From the Vault: Into the Uncanny Valley

simulations of the human likeness. Yeah, yeah, this one, this one is one that that I think back to a lot as well, just because, yeah, you're always encountering some imperfect digital recreations of a human likeness, and then we have to sort out how we feel about it, but also how we've perhaps been conditioned to feel about it just by knowing about, uh, the idea of the Uncanny Valley.

So yeah, this is gonna first of two episodes of this episode is Into the Uncanny Valley, originally published April four, and then in the subsequent episode, we're going to go Beyond the Uncanny Valley. Welcome to Stuff to Blow Your Mind from how stuff Works dot Com. Hey, welcome to Stuff to Blow your Mind. My name is Robert Lamb and I'm Joe McCormick. And Robert, I want to take you back to a conversation we had. I think it

was last December. It was right after I went to see the new, the most recent Star Wars movie, Rogue one. Oh yes, uh, And I am on the cusp, the very cusp of seeing it myself and waiting for it to become rent rental options. So oh it's not yet. Still got still a week or two out. So has everything been spoiled for you so far? No? People have

been a little um cooler on this one. Um. I think some things are probably spoiled for me, but not like that last one where every just really felt the need to, you know, just lay it all out on social media. Let me let me spoil one thing for you here and they go to space and there's a war there. Yeah, stars among the stars. But there is one thing in the movie. Okay, so mild spoiler for Rogue one coming out. It's something that probably everybody already knows.

It's also not really about the content of the movie, but just about which characters you see. But if you are ready, are you ready for the mild spoiler? Okay, we get to see if you remember back to the original Star Wars. Back go back to the seventies. Peter Cushing as Grand mof Tarkin, the guy who was in fact Darth Vader's boss on the Star Him. Yeah, he's enough of this, Vada release him. And we love Peter Cushing because he was in all these old monster movies.

He goes back to the Hammer movies. He was. He was Dr Frankenstein, like the villainous Dr Frankenstein of the Hammer films. And he I think was the hero of the version of the Mummy that has Christopher Lee as as the Mummy. Uh, the one. I've got the poster for it in the House six up. It's the Belgian posters, so it's La Malediction de Feron. But yeah, so Peter Cushing was the original Grand mof Tarkan, this bad Empire

guy who was Darth Vader's boss. And the thing they do in the New Rogue One is they bring him back. He's dead, he has passed away. This movie takes place a little bit before the original Star Wars is supposed to have taken place. But they bring this character back and they have an actor stand in as him, but

it's not just a recast role. They try to make it look as if this is Peter Cushing standing here delivering the lines with c g I. And this is an odd choice because all right, so you're gonna have Darth Vader in there, that's easy to do. Darth Vader is a dude in a suit, voiced by James Earl Jones. James Earl Jones is still alive, so you can check that one off the list. But Grandma Tarkin, like you said, uh, the actor is dead. So it seems to me like the first easiest thing to do is just don't have

those scenes. If you know it's going to be ablematic, don't even mess with it. Or um, just use an actual living actor such as Wayne Pigram who played him in Revenge of the Sith. Or go with Ben Cross, who's another actor that I've seen over the years brought up as potential Tarkan casting, or head go with Ralph Finds, Like, clearly you have the money to throw down the well of expensive c g I equipment, Just go ahead and hire Ralph Finds. I know he's pricey, but he's great

and it's consummate evil Brett. Yeah, he even kind of looks like a younger Peter Cushing. He's got that same kind of angular face, like the thin long face with the jaw and the scowl. It's all there. And it's not like fans of various franchise are not clearly cool with recasting. It's not like we're gonna be thrown into, you know, a traumatic spin, because you can look to a Game of Thrones, James Bond, Twilight, Harry Potter, etcetera, like we we get it. We can roll with a recast. Now.

I want to go into completely different directions thinking about this c g I grandmof Tarken. One is that I didn't like it in the movie. Okay, I saw it and I was just like, I don't want this. It pulled me out of the movie. It made me stop being in the story and just thinking about how did they do that? I don't On one hand, it looked great like when you see the movie, I think he will kind of have to agree. It's unless I'm missing something. It's the best c g I simulation of a real

person that I've ever seen. Like it looks amazing, but it still looks not quite good enough that I can just accept it and go with it. I kept continually thinking, like, what am I looking at? It's almost really him, but it's not quite really him, and it made me feel icky. So in this it made you to send into what we've come to know as a as a as a

species as the Uncanny Valley. Right, So today is going to be the first of two episodes we want to do about the Uncanny Valley, and this first one we wanted to descend into the Uncanny Valley, but not us talk about it in terms of the standard pop culture phenomenon, because this is one of those side tech concepts that is totally filtered down into the mainstream. Everybody talks about

the Uncanny Valley. It's a totally normal, ground level pop culture phenomenon now, especially with as much bad C. G I as we encounter in the movies. But there it's also a scientific field of study. It's something that people are looking into with empirical research to try to figure out does it really exist, If it does really exist, what causes it, what can be done about it. So we want to look at it from both of these angles today, Right, so we should probably roll through some

just fun examples of this. We're gonna try and not to go too long on this. If we do, we'll cut it and save it for trailer talk. Either way, we'll probably do a Facebook live trailer talk on an upcoming Friday about some of these movies. Okay, so I want to go back to a much earlier experience for me, Robert, did you see The Mummy Returns in two thousand one? Remember this one? I don't think I saw The Mummy Returns.

I saw the Money and I remember digging it at the time, but not not the old hammer one or the universal one. You know, the Yeah, the the re the reboot of the Money. Is his name, Brendan Fraser. Yeah, and what Arnold Vosslo? Yeah he was he. I I enjoyed him as they kind of brought in some of these aspects of the tragic Mummy figure, which I liked. Yeah, yeah, yeah, okay. But in two thousand one we got the Scorpion King.

