The Uncanny Valley Phenomenon: The Impact of AI on Human Behavior

Welcome to beyond the Screen, an Ionis podcast where we share insights and tips to help you scale your business's online presence, hosting genuine conversations with the best in the web and IT industry, and exploring how the Ionis brand can help professionals and customers with their hosting and cloud issues. I'm your host, Joe Nash. Welcome to another episode of beyond the Screen and Ionas Podcast. Joining us today is Ketu Shah, a software engineer at Gusto in New York City. He's an expert on AI and technology, working for some big names in technology in recent years, including IAC and iis. He studied his Master's of Science at Stevens University of Technology, following four years studying his BE in computer engineering at Gujarat. When he's not at work, he's a translator at Translators Without Borders. Welcome to the show K2. How are you doing? I'm good, how are you? It's been a Monday after 15 Super bowl, so I did notice trains being empty. Yeah. A quiet day at work. Yeah, absolutely. Coming from a country that doesn't really do the super bowl, it's always a strange time because I feel like half of the world falls off its orbit revolves around this one thing. And I'm just in blissful ignorance over here. But I'm glad you made it into the office today to chat to us. You know, we mentioned a couple of aspects of your career thus far and your education in that intro. Could you walk us through your career so far and how your motivations evolved throughout that time? Sure. So grew up in India, so I do fall under that stereotype that how the Indian people are software engineers. But I studied computer engineering at just like any Tom, Dick and Harry University and Gujarat. But it did provide me really, really great segue way into like what technology can do. And coming from still a developing economy, you can see like so many issues persisting and how technology can help evolve, especially with E governance and whatnot. And that provided me more of a motivation of experiencing education and culture abroad. Because, you know, when you are in India you are sort of in a bubble sometimes. Like if you want to solve a problem, you need to look at it from the outside. And that's what I did. That's where I came here in U.S. expanding my horizon. It's been really, really humbling, resourceful experience so far. I have suffered like anytime I change a job, there are like first few months where I had the imposter syndrome. I don't know a thing. But yeah, eventually, like once you get into the zone and whatnot, once you get used to it things start to unravel easily. I would say, like, biggest motivation for me has always been solving problems. I don't like to say like I want to solve technological problems. I like to solve problems that touches humans. I want to solve people's problem, basically. And that has been my biggest motivation, like what sort of impact my work will have on something that I'm doing. I do lean towards a B2C company just because my work will actually have an impact on actual human beings. And that's something that I have seen, like for a lot of engineers usually the motivation has always been, I'm a tech savvy, I'm not that sort of an engineer. I wouldn't like to say that I'm a tech savvy engineer. I like to say that problem savvy engineer. I do take really cautionary approach to everything. I like to see like all things that can and approach things from there. That's awesome. Yeah, I mean, that makes a lot of sense. I think being motivated to solve people's problems is definitely a very fulfilling way to approach a career in the tech industry. So one of the things you said there, which I think is interesting given your introduction, was you said you're not a trendy engineer, not an engineer who's on trend. But you know, we're here to talk about AI, which is obviously the current trend at the moment. And you and I were talking a little bit before the show about this current moment and how we feel about it. And you mentioned being cautionary, I think was the word you used. Can you tell us a little bit about how you think about AI and how you're thinking about using it in your work? So artificial intelligence first of all has the word intelligence in it. And when we think about intelligence, we think about human brains. Now we really know so little about human brains. In recent years, we have made so many good strides, but that is just such a vast unknown out there. And we are trying to create an artificial brain that's going to solve the problem. So while we are trying to create something that's similar to something we just don't know enough about. And when you come from a technological background, if you don't understand the problem, you can't solve it. And we are trying to mimic the human brain. And when it comes to AI, we are trying to actually even bypass the capabilities like a normal human brain because it deals with so much data out there. The world has so many things out there, like it's gazillion trillions and whatnot bits of data that it's storing and processing. So when I think about AI, I like to think about, first of all, there might be biases. There are developers, so there are just so many biases, first of all in the content itself. Then AI has to go out in the lab, basically you have to train the models and it has to scrape the web, understand what's good, what's bad. And usually opinions are not binary, they are on the spectrum. And first of all, since I don't develop the models, I just don't know whether