of time machine. Was guessing that there will be androids in the future. But Um basically said that these machines would need to be able to put together a conversation like a person in order to really be defined as intelligent. Um. But he said that he didn't really believe that it was possible for a machine to exhibit that kind of intelligence because they would also need and I'm not quite sure what he

means here, all the organs necessary to live. I mean to react to everything that they would encounter. I'm assuming that there he's saying. You know, you wouldn't have skin, so you wouldn't be able to go, oh, we have that panics hot whorse, that would be a lot more. Again, this would suggest again that he does not have a time machine, because yeah, actually, now there are sensors that you could embed in Yeah, there's actually there's there's so much advancement in as far as being able to sense

the environment. Is concerned that we have some rudimentary artificial intelligence examples out there, things that are able to uh to examine and react to their environments. Um, things like you know as Amo. As Amo, the robot from Honda can walk around the room and recognize an obstacle in its path and then reroute its path so that it can get to its end point um by you know, walking around or whatever. Uh it, But it still has limitations. It can't it can't deal with anything completely new. It's

not capable of doing that. And even something like the it's able to go up and downstairs, which was a big deal, like it was able to balance while going up and downstairs, right, Daleks can't. That's all that takes to ruin the machinations of the Daleks. A staircases. Um, although I've been told that can fly, yeah, but I mean, come on a stairwell. It's not gonna fly up a stairwell anyway. So the stairs have to be pre programmed

right into the Asimo. It can't. It can't encounter a new staircase and be able to walk up or down that staircase all on its own yet, but it could encounter a new room and be able to walk from one point to the other and recognize that there's something in the way and move around it. So there are limitations even now. But um, you know, we we're seeing

advancements all the time. And uh, I do think it's just a question of when we get to the point of something that is capable of learning from its environment, like the Cornell example kind of proves that. Um, whether or not it develops a personality is a totally different question. And I actually think that the Decart example was fascinating.

I want or if perhaps part of the problem with defining artificial intelligence is that we have a very you know, we have a very narrow perspective of what intelligence is. You know, we can't we're humans. We can't really conceive of an intelligence outside of our own human intelligence in the sense that there might be some way of having intelligence that is so foreign to us that we would not even recognize it as such. Yeah, we're we're basing

it on our own understanding of what intelligence is. You have to there, there's no way we can't because we are human beings. But that's that's a good point to make, is that there could be let's say that ten years down the road, we have an artificially intelligent device, but we can't really recognize it as such because it's not intelligent the same way human beings are. And I do think there are artificial intelligences. Yeah, I've met a lot

of them, mostly on election day. Oh man. Now, I mean but I think, you know, going back to your to your Asimo example, Yeah, um, I I think they

are intelligent. The thing, the things that we manufacture that have an artificial intelligence to them are artificially intelligent within the boundaries that we set for them, like there are you know, I've played games that have the ones that are described as having good artificial intelligences, and the ones that are that are really good are the ones that seem to adapt to the way that you play. They say, Okay, I know he's gonna come at me from this angle.

Now I'm going to do this to defend myself and go in this direction. Um. But of course, if you wanted it to make you a grill cheese sandwich, you know, really that's not gonna help much. Yeah. The intelligence intelligence is limited to the set of instructions that were given to the intelligence at the very beginning, right, So Asimo couldn't direct an orchestra, but it could climb a pair a set of stairs, and then they programmed it to

conduct in order. Was just about to say conduct it did, but it was actually it was actually um mimicking someone else's movement, accorded someone else's movements, programmed that into Asimo. So asthma was not that was that That was like putting a tape into a VCR or a CD into a CD player and pushing play. There was no way for Asimo to depart from that series of movements. It

could not It couldn't interpret the music. It couldn't stop and try and hold a note a little for a little more time to get the biggest emotional impact of the piece. It could only conduct at the the tempo and the all the motions that were involved. Um that we're programmed into it beforehand. So that's a very good example. One thing that that Honda's robot will do. I mean, since it is from Honda, it can reach in a chord. Oh my gosh, uh Michael, um so that I see.