This is a character that appears in the Mummy Returns and he's pretty much if you want to picture it if you haven't seen the movie, actually you should look up video of this. We're gonna tell you to go look up images and video quite a few times in

these episodes because some visual aids will help. But if you want to picture it, picture the concept of a centaur, except replace the horse parts with scorpion parts and some other random arthropod bits, and the man part on top is Dwayne the Rock Johnson, except it's not Dwayne the Rock Johnson. There's a there's a bit of a problem

with the rocks. So the corpion Scorpion, the Scorpion King, scuttles into action in the film, and you can tell immediately something is wrong because it's not just the rock. It's this c g I upper body designed to look like the rock. It's supposed to be him, but it

doesn't look right. It looks like somebody took the rock, skinned him, and then took the skin suit and then boiled it and then maybe ironed it and rubbed it down with wax, and then stretched it over somebody else, like a bald Crispin Glover wearing a waxed up the

rock suit. And and that would be fine if that's what they were going for, But I guess the disconnect here is that clearly they wanted this to be like the Rock as a scorpion center and not this um creepy in human above the nation that you've described, right, And it's sort of I guess works because it's okay if he's creepy, because he's a monster. But he was creepy in a way that he clearly wasn't supposed to be. It wasn't just that, oh he's a monster. He looks weird.

Something looked wrong with him. And this was at a time when computer generated animations were hot, right two thousand one. They seem to be getting better all the time, and yet they were terrible producing these characters that were not only not convincingly human, they were literally physically unpleasant to look at. They were repulsive. Yeah, it was a period when everyone was just foolishly optimistic about what we could achieve with c g I, and you know, into in

in a sense, maybe that hasn't gone away. We're still very the Rogue one example, Like, clearly everyone was very optimistic about how great this looked, and even though to your point, it does look great, but within the context of the film, something doesn't quite work. Yeah, I would say, now for some people we can get into this more. I think, especially in the next episode when we then the next one, we want to try to go beyond

the valley, the Uncanny Valley. But I will say at this point, for some people, Tarkan was not over the line or under the line. I don't know where you put the line. But for for some people it worked, and I do think that's an interesting thing to acknowledge that while for me I I experience this, uh, not to the same extent as the Scorpion King, but a

kind of Scorpion King revulsion, not everybody did. Now, one thing about Tarkan is that the Tarkan c Jack character is correct me if I'm wrong, because I have not seen it myself yet. But he is interacting with human actors in this in his scenes, Yeah, I think, so okay, or at least he's in a film with other human actors, even if he's not sharing the exact same scene with them. So you might think, well, if you just had a movie just full of brilliant looking Tarkins, maybe it would

be okay. And maybe it would But some of the classic examples of Uncanny Valley happened to be films that are filled with nothing but c g I characters. Yeah, how about one from the same year as The mom of Your Turns two thousand one, if you go back to Oh yes, Final Fantasy, The Spirits Within, I remember kind of liking it. I do too. It was a film that I think I kind of half watched, half

worked on, like some just college course workers something. It was just on in the background and and maybe that was the right level of immersion in it. But I remember digging it. But at the same time, there are a lot of dead puppet eyes in this movie. Oh yeah, and it's so I saw it at the time. I remember having mixed feelings about the animation, like in some senses, I remember thinking, Wow, that looks so cool. That again

may have been a product of its time. We can talk about that more, how our expectations changes things go on. But also I don't know. There were multiple things wrong with that movie, one of which being that the last line of spoken dialogue in the movie, as a friend of mine pointed out at the time, was oh it's warm. Well I can I don't remember the line, so I can't speak to it how well it landed, But I

can see that being a problem. Role credits Now. Another big one, this came out just three years later, is of course, The Polar Express. Now, when people talk about the Uncanny Valley these days, i'd say this is a top three mentioned. Yeah, this is one of the defining nightmare of our time now based on a wonderful children's book about the magic of Christmas time. Yeah, the book

is wonderful, but it's certainly one of these examples. If you take a very brief children's book and you try and adapt it into a feature length motion picture, that's very difficult to do. In fact, I'm really grasping for an example where anybody actually pulled it off. Like, the best adaptations of children's books that come to mind are

all very short, very short films. Generally, I'm thinking of Dr Seus's adaptations from the seventies and eighties, not The Polar Express, which is just an exercise in psychic trauma brought on by just seemingly intentionally weaponized Uncanny Valley. Um, you know, the soulless puppet people. I've never seen this movie, but I looked up clips to see what people were talking about, and oh man, they they are not kidding it. I don't know how children made it through this movie.

It has these it has these creepy elves, It's got a creepy Tom Hanks as a train conductor. Nothing seems right, Everything seems like it's just about to everybody's about to start melting and screaming. Yeah. I think this is one where it was a poor idea in my opinion, and uh in the technology was not there to to rescue the idea. Now the next one we're going to discuss though. I think it was a great idea on paper, but it just didn't work out on the screen. And that's

of course. Two thousand seven's Bayowolf. Now as this Robert Zemeckis who did this, Yeah, Robert Zemeckis helmed it. And then the writing it was Neil Gaiman and Roger Avery. So some you know, some some some big names just attached and to the the ideas behind this. Uh this movie, and of course based on the story of Grindel and Beowulf, which is a classic you would think, you know, hard to miss action narrative. I think The Bao Wolf could make a really great movie if somebody did it right. Yeah,