or not AI understands that things usually are on the spectrum. So those are like all the things. And then a lot of times I also think about social media and whatnot. And recently I've seen like so much trend in terms of I don't do TikToks, but I use Instagram and reels and stuff. So I've seen like so much things about AI generated content and whatnot. And I really like to think about what sort of impact it would have on creators. And then there is this human psyche and what sort of impact it will have on human brains or on our evolution. Because things have been advancing like so fast that I have lived in a time where when I had to solve a problem, I would go to Google and find Stack Overflow. I write up my problem and I'll go to Stack Overflow. And you know, usually it's Never like the first link solves your problem. It's usually like 10th link that solves your problem. Now this used to provide me so much more information for like future like, oh well, the other nine answers did not solve a problem, but they did provide me some sort of insight. Now in the future, if I do come across something, a problem, if I do like. And again, we don't understand memory and stuff, but somehow my brain did provide me that information. Oh, you came across this a while ago now with artificial intelligence and you know, more and more use of it. I don't think that developers who use AI, again, it's a really good thing. But as a developer, like if I'm using an AI just to like, I have this problem, I'm in a hurry, I have a deadline coming. Let me just start the chatgpt or whatnot, write a code for this. It will solve the problem. I will have to like make some adjustments because, you know, edge cases and whatnot. But I'm missing out on the opportunity of finding the imperfect solutions which in future might help me develop something that's, you know, the wider background knowledge. Right? Yeah. Exactly. A really interesting perspective. So I guess one of the things we're getting out there is that in some ways ChatGPT is a shortcut to the research we're used to doing as developers. I get very frustrated personally whenever I run into ChatGPT or chat boxes. Going through ChatGPT because it feels so much slower to me, like having to have a conversation with this thing when really I just want to like fire off. Like the subject of evolution here is really interesting because I think my brain is very much geared to like, I have been forged in the fires of knowing how to construct the shortest number of keywords I need to put into Google to get my answer right. And so now having to like go and talk to a thing to get it to give me the answer, it feels a lot slower to me. So it's interesting. You're correct, like the actual process of extracting the information from the search in the way we're used to doing it is a lot longer. But I'm so much more frustrated by the process of constructing the search in the chat. Yeah, I think it's a really interesting point. So you mentioned evolution there and you mentioned there is that difference. Do you see this like next generation of the Internet users growing up now learning the program with these things rather than how we did. How do you think it's going to change how they interact with the Internet? I feel like again, as humans we are getting smarter now. Up until recently that getting smart was by our own research, by our own interaction with the world and information and data. Now we are making it easier to access this information to solve these problems. And I just feel like people who are actually developing all these models, they will be really, really smart. And then there are everyday humans like me who are not actually developing these models, but who are consuming it. And given that things are so much easier for me to just extract or you know, like usually everyone is, I have this thing and I need to get it done and everyone is in a hurry so they will just ask that. And it's so much easier just to input something into ChatGPT and get the response that I don't have to do the research. And again, there are always exceptions. So I'm just talking about majority of the human population out there and given that things are so easy so they would not be able to develop that critical thinking aspect of whatnot and critical thinking deduction, discarding what's good, what's bad. Also like what's something that really, really bothers me is the over personalization of things out there where you're just not getting a competing perspective. So lack of perspective. So I'm just scared that we might have lost productivity that we would have gotten from them or some of the insights that we would have gotten from them. So it will limit the their evolution of the brain. But again, they might understand how to interact with technology even more and more. And again, all these AI models, they do make access to information real, really easy. So in a way that is still a really, really great potential. But then there are these downfalls and we just don't know it. There are just so many loans that. Yeah, it's just hard to fathom right now because the thing itself is evolving. Things are still remaining to see what impact does it have. Plus, it's a new field, we are still new to consume IT models are still being developed. But then there are this all sort of philosophical aspect and ethical aspect of AI that we still need to think a lot more about. And I think developers and companies and leaders who are driving all of this progress, they need to keep in mind what sort of impact this will have on