That was that was not Honda. I was gonna talk a little bit about capture as well, and it's the way it plays into artificial intelligence. Now, for those of you who aren't familiar, capture, of course, is the uh it's it's used a lot in verification and when you're you know, signing up for a new account with something, um it's quiggly alphabetical characters that look like they've been

type written around a piece of string or something. It's it's some Usually it's a It can be either a string of random characters or if you're lucky, it's a word.

Because some of these can get so difficult to understand that you know, you can't like, is that a B or is that a you know A Yeah, if it's a word, you can at least you know, your brain will put that together, which is as another good example, like if a computer is not able to figure out a word based upon a couple of you know, letters in there, then obviously that shows that it fails that test.

So the idea mind capture is that it's supposed to be a test that human beings can easily pass, but computers cannot, And the purpose for capture is to try and avoid having automated bots register tons and tons and tons of spam accounts for whatever, you know, whatever the thing is like. Ticketmaster is a good example. Let's say you're buying tickets on Ticketmaster. There's a capture section where you have to fill it out before you can complete

your order. This helps prevent scalpers from setting up an automated program to buy all the tickets from for a certain venue for a certain show, and then scalping the

heck out of those tickets. All right, so it makes sense now, Um, there are people who are figuring out ways to program computers to recognize captures, and in a way you might say, oh, well, this is terrible because it means that the security now has taken a back step and now they're gonna have to figure out a new way of doing this, which is gonna make it even more difficult for humans to figure out the damn word. Right, Well, but you're you're talking about. The silver lining is that

it means that it's an advance in artificial intelligence. So on the one hand, computer security has gotten a little more difficult. On the other hand, artificial intelligence has gotten a little better. Now. Of course, this doesn't apply if you happen to employ hundred people out in Indo, China to manually go and type these captures in because that's what a lot of you know, quote unquote hackers are actually doing. They're not creating a program that understands capture.

They're just hiring someone who will work for pennies on the hour to type these things out. Um, that's obviously not artificial intelligence. That's just being a scumbag. Uh, because there are better ways to help people in need than to employ them to create junk mail accounts. Um, well there goes my business model. Yes. So yeah, So that's the whole idea behind the capture is that even when it fails, it's still a success. It's just a success

in a different field. Alright. Then it's kind of it's kind of a weird way of thinking of it though you would think that would be sort of a negative all the way around. And and you know, we've seen other elements of artificial intelligence pop up in devices. You have things like the collision detection in cars, where it alerts you before you actually were to uh collide with something. And then there's it appears that you're trying to write a letter, Yes, there's Clippy Clippy, which is not really

that intelligent. No, it's artificially incredibly irritating. Actually, you know, I haven't seen Clippy around. What's he doing these days? Hopefully hard time because I hated that thing. Oh my gosh, I give you a hard time. Yeah. Yeah, you type in three words and it suddenly decided what you've tried to do, and it's trying to be helpful even though you didn't ask for help and you didn't want it. I hate Clippy. It appears that you're trying to extort

some money. It appears that you're a Nigerian prince. Who is? That's just thanks Clippy, you've made scamming people so much easier. But artificial intelligence, it's one of those truly multidiscipline type of uh fields, right, I mean it has it has everything that it's everything from philosophy, computer science and euroscience. Uh, I mean biology even because there are people who are staying the human brain and trying to figure out how

that works and apply that somehow to computer science. That's the tricky tricky thing because what what we know about the computer the human brain is so incredibly it's such a tiny fraction of what there is to know from what we can tell at any rate. I mean, there's so much we don't understand. What is the mind versus the brain? We can't answer that question, which is kind of ironic when you think about it. That would be philosophy right there, Yes, philosophy, but it's also biology. I mean,

it's the same thing. It crosses over. The mind versus the brain is one of those questions that various disciplines have been trying to answer for for generations. And um there's some who who believe that a and artificial intelligence could develop a mind, even if we don't ourselves understand how the mind works. If we build a brain that's good enough, a mind could spontaneously generate from that brain. If the brain was effective enough and could could make

its own decisions, it could develop its own mind. This is completely theoretical because we don't have the ability to make a device that complex right now to test this out. But it's also kind of terrifying because then we arrive at the scenario you mentioned at the beginning. The intelligence takes a look around, looks at what it needs, decides whether or not humans are needed. If humans are not needed, or should they be eliminated. If they should be eliminated,