I think so too. I have yet to see that movie. But but but certainly has all the potential in the world. And they had a pretty cool vocal cast as well, I think, and Angelina Jolie is in it as the monster's mother. They have ray Winstone as Bowel. Yeah, he does the voice of of Beowulf, and uh and who was it that plays the monster? We were just Crispin Now I've seen it all back home. Not one of my favorite monster depictions of Grenville, by the way, but

he's a monster. We can get past that. But everybody else in the film really has the uncanny valleys that going on to to a high degree. I think I read a quote somewhere where film critic was talking about how the monsters in the movie were only slightly less frightening than the humans. Yeah, yeah, the humans. It just

it just didn't land. Now at this point you're probably thinking, well, how about video games, because they're certainly when you're thinking about computer animated human beings interacting with each other, staring right into the camera, you think of video games. Yeah, and and I think, you know, here's here's the thing here. I have to say that I haven't noticed it as often these days. I think a lot of game animators

have found ways to get around the Uncanny Valley. I don't want to get to ahead of our flow here, but I think one thing that I've noticed they sometimes do is that they don't actually go for photo realism, and they go for a kind of more real than real combination of like a comic book style type character illustration, and then these other realistic aspects that when you when you look at a video game character, you would never mistake it for a photograph of a person, even even

one that's got really good graphics. But much like the way dialogue is written in films, you know, you don't want to make dialogue sound like real people talk, because that would be horrible to listen to, but you do want to make it sound quote realistic. You don't want to make your characters look too realistic in animation, but you do want to make them look quote realistic. In

other words, they feel real. Yeah, this reminds me of a game franchise that I haven't I don't think I've ever played more than a demo of this, but the Gears of War series. So all the people in this kind of look like like, if you're gonna be critical of it, you might say, well, everyone looks kind of like they're weird guerilla people. Like. It was a like we're in an alternate world where unrealistically huge upper bodies. Yeah, is if evolution took a slightly different turn into an

intelligent primates. Uh. And yet they look real. They don't look like they don't get an uncanny effect rolling off them, Like you know, you look at them, you can see pores,

you can see hair follicills. They look real, but they are but they are certainly not going for authentic human being there all right, Now, I want to put out one more example here before we move on, and it's a rare example of uncanny valley avoidance, a very specific type of uncanny valley avoidance, and that is from a fantastic stop motion short that was produced by the National Film Board of Canada. And you can find the online

if you just do us a search for it. It's Madam Tutley Putly and it's a wonderful little little film, very very French feel to it. Characters on a train, weirds, frightening things occurring. Uh, definitely check it out. But the trick to it there, these are stop motion animated characters and their eyes just feel so alive. They stare right into you, and you don't you don't question for a

second that these are that these are people. And the trick that they employed is that they used real human eyes, not in a you know, depraved, evil puppet master kind of a way, either. They videotaped the eyes of human actors and then blended the footage with that of the puppets. That sounds like an incredible gambit, because that sounds like that could have produced some of the worst Uncanny Valley

feelings ever if it went wrong. Yeah, and and I don't know, there may be some people who watch this short and and have the opposite effect and think that it's super creepy. I found it to be like this, this interesting example of circumventing the Uncanny Alley. But I'll leave it for you guys to decide. I'll include a link to this one as well as some of the other sources we're talking about on the landing page for this episode. It's stuff to blow your mind dot com.

All right, well, we are going to take a quick break and when we come back, we will get into the origin of the scientific idea of the Uncanny Valley and its history. And research. All right, we're back. So the uncanny Valley. Where does this even come from? Right? So we've already been talking about it because most people have heard of this, they're somewhat familiar with it. I

was talking to Rachel about it though. She was saying, you know, at least to her, it had this connotation of just generally synthetically generated images being creepy in one way or another. So maybe we should get into the specifics of the origin of the idea. So let's go back to the year nineteen seventy. Everything's great, Wait is it? I don't know, but but everybody, everybody's looking forward to the future in terms of creating humanoid robots. What are

we going to be able to do well? The Japanese roboticist massa Hiro Mori of the Tokyo Institute of Technology. He wrote a paper that was published in this Japanese journal Energy that coined the term uncanny Valley to describe a problem that he was predicting with increasingly humanoid robots. And this was based on just some observations he'd had of of different events. So you might say incidents in the progress of designing humanoid robots such as consumer electronics

shows in Japan and stuff like that. So what he predicted was that as you had a humanoid robot, robot that looks like a human and it's likeness to a human increased, our attitude toward them would improve. Our affinity would go up as they became more human, until they reached a certain tipping point of similarity to humans, where suddenly our affinity, our friendly attitude, almost immediately shifts and

plunges down into strong revulsion. Being human is likable, being sort of human is likable, but being almost human is horrible and repulsive. And then of course at the final end, uh you you would have a real human. So you can think of the uncanny valley as a phase in a graph, an X Y graph, and along the horizontal axis on the bottom, you've got the degree of similarity to a human, and then on the vertical axis you've got the degree of our affinity for the object. And

more hypothesized, this graph would have these two peaks. You'd start with zero on both axes, because a thing that has no human like traits basically gets no human affinity response one way or another. And we just don't you know, how much do you really like an industrial conveyor belt that you're just sort of neutral on it. But as you increase the humanity, you give a robot arms or something that looks like a face, eyes, limbs, you climb

this gentle, gradual slope to the fur peak and affinity. Um, you know, and he didn't name the peak, but I think we should name the peak. I think this first peak should be called something like the cuteness peak. That's not exactly right because it's not exactly cuteness, but it's recognizing something kind of human about what you're looking at. Yeah, Like, I mean, we don't have to describe cute to everyone here, but certainly this is hello kitty territory, this is the

this is the domain of large eye. It's vaguely infant or kitten like creatures that would never be mistaken for human or real, but they resonate with us for a number of reasons. We could do a whole podcast. In fact, we have an old podcast episode about the science of cute. Why that connects with us? Yeah, so they would include that would include all kinds of robots that are just kind of have general, very basic faces that don't try to have human skin or anything like that that just

might have like kind of a mouth and some cartoonish eyes. Yeah. Sure, there you go. That the C three po boldly on the cuteness peap. But at a certain point after this first peak, this graph drops off steeply. So you keep going along the x axis, but then the y axis drops off, not just down to zero, but far below zero, into the negative affinity range. And this part of the graph is the uncanny valley. As the similarity to a

real human continues to increase near a dent. In other words, as it becomes indistinguishable from a real human, our affinity sharply shoots back up the second peak toward reality. So I'd call this second peak the reality peak. It's when you become, for all intents and purposes, a real human being. Yeah.