human kind, basically. Right, that makes sense. One of the things you mentioned there was the presentation of these things and the personification, I guess so. One of the concepts that I guess you've attached to this field is the uncanny valley. Can you tell folks about what the uncanny valley is if they haven't heard of it? Yeah. So uncanny valley. My understanding of it is whenever an AI system, or not even system, but right nowadays, like we see a lot of digital avatars whenever an AI personification is so close to reality, where you know that that's not a real person, it's an AI. So there are certain imperfections, but you still know that they're not human. But it's so much close to actual representation of humans. Some things that AI can say, like sometimes they make you think, oh wow, am I really talking to the system or is it an actual person? So while you know that it's an AI, but it's so close resemblance to humans or human interaction that it makes you uneasy or it gives you like discomfort. That's what my understanding of uncanny values. Yeah, yeah. I think that is broadly what I've heard as well. I think one particular aspect that's really interesting of it is that it's specifically the feeling of unease when something is close to being human, but it's slightly off. And I've always seen this like conspiracy theory about that, which is like, you know, why did we evolve to Be freaked out by things that are almost human, but like not quite. Like what did that evolve in response to? Which has always been like one of those like horrific to think about notions that really entertains me. But as it comes to AI, most people will be familiar with Uncanny Valley from things like cgi, when a character looks almost perfect but it's just a little bit off. How does it apply to the type of AI systems that we're talking about today? Yeah, so nowadays like given that all the companies, they are trying to utilize the good aspects of the AI, again, it makes life a lot of times easier for our own clients. For example, like we work in an industry where we have like a certain knowledge base which, like FAQs and whatnot and you can just feed that all of that into the system and make it conversational. Otherwise a person would have to like Google it or search in our own company's like own website and stuff. And again, it's, it's just hard. So basically just creating a chatbot that has all of that information, all that knowledge base, you ask the chatbot a question and it will just give you an answer. So that's something that I've seen like has been going really, really well. It does make life a lot easier, except for the time where you know that this is a specific problem tailored to you and you will not get an answer. And that thing keeps on trying and trying and trying to. There is that aspect. Plus nowadays I have seen like a lot of times in the news articles like AI has gone sentinent and whatnot and it just makes people creeped out. But again, so in terms of like actual use cases, chatbots are something and digital avatars. And I think I have seen a lot of users in therapy as well. So again, I would be really cautious whenever it comes to like especially like health aspect of it. But those are some of the use cases that I have seen. I, I think they do help a lot, especially with productivity and solving the problem faster. Yeah, yeah, absolutely. Yeah. This avatars one is interesting. I feel like there was a period that's maybe quietened down a little bit, but a couple of years ago, you know, as everything was going to be the metaverse at the time, there was a lot of putting distro avatars into everything where there wasn't necessarily a clear use case. You mentioned that as being a area where there is utility. Where do you think usage of digital avatars in a AI driven or more realistic fashion than like the kind of things that we have traditionally used These things like profile pictures and whatever and where do you think those might be relevant? Nowadays video content is going. Yeah, it's a primary form of our consumption of media has changed from text to audio and video. I'm so tired of hearing that same mail voice of the digital. I know that it's an AI voice because, you know, a lot of times you can just tell that this is an AI voice. So people have been using it so many times. So that's one use case, the other one being like for example, I'm a shy person. If I ever create a professional account, I'm not going to put my video talking. If there is an AI digital avatar out there, I'm just going to use that and have that person talk and whatnot. And then you mentioned AR&VR. It is Apple came out with their own product. We already have Nordics or Meta, except that it's doing really, really great with augmented reality. Plus all of this. Eventually the only way for all of them to make money is advertising. So eventually I do see a lot of use cases and just a new domain of all these digital avatars. So you can have a text, audio avatar and then video and plus eventually content creation is something that's leveraging AI really, really well. It's really interesting for a couple of reasons. So bringing up, you know, the AI voice generators is really interesting. I have like two thoughts that I'd love your take on this. So first of all, like, you know, the usage of the avatars to replace you in a video or especially your voice. I know a really fantastic educator who does really great game development educational content and he's not a native English speaker, but he still has a really good grasp of English