what's the most effective way of doing this. If it's got the self recursive improvement, it's going to figure out better and better ways of getting rid of people until you don't have people anymore, or you get assimilated and you become part of the borg. Well, you know, that's a cheery outlook on it. Or it absorbs our intelligence in some way and we all become virtual people and are fleshy, meaty bar These eventually die and decompose and say farewell and shovel off the mortal coil. But our

intelligence lives on. Now that's kind of an interesting thought. It also is kind of depressing because he realized that it's not the intelligence that's living in you right now. It would be a copy of you. You know, I'm guessing that when Alexander wrote to us that she had no idea whatsoever that we were going to get into this deep conversation about this stuff, that's because I'm not sure that I and my brain is kind of hurting at this point. Yeah, it's well, this is a this

is a tough topic. Now we're all to answer Alexandra's question artificial intelligence, to some extent, is already in a host of different products, from cars to even things like smartphones. Smartphones able to detect where you are, how fast you're moving, all this kind of stuff. Um, it hasn't gotten to the point where they can start to make their own decisions or or improve upon pastic sperience. But I don't

think we're that far off. There are people who believe that artificial intelligence may maybe there may be a breakthrough as early as next year. Of course they said that last year, right, Well, actually in two thousand and seven they said any time between three and seven years, So next year would be the first year for that major breakthrough and then for the singularity you're looking at around depending upon who you ask. So uh, I mean that whether we actually hit the singularity, A lot of that

depends on if other things hold true, like Moore's law. Yeah, but of course you know, Ray Kurtzweil and others are counting on it. Yeah, they've they've got an entire lecture series devoted to it. And they are very smart people, I would say, far smarter than I am. So it may be that they're onto something. Um, it's hard for me to say because it's you know, I also look at other extremely complex machinery that doesn't yet work. Yeah,

like you know large Hadron collider. Well, there are theories as to why that's We're going to get into that, though we can't. We can't talk about that. That's coming up in an upcoming episode. I wonder if artificial intelligence is behind that. Sounds like a conspiracy to me. That's a hint. All right, Well, let's uh, I think are do you have anything else to add on artificial intelligence? This was a very philosophical discussion, very different from our

normal episode. I hope you guys enjoyed it. We don't plan on doing a ton of philosophical discussions, but it's gonna do something a little different. It is. It's nice thought experiment, which, as it turns out, very important in artificial intelligence. But now that we've exhausted that topic, at least as far as our understanding goes, you could you could take entire college courses in artificial I mean, there's far, far, far more out. Yeah, this is not even scratching the surface.

That would be an insult to the field of artificial intelligence. But it does now wrap things up, which leads us to our second bout of listener May This listener mail comes from Thomas, and Thomas has this to say, Hello, Chris and Jonathan. I just recently listened to your podcast on a r g S and I was wondering if I could get a few pointers. I'm getting started and participating in one. So r g s are the alternate reality games, which we talked about several episodes ago. This.

Two sites I would recommend anyone go to if they are interested in getting into alternate reality games are www Dot a r g N dot com or www dot unfiction dot com. Both of those are good place to start if you want to look into alternate reality games and kind of get an idea of which ones you want to play. Some of the old games are still semi active, so you could actually play through parts of

the game now. Granted, anything that realizes on interaction between the game master and the players that's long since over right, so you can't you can't play that part. But the puzzles that are online you may still be able to play through, or at least be able to see what the puzzles were, how people work them out, get an idea for the way the games are played. Um. Both of those sites also list active games. When I looked, there weren't a whole lot of active games right now,

but that could change at any time. So thanks Thomas. If any of you have any questions, suggestions, criticisms, comments, that kind of thing, right to us tech Stuff at how stuff works dot com. We have a lot of articles on the site that relate to artificial intelligence. Uh not one specifically about how AI works, but how it relates to things like robots, video games, all that sort of stuff. I recommend you check those out. That's at how stuff works dot com. Remember we have a live

show every Tuesday one pm Eastern. Check that out again. Blogs, So you would look at the House of Works dot com on the right hand side there's the link of the blogs. That's where you're gonna find that information. 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