I would also say that if if robots were candy, the bottom of the uncanny valley would be banana flavored candy like that for me has always been a flavor where it's like clearly like not only like runts that have bananas, I think, so like like great candy, like grape candy doesn't really taste like grapes, but it's if it's enough from the Uncanny Valley of Candy the year You're okay, whoa, You're right, banana candy actually does taste

like bananas in a way that makes it not really good. Yeah, Like I've candy fans, I don't eat that much candy anymore. So maybe the technology has advanced, but uh, my memory of the banana candy is is that of an uncanny experience. Now, one thing we should note is that so this original paper was published in nineteen seventy twelve. English translation was published in uh the I Triple A Robotics and Automation magazine. And that's what I was using, is my reference, that

English translation from twelve. Uh. And so it has some graphs here, It has Mori's original graphs or interpretations of them, and we can get into a little more detail on the nuances of Morey's theory. But one thing I did read was that many years later somebody contacted more and he and they were talking to him about this idea he had of the uncanny valley, and they were like, well, does does anything lie beyond the peak of reality? And he said, hey, oh yeah, actually there is such a thing.

And he said, you know, beyond the real human, you'd have sort of like artistic ideals. Oh wow, like the realm of forms even right, Yeah, so well, I think he used an example of like a statue of Buddha, you know, a beautiful, perfect statue of Buddha. It's almost like it we have greater affinity for it than we have for a realistic human because we've been, well, to

a large point, we've been conditioned to rite. Yeah, that that kind of gets into this, this idea of conditioned familiarity that we not only have with religious icons, but also with pop culture icons. So not only the Buddha, but also Robbie the robot, or or even the Terminator or well, yeah, that does make me think that in some ways, if if aesthetic ideals and things that were familiar with through our culture might be even beyond humans. I mean, again, this is not like rigorous research, This

is just what more he says, he inks and predicts. Uh, could could there be like a robot that we really love that's actually better than a than a normal human? Well, you know, there's a study that came out last year, I believe from Penn State University that was kind of

interesting alone in these lines. So The researchers survey three seven nine adults ages ages sixty to eighty six, and they asked them for specific memories of robot films they'd seen in their general attitudes towards robots and and the you know the age here. As you might imagine, they're really looking at at potential care robots, like the idea of like, what kind of robots should help you use

the bathroom? Do you want something that looks like kind of like a person, or do you want something that looks like a forklift with a forklift mated with an easy chair. But if you look at the ages used here and twenty sixteen, when the studies took place, you can say that these people grew up with science fiction. Oh yeah, I mean they might not have personally consumed a lot of it, but it's in the culture, right, Yeah.

They they definitely had access to it. And researchers found that individuals who could recall more cinematic robot portrayals were increasingly likely to hold positive attitudes towards robots in general. So it didn't matter if they remembered murderous kill bots or well meaning helper bots. Uh. They the mere memory of multiple robotic portrayals correlated to pro robot vibes, so

the study findings. They also backed up the importance of human looking human esque robots to invoke a sympathetic user response. But the researchers stressed that robot designers might want to incorporate robotic features that older adults will remember from their cinematic past. So it's saying that, like, don't just try to make it like a human. Try to make it like the robots we have known and loved. Yeah, Like

make it fun. You know, if I'm if I need a robot to help me go to the bathroom, make it make make them the robots from Silent Running, you know, Hulie Do and Louie, the little little guys. Then at least I can engage my nostalgia a little bit totally. So I want to look at a few more nuances. Maury's original paper in nineteen seventy so one thing, I do think it's very interesting and I want to come

back to as we explore this topic more. More hypothesizes in the original paper that our perception of an uncanny valley might depend on the context in which we're we're viewing the being and the example he gives here is he's talking about Buon Rocku puppets, and so he says, quote, I don't think that on close inspection of bun Rocku puppet appears similar to a human being. But when we enjoy a puppet show in the theater, were seated at a certain distance from the stage, the puppets absolute size

is ignored. Its total appearance, including hand and eye movements, is close to that of a human being. So, given our tendency as an audience to become absorbed in this form of art, we might feel a high level of affinity for the puppet. I think that's interesting. So it's it's not just the object, but it's also the context in which we experience the object. You might have very different feelings about a and rocky puppet lying on the floor versus one that you go to see in the

context of staging a play. Yeah. I think that the puppet argument is something to keep in mind throughout considerations the Uncanny Valley, because there are a lot of people that there are a lot of people who have kind of um an irrational version to puppets in general, and certainly if you take just a still puppet and you hold it up. There are various puppets that one might find a little bit uncanny or creepy, etcetera. But in the process of performing with a talented performer is going

to bring that to life. Like that's the art form. And and there's so many different varieties of puppetry. Certainly they're they're too broad. Categories are its situations where the puppeteer is visible and puppet situations where puppeteers not. You know, so you have your basic the muppet situation where you don't see the puppeteers, but there are plenty of art forms of puppetry performance styles in which the puppeteers very visible,

either completely or just their face. You see their eyes, You see that there's a person involved here, and uh, and there's not this this mystery or this sense of deception, right, yeah,

I think conceptual clues like that are very important. Also is when you consider the the idea of going to a puppet theater, it also includes a certain attitude charging effect in the audience, Like an an audience member goes to a puppet theater prepared to suspend their disbelief, like you know what I mean, Like you put yourself in an intentional state of open mindedness about what you're viewing, and you give yourself an interpretive framework through which to Like,

if you were not prepared to watch a puppet theater story and suddenly a puppet was just moving around, that might be a lot creepier. Yeah, I agree. So part of the Uncanny Valley effect is probably also in the viewer themselves and in the so the context is not just where you are, what's going on with what you're looking at, but what you're expecting to see. Now, one more thing that Morey points out is he thinks that they're going to be very different rules governing the Uncanny