and he is able to get his points across. But he's really concerned about his accent and so has started replacing his voice with an off the shelf AI1. And there's part of me that's like deeply uncomfortable with that. Obviously it's all about his comfort. He's a content creator. Whatever enables him to create what he wants to create is great. But on the other hand I'm like, this just feels like we're just paving over individuals and their stories with like the same AI Beige mush. Right? Am I weird for finding that uncomfortable? No, not at all. I'm glad that you brought it up because I've been seeing this ad on Instagram a lot where it's talking about how one of the product out there, it's replacing people's accent. The ad that I see a lot is basically again, Meta knows that I'm not a native English speaker. So the ad goes something like the preview is someone French speaking English. Now French have the accent and then there is a content. So I think like it speaks a lot about what sort of audience people want to target. I do find it really, really uncomfortable. Plus it's really hard for me to connect. And I mean I'm still a human and not humanoid. So having that authenticity being replaced with an AI. And I mentioned earlier that how I am, I got like frustrated with just hearing the same AI voice because everything sounds like TikTok voice, right? Like exactly. So it might be easier for folks if you're trying to expand your reach. It might make things easier for that target audience to parse that content easily. But then you will be lacking that authentic connection with all those people. You will be replaceable because you know, if they ever hear you giving interview somewhere, they will realize, oh, this is how this person sounds like. And this is usually what I hear. So they might just not make the connection. Again, we still don't know how human brain works. I mean we do know some parts, but I'm just going to like generalize. But there are like this things that, you know, how things relate with each other. That's one of the unknowns that's out there. So there might be that disconnect that you might feel and you might lose that audience or you might find more engagement from those people. So I do find it weird. Again, it will help you expand your reach and whatnot. But then there is this, you know, there's trade offs, right? Yeah. So I mentioned those two things. I guess the other one to I guess link these things together. You mentioned being able to recognize that voice is not human, it being particular AI voice. But then also that lots of people are using these things for content creation. And even Aside from the AI voice generators, just for example, text coming out of ChatGPT, I think a lot of people have identified that there's like a particular feel to it, like they can recognize when Something's written by ChatGPT. It has ChatGPT voice. Do you think that's a form of the uncanny evali effect in practice? Is that what we're seeing there or is it a different phenomenon? It can be binary. Now it depends on how often you interact with ChatGPT plus what sort of questions you ask in ChatGPT. So if you're having a philosophical discussion with ChatGPT, chances are really high that it will spit out something that will give that Uneasy feeling, mess. But if you're just using it as a transactional tool, in that case, I don't think that you may experience any sort of that unease, except for the fact that, oh, this thing can replace me. So again, that is a spectrum of uneasiness that one can feel from all these AI tools. One being, I haven't used it to have like, you know, for the software discussion because I like to do it with humans, just because it just gives me like a personal perspective. And philosophers are someone that AI will never be able to replace. They'll be happy to hear that for all the philosophers who are listening. Yeah, yeah. Plus also like business leaders as well as like long term roadmaps and whatnot. AI can never innovate. Well, I wouldn't like to call it never, but as humans we know like what sort of problems there are. And in India a lot of times we do like lack sometimes a lot of tools that are sometimes just so easy to get out here. So any sort of like weird problem which is just like annoying and this is about, you know, like mechanical problem instead of like solving it, sometimes I fix it. And again that are there fixing something and solve something. That's a special word for this, right? Jugaad. Yeah, that is this all this jugaad. AI will not give you, at least that's my belief that AI cannot give you this Jugado. Yeah, that's something that we have to cover because this is what I have now. How do I make this thing work with whatever I have? We are not there yet, at least in the evolution of AI because you know, AI itself is evolving. Who knows what new version of the thing will have. But again, there are all this weird things and you know, small aspects, but human brain, I still believe that it's irreplaceable, acquiescent. We actually understand 100% how it works. Yeah, that spirit of innovation that allows you to make a really, truly Rube Goldbergian solution to a problem in a very hacky. The bodging it with duct tape, as would be said, I guess in a very British way, is a very human instinct. So you mentioned popularity of content creation and some of the positive ways that's being used. But obviously it's being used a lot for fake content, for harmful content. I feel like I'm becoming more and more obsessed with the genre of AI generated ads because they