Valley for still objects versus moving objects. And essentially his hypothesis is that movement is going to amplify both the peaks and the valleys of the graph. So if you imagine the graph we said earlier, gentle slope up to first peak, you know, kind of cute whatever has some human characteristics, then a dip down into too close to human but not there, and then a final rise up to actually human. He he would say, if it's moving, the peaks are going to be higher and the valleys

are going to be lower. So a thing that is moving gets greater affinity if it's good if it's at one of these two peaks, but it's even more revolting and unpleasant if it's at the valley. Uh So, this this makes me think of Samara in The Ring, those scenes where Samara is emerging from the TV or the well, her movement is is jerky and and I understand that they created that effect by having the actor or actress

walk backwards and then reversing the footage. So you have this, you have this this movement that is, you know, natural, but being reversed it it feels very unnatural and it's hard to really pinpoint what's not working for you about it. Right. So More in the end concludes he gives this recommendation based on his hypothesis. He says, don't go for realism, right, It's gonna be so hard if you're designing a humanoid robot. Now, a lot of what we're we're talking about in these

episodes is animation. He's talking primarily about humanoid robots. But typically these uh two fields get somewhat conflated in discussion of the Uncanny Valley because in both cases you're trying to create something that looks pleasingly human. Um. He wrote, it's gonna be so hard to get out of the valley up the second peak, that that's the reality peak is so steep. Instead, roboticists should not try, and instead they should aim for the very tip of the first peak.

Stick on the cuteness peak, because we know we can get there. Think think Wally or other cute humanoid robots. The first peak is not really that hard to attain. People respond well to it, So why do you need to try to go past it? Um? You know. As for animated humans, I think a good analogy might be here's one Pixars The Incredibles versus Final Fantasy the Spirits, within which we already mentioned. The former they don't look like real humans at all, right, they're cute, cartoonish, non

realistic humans, but they're quite pleasant. The latter goes for and fails at photo realism and creates these characters that are stiff and unsettling. In other words, he says, don't try to climb out of the valley, just don't go into the valley to begin with. Yeah, this this really brings to mind just the idea of like filmmakers and creators standing on the edge of this physical valley and there's a local guide. They're saying, don't do it. Don't

do it, the value will consume you. And they're like, no, you're Lucasfilm. We can do it. Yeah, we got all this high text gear. There's no way that anything's going to take us down there. And then they go down there and it's just Jurassic Park or Congo with they just get torn apart. You know, I do. I do think the dinosaurs in Jurassic Park come out of the

uncanny valley for dinosaurs. They do. Yeah, And that introduces an interesting wrinkle in that, like Maury is talking about human wide qualities, it would probably be a related but different thing to just say animal reality versus specifically human reality. Yeah, because I mean when for non human creatures, certainly we've been able to nail that. For ages, to stop motion creatures, often, even if their movements are kind of herky jerky, they

feel great. Um that heard of stop motion robots in older films, so that I've never had a problem buying into them. And yeah, you're look into the eye of the track or the t rex or the velociraptor and you never doubt for a second. Yeah. But so one thing I think we should point out is that as prescient as Mori was of what would become this widely recognized pop culture phenomenon his paper, it's not it's not research really, it's just sort of observation and interesting speculation.

So what we should shift now to do, I think, is talk about whether there's really any evidence that the Uncanny Valley Number one exists at all? Is it really a thing? Number two? Is it a is it a unified phenomenon, or is it there there's some separate things getting pulled into the net together here. And then finally maybe we should look at if it's real, what causes it? Why do human brains tend to react this way? So maybe we should take a quick break, and then when

we come back we will get into more recent research. Thank. Okay, so there's there's really no denying that there is some kind of creepy humanoid synthetic figure effect. Right. We we've all seen these c g I movies. We've all seen these creepy robots and had that feeling don't like it. But that doesn't necessarily mean that the uncanny Valley, as described by more or as popularly conceived in culture, is

in fact a correct description of what's happening there. Right, just because it it feels truthy, just because it lines up with to a certain degree with how we feel about the world, doesn't mean that it is. You know that it is an actual effect that's taking place, and or that it's even a fixed effect, et cetera. There are a lot of factors to contemplate here, Like for my own part, I've always found it interesting and I

definitely think there's something to it. However, you line it up with similar cases in life, such as say, individuals that you may encounter who have some degree of facial disfigurement, and it might be extremely mild, it might be it might be nothing more than a uh, and then you know a lazy eye, or or you know some sort of cleft lip or cleft palate scenario, or it just might be like their faces maybe not all that symmetrical, right,

and you know nobody whose face is perfectly symmetrical. But with all of these individuals, you interact with them, you get to know them maybe, and whatever kind of like initial um reaction is present, be it just kind of a huh or that goes away, and you can unless you're a total jerk, Unless you're a total jerk, or you're gonna be able to relate to that person. You're gonna be able to communicate with that person, and you're not going to be thrown for a curve every time

they make eye contact with you. Yeah, I would agree with that. So there is certainly, like in Maury's original formulation, he he would, I think, put different kinds of um physical abnormality somewhere on the ascending slope, on the on the on the uncanny valley slope. So you have a normal, healthy human up at the peak, I guess somewhere below the artistic ideal of the great Buddhist statue or something.

But you'd have normal, healthy human. Then somewhere below them would be people who have who look like there is something wrong with them in terms of having uh, you know, perfect health and symmetricality. I mean, certainly because just an ill person. You encounter someone who is clearly a little bit sick or a little bit hungover or whatever you can tell, and it causes a light to go off

in your head. Yeah. Uh, And and yet we can quite easily adapt to people like, you know, you see somebody like that, it is you just know it is not proper to react to somebody with revulsion. Oh yeah, like right now in Atlanta as we're recording, this pollen is everywhere. So there are several people in my life. I'm not really affected by the pollen so much, but it totally debilitates some of my coworkers, some of my

friends and red face puffy eyes. Yeah, and and sometimes they're like walked out on allergy medication to boot and you just get you just you know, you accept it. You realize, oh, well, you know, my my friend here is going to be kind of a pollen zombie for a couple of weeks. But that doesn't mean we can't hang out. It doesn't mean we can't work on this or that. Yeah, so definitely, The Uncanny Valley has plenty of critics, and plenty I think a very fair criticisms

leveled at it. I just want to go back to one popular article. I came across a two thousand ten article in Popular Mechanics by Eric Softge where he sort of points out that at the time people were as I think they are still now treating the Uncanny Valley as a proven fact, but in fact, at the time, he says, you know, there's really almost no convincing evidence that such a thing even exists, and he speaks to an expert named Carl McDorman, director of the Android Science

Center at Indiana University, and McDorman, who has conducted research on the valley, offered his opinion in the article, saying, quote, it turns out that there may be more than one Uncanny Valley. It's not the overall degree of human likeness that makes a robot or animated care acter uncanny. It's more a matter of mismatch. If you have an extremely realistic skin texture but at the same time cartoonish eyes or realistic eyes and an unrealistic skin texture, that's very uncanny,

uh and the art. So that's an idea about the perceptual mismatch that I do want to revisit later in this episode. But the article also speaks to a guy named David Hanson who's a roboticist who specifically specializes in creating very realistic humanoid robots. I think he did that that Einstein head thing. Oh yeah, nothing. So Hansen claims that even if people find overly realistic robots creepy at first, they get used to them within minutes. This is sort

of what you were just talking about. I think, you know, you become acclimated even to something that you might uh, at some kind of base level, have a negative reaction to. Yeah, I keep thinking of having an isolation in this because it's the game I'm currently playing, and uh and I feel like that the c G characters are are a pretty well done in there. I haven't felt that the tinge of of Uncanny Valley washing over me. Some of

the voice actings a little weak. But but but speaking of the voice, like the the the androids you encounter though with the sex and uh androids that key. Yeah, and when I first on Canny Valley, well, yeah, but when I first encountered them, yeah, they had the uncanny intentionally kind of creepy appearance and the very creepy robot voice. But yet when they were not actively attacking me, I kind of was. I was kind of cool with it.

It wasn't until he started becoming violent that that that the mere sound of their voice or the appearance of one uh down there, you know, in the distance down the Hallway would would cause my nerves to react. I mean, those things are funny, they're uh, they're a good part

of that game. But anyway, So in this UH article, the author also cites some other unnamed robots roboticists, as well as his own experience when he's talking about meeting robots that he had previously seen on video, and one thing he says is, you know, an Uncanny Valley effect that was present when I saw a video of this robot went away when I saw it in person. I don't know if that's generally true of people. He claims

it's true. But even if this is truly the case for robots, I'm not sure how it would apply to animations. Probably wouldn't apply to animations. Um, But I think that there are some good threads to start tugging at here, because it's probably the case that there are more dimensions to the Uncanny Valley than more he imagined in nineteen seventy, meaning more than just that X axis of um closeness

to realistic human appearance versus distance from realistic human appearance. Yeah, I mean, just what makes a person human, what makes a lightness human? There's arguably a whole chorus of things going on there. Yeah, so it would make sense that that that chorus would play into the Uncanny Valley. Yeah. So I do think that there are multiple other dimensions to be explored, But I also don't think that means we can conclude that there's nothing to the Uncanny Valley.

And in the past decade there's actually been an explosion of research on the Uncanny Valley. So I think we should look at a few interesting studies on the effect. All right, Well, first one here that I came across was a two thousand nine Princeton University study and they looked into the effects of uncanny value of the Uncanny

Valley on maccaque monkeys, so so non human subjects. Yeah, because that that makes sense, right if Yeah, if you want to see if this is an evolved response, let's look beyond the complications of human intelligence and human culture and looked something closely related to us. Is it biological rather than say cultural? Right, And so they showed a selection of the primates close to real quote unquote computer visuals of macaques to see if they responded with coups

and lip smacking as they do with their fellow monkeys. Uh, and these these close to real computer visuals were essentially lawnmower man monkeys. If you see lawnmower man. They kind of asking if I've seen lawnmower man, Robert, you know I've seen lawnmower man. Yes, so yeah, I think lawn more man. Uh, and you kind of have an idea that that level of computer animation, and the monkeys did not want any part of it. They averted their eyes,

they acted frightened when confronted with lawnmower man monkey. So it's not much, I admit, but it's a little experimental evidence for the argument that Uncanny Valley is an evolutionary response. Right, So if you can observe it in monkeys, there's probably some element of it that that is biological in the brain. It's instinctual and and not just something we've all learned

to say about weirdly looking animated characters and robots. Yeah, and that would be maybe a weak piece of evidence, but still a piece of evidence you could put in the column of saying there is something there. The valley does to some extent exist. Right now, the next study that I ran across this comes back this to one of the graphics that you pulled out of the believe the original h study correct, Yeah, yeah, the original Morey's

original graphs. So in this graph and we talked about diving down into the valley and then steadily trying to claw yourself out on the other side, very very steep ascent. Yeah, so you hit bottom and that's where you have a zombie. And as you begin to scale out of the uncanny valley, he has um uh myo electric hand and prosthetic hand down there. As you climb back up, eventually hitting ordinary doll and puppets and ill person and maybe hitting healthy

person at the very top. Again. But it's interesting you have prosthetic hand down there, because this next study looks at prosthetic and robotic and human hands. Yeah, this is in the original study. More He talks about the variable creepiness of prosthetic hands. And I found I found this interesting because I don't know about you, but but growing up, I felt like crazy robot hands, especially we're everywhere like every G. I. Joe show or he Man type franchise,

there's always somebody. It could be a villain, it could be a hero. But there were crazy robot hands galore. Uh, And I always found them cool, and I feel like a lot of us probably even fetishized them to a certain point, Like we we didn't understand what it would necessarily be like to lose and lose a hand and the shortfall and the ability of technology at the time and even today to replace that missing limb, but we thought, well, that looks cool. Superpowered robot hands signed me up. Right.