are obsessed with lions for some reason. But you know, obviously a lot of these things pose risk to consumers and to people on the Internet. From your experience recognizing these things yourselves and using them. Would you have any recommendations for consumers for recognizing and being aware of false and fake AI generated content? Some of the moments, for example, blinking and AI cannot really, really imitate how humans blink. You mentioned about. I'm glad you brought it up. Like defects which are really, really a big nuisance out there. And it's an election year. Plus it's like most of the times, like if someone is creating all these defects, it's usually just harmful. So usually like really, really smooth movements. And a lot of times their moments are weird. I mentioned about eyes a lot. Also, like lip moments. And then sometimes like they don't have wrinkles and whatnot. So there are these tiny details. They can get like voice. Right. They can look alike, like, but then if you really do pay attention, you will always notice that, oh, that person is blinking differently or it just feels like unnatural. Right back to the uncanny valley, right? It's the little things that tip you off. No, even like as we talk, I'm moving like my hands around stuff like AI cannot. Oh, how is this person going to move his hand when he's talking about this? So there is all that aspect, even like moment of head. I'm doing the hemi type thingy in the shoulders and whatnot. So all those like special aspects, like background looks weird. Sometimes it just looks like too perfect. And nowadays like actual content creators, photographers especially, because I do love photography, they are introducing imperfection just so that people can be sure that, oh, this is human hands will shake. So yeah, you have to pay really, really good attention. Sometimes it's going to be really apparent, but sometimes it just won't. And a lot of times AI cannot generate content which will inspire feelings or whatnot. So a lot of times it's just like robotic or that sort of response that you may get, like voice tones, like humans voice. For all of that, you need to pay attention. And that's something that's then trending on the low note ever since the evolution of shortfall content. Yeah, you need to have been tipped off that something's off to then start inspecting it in that way. Right. If you're just scrolling on your feed, you're not going to give anything that amount of time. Exactly. You know, all this technology being available and with the upcoming election year, it's going to be an interesting time for sure. But as we get towards the end of our time to wrap up on a slightly happier note, you are working software engineer, you're keeping tabs on this new emerging technology, both technically and also philosophically, are there any books, websites, podcasts, et cetera that you'd recommend people follow to keep up to date on these topics? I would say like, first of all, this podcast because there are business leaders and all sort of topics that one can get into. And then I always recommend TechCrunch, usually have really, really good content. And then I just sometimes just scroll through medium, even LinkedIn. So usually like I don't do like a specific thing just because tech is a topic where people have so many different takes and I still want to be irreplaceable. So I still like to get so much like a different perspective. So usually I still like follow the generic search mentality and even like Reddit, a lot of Reddit subreddits, they have such a good information. But again, Reddit is anonymous. So a lot of times, like you have to tread carefully. And that's also getting polluted by AI. They mentioned LinkedIn's really interesting. I feel like. So this is a bit of a common segment for our show. We like to ask this question and more and more we're hearing LinkedIn come up. Yeah, it's benefiting a lot from other social networks having problems. Plus I try to, instead of scrolling on the feed, sometimes I just search for a keyword and just look for the post. That's because lately, like the timeline has been depressing just because of all the layoffs and whatnot. And nowadays like people have been obsessed with marketing themselves so much just because, you know, they want the visibility. That's something that I also really worry about humanity in that aspect. What's this going to do? Like anyone who is like really good salesman, they will be like up at top. No, regardless, you know their competency. But again, with humans, it's how it always has been. And Medium has been also like something really, really insightful. I would say, like it's more nuanced. So yeah, I would definitely. People like tabs and then you can also subscribe to Digest. So I have been using technology since the Quora times. The last core user left. A lot of times it's interesting. Plus like you can always have all this stuff feed into your RSS and whatnot. Perfect. That's great recommendations. Thank you so much. Yeah, thank you, Katie. This has been super fascinating. Really appreciate your time today. Thank you. It's been Our pleasure to Dr. Yazo. Thanks. Beyond the Screen, an Ionis podcast to find out more about Ionus and how we're the go to source for cutting edge solutions and web development. Visit ionos.com and then make sure to search for Ionos in Apple Podcasts, Spotify and Google Podcasts, or anywhere else podcasts are found. Don't forget to click subscribe so you don't miss any future episodes. On behalf of the team here at Ionis, thanks for listening.