But back to the study two thousand thirteen University of Manchester study, and they looked at prosthetic hands. UH. They used of forty three right handed participants, thirty six female and seven male, and they were all looking at photos, and the photos were divided into three categories human hands, robotic hands like no question about it, that's a robot hand I'm looking at like straight up terminator exoskeleton or

or or even less human, and then prosthetic hands. The results, I have to say, reading through some some of the writing about this UH and the original press release, the results were kind of confusing sounding. They the subjects here preferred human hands and robot hands, but but rated and certainly rated prosthetic hands is more uncanny, But prosthetics that looked more human were less eerie. Okay, so, so something that's clearly a robot that's not too creepy. Something's clearly

a human that's not too creepy. If something is a robot trying to be human, that might be more creepy, but as it gets better at being human, it's less creepy,

I think. So, I think that's my take. I mean, it also makes me wonder if if the hand alone is an is like a subset of the uncanny Valley, because certainly if you're if you're just working with a hand and trying to replicate the movements, the look, the feel of a human limb for an observer, not we're not going to even get into the the you know, the problems of creating something that the user can experience as a life like limb. But if you're just looking at it, if you don't have to worry about its

eye contact, you don't have to worry about micro expressions. Uh, it seems like it would be an easier peak to surmount. Yeah, so that if that is in fact the correct interpretation, that would seem to undercut the steepness in Maury's original graph right on the on the final peak. Yeah, that's I mean, that's what I'm wondering, because the hand had taken in isolation, is if you thinking to be easier

to replicate? Yeah, uh and uncanny Valley. Let's face it, when we talk about it, most of the time we're talking about faces. Right now, Speaking of faces, there's another study. UM. This is a two thousand and eleven University of California, San Diego study. UM. This is published in the Social Cognitive and Effective Neuroscience. And they did exactly what you'd expect researchers to do when confronted with the Uncanny Valley.

Grab the f m r I and see what our brains are doing when we're looking at all these images. So all these fMRI I studies, all right, well, what what did they find? All right, I'll roll through the basics of the study here. So twenty subjects, not a not a huge study here, aged twenty to thirty six. And here were some of the caveats they had in

selecting these individuals. No experience working with robots, no time spent in Japan, no friends or family from Japan because they wanted to avoid uh, any you know, potential cultural exposure that would have made them would make them more accepting of androids. Okay, so the idea is that maybe in Japan people just experience humanoid robots way too much. Already,

they're too they're acclimatized to them. Yeah, that that's the the argument they made, and laying out the study, let's let's not even go there, let's just deal with people who have less exposure to robots. And they were shown twelve videos of a humanoid robot named repley Q two. Oh man, I'm looking it up right now. It's it's it's rough. But well, they watch video twelve videos of this robot doing there is things, and they were shown

videos of humans doing the same things. And in fact, the robots movements and mannerisms were patterned directly after the humans. So you had a you had a human version of the actions, you had an android version of the actions, uh, you know, a lifelike robot, and then you had a a stripped down version of the androids. So basically the android of all its skin ripped off, so it looks more like a robot, clearly a robot, and it's doing

the same motions as well. So this broke it all down to a human with biological appearance in movement, a robot with mechanical appearance and mechanical motion, and a human seeming agent with the exact same mechanical movements as the robot.

Then in came the f M R I scans. So the main brain area of note here, the the area that that that that lit up where we saw the most activity, the parietal cortex on both sides of the brain, specifically in the areas that connect the part of the brain's visual cortex that process bodily movements with the section

of the motor cortex thought to contain mirror neurons. Okay, so those would be like the empathy parts of the brain where you know, we we see something going on in some other creature like us, and we empathize with it exactly. Yeah, So when viewing the human looking android, the brain lit up at the recognition of a human form, but registered essentially a computing error over the movement. Something

didn't match up. Uh so it's it's not. According to this study, it would seem that it's not the biological movement or the biological appearance, it's the congruents or lack of congruents between the two. You look alive, but you're dead, you look dead, but you move you or you speak as if you're alive. Um. So, the researchers noted that this is something that could be retuned through exposure, but it could be at the heart of what's going on

with the Uncanny Valley. Interesting. Well, I think we should look at one more study potentially providing recent support for the existence of the Uncanny Valley, and then maybe after that we should break and then come back next time to get into the causes, what what would be causing this effect and uh in the future. So I want to look at a study that came out in two sixteen in the journal Cognition by Mather and Ricling called

Navigating a Social World with Robot Partners. A Quantitative cartography of the Uncanny Valley. Cute invocation of map making there because it does kind of make sense. I like the idea of mapping the valley because that indicates that it may expand beyond just the one dimensional dip and is in fact more of a topographical space, you know, like

we can extend into three dimensions. But anyway, so to get into the study, the author's note that while the Uncanny Valley has very strong intuitive support, people tend to take it as fact. Experimental evidence for it has been limited and inconsistent. As as we mentioned earlier, some studies seem to find evidence for the valley. Others don't you know, they say this, this isn't necessarily a thing. So there are multiple experiments here. First, they did a thing that

I think was pretty smart. If they were trying to chart a linear progression of the up and down peaks and valleys, they tried to generate an objectively determined gradient of more and less human looking robots. So what a lot of these studies do is maybe along the macaques study ideas, they show you a lawnmower man, they show you a real person, they show you a robot, uh, and they ask you to characterize you know, how do

you feel about these? What they did here is that they gathered a very large sample or relatively large sample of eight images quote from the wild, meaning from the internet. So these wild type robots samples, and they had a bunch of inclusion and exclusion criteria. I don't want to get into all of them, but they tried to limit it to where it would it would kind of throw

out all these variables. They could complicate things like they tried to keep just certain types of pictures of faces of real robots that are built and uh, and they had some exclusion criteria like it couldn't be a well known character, a famous person, um, it couldn't have objects overlapping the face, It couldn't be a toy, it had

to be a real humanoid robot. And then they had subjects rate these images on what they call the mechano humanoid scale, basically to come up with an objectively derived score for each image by using this this empirical research, by going to a bunch of people and saying, hey, how mechanical is this? How human is this? And then after they had a rating for each of these eight images, and Robert have included an image, uh, I think down here to show you, like what all these robots where

you can kind of see. It starts with things that look not human at all, just like a lump of wires and junk, and then it proceeds up to something that looks like a picture of a guy. Yes, yeah, very much. So. You start off with very kind of wally asque heads. When you move in through like like skinless gremlins, and then through the sort of the the expected hierarchy of humanoid robots. Okay, so they've got this thing, and then they rate all these images and sort them

into an ascending scale of humanness. And then they took ratings in multiple different ways of likability and trustworthiness. Now, in likability, they claimed to find a robust uncanny valley effect, where likability increased linearly with humanoid qualities up to a certain point, and then it took a negative dip as the humanoid qualities continued to increase past that point, and then once again began to rise at the far end

of the scale. Now, one thing I want to say, just looking at the results is it does not appear that people were the most bothered by the things that were the most human looking. Like given my understanding of the uncanny valley, I would have expected the stuff at the very top end the scale to be the most disturbing. But they actually kind of liked the stuff at the very top end of the scale. It was somewhere closer to the upper half middle of the scale that they

really didn't like. Um, So, to whatever extent there is a real uncanny valley, it might not lie so close to the quote realism into the spectrum as we think.

They also performed some trust experiments by creating a scenario where subjects would be asked to trust these robots to invest money for them, and the results there were basically they claimed that the trust uh experiments did show some Uncanny Valley effects, but the results were a little more complicated than on the straightforward superficial likability scale, the likability

really did look like Uncanny Valley was being displayed. They also performed experiments with a more traditional quote controlled series of composed face images, so it would just be a series of basically the same face as a robot than a little bit more human, little it more human, little

bit more human on this gradient of human nous. And they generally claim to find that there was evidence for the Uncanny Valley effect in both likability and trust with both the wild caught robot image samples and with these composed face images that they came up with, But as always, more studies are needed. But that looks like there is one study showing pretty solid evidence that there is something

like an Uncanny Valley effect. Yeah, and I like the idea that that that that it's it's an uncanny valley. But maybe it's just a more more nuanced from a topographical standpoint. You know, they're they're more a little little bumps and little valleys within the overall valley, little caves you can crawl into and just yourself inside, and maybe even caves that turn into tunnels that emerge on the

other side. Yeah, that that's an interesting thing. I mean, like they point out that there's a lot of variability in their data. Actually, like it wasn't um If you look at their their plot chart of where all the data points fall and then they plot a line going through it, If you plot a line going through all their data, it does show the uncanny valley effect. But you know, there there are outliers all over the place, like there is some there are some robots that are

just consistently more like more than the other ones. I find it interestingly that the some of the higher rated ones, or at least I think what number seventy nine in particular, kind of looks like a generic human as opposed to say, go down to seventy four that looks like a very specific human, Like if I had to pick him or pick the human he's patterned after assumingly out of a police lineup. I feel like I'd be able to do it,

but also seventy four looks angry. I'm sorry, folks, you can't see what we're talking about, but it's frowning at you, kind of like should I kill all humans or just shrug it off? And maybe two day's the day that does introduce There are a lot of complicating factories here, and the authors icknoled this, like these images don't all have necessarily the same emotional affect, like some of them

seem happy, some seem unhappy. There's enough variability across the board that you can think you're getting a reasonably decent answer when you plot reactions across all samples. But yeah, there's definitely a lot of different stuff going on here in addition to just being more or less human. I like how thirty four on our on our chart here it seems to rely heavily on animated mustache and eyebrows. Oh yeah, what is that? It looks like a It looks like a very mustache. I can't add to what

you've just said. It's got a white mustache and brow and beard, and it's saying, oh boy, it looks like a lot of these incomplete puppets are stripped awaite puppets, you see where they're like, all right, we got a lot of work to do on this thing, but at least we got the eyebrows and a mustache in place. But see, I find that one very likable. It doesn't look very human at all, but it's very I want

to play with it. Yeah, okay, Robert, Well, we've got a bunch of more stuff to talk about, but think we should call it there and come back and finish our discussion of the Uncanny Valley next time. Yeah, we'll get into we'll go be on the Uncanny Valley. Yeah, so we'll we'll talk about what might cause the Uncanny Valley effect to whatever extent it does exist, and we can talk about you know, what happens when you ascend

that that far slow? All right? Well, hey, in the meantime, head on over to stuff to Blow your Mind dot com. That is where you will find all the podcast episodes.

You'll find videos, blog posts, as well as links out to our various social media accounts, and the landing page for this episode should include some links to some of the resources we're talking about here today, and if you want to get in touch with us as always, with feedback on this episode or any other, or you just want to say hi, or you want to let us know an episode topic you'd like us to cover in the future. You can email us at blow the Mind at how stuff works dot com for more on this

and thousands of other topics. Is it how stuff works dot com. They can't even big, Big,