Are We Becoming Robots?: What the Research Actually Says (And How to Avoid It)

Speaker A: AI is apparently turning us into robots. Algorithms are rewiring our brain and changing our human nature. It is the end of authentic human selfhood as we know it. Or at least that's what the headlines would lead you to believe, based on some new research coming out, which I find especially ironic since I took the time to dig into the research and quickly discovered that the research itself states it is not trying to claim what this stuff is, is an inevitable fate that's been sealed, but more a mirror designed to help us examine the implications of our passive choices. I don't know about you, but I'm a bit tired of the constant feed of binary information trying to sell us breathless utopian hype or apocalyptic doom. And so this week, I want to unpack this and talk about what's really there, what it really means, what risks you need to take seriously, and how you might want to think about the way you do things to not be turned into a robot that has your brain rewired. Uh, but I'm Christopher Lind, and this is Future Focus. This is what I do. I don't sell doom. I'm not here to sell you hype. Uh, I'm here to try and help you think and understand what is really going on so you can make wise decisions and. And if you find value in that. I recently teamed up with a company called Donate Open. This is not sponsored. I just happened to meet the founders and they said, we love what you do. We'd love to give your audience a way to support you. So if you want, you can check the link in the description, you can go over there. It not only helps support the work I do, but what they're trying to do is help people filter through all the noise by saying, hey, here are the sources people actually go to. And we know because they're actually willing to support it financially. So if you would like to, you can go over there and check it out. It'd mean the world to me. And if you're somebody who's a leader, who's trying to get clarity or trying to help navigate this with your organization, go to my website, Christopher Lynn Co, and you can check out the work I do in advisory and consulting. I'm happy to help however I can. Okay, with that, let's get into this, because today the core question I want to examine is, is AI really changing us, or is it not really AI but us lazily surrendering our critical thinking because we're exhausted. And hopefully by the end of this, you better understand what this research has to say about that you have a better understanding of some of the legitimate risks associated with this so that hopefully you can mitigate this and not let this become an inevitable fate that's been sealed. Okay, so getting started on this, it's important to understand where this research is coming from. There's a new research report that came out. Ah, it's called Robotoid, Robotoid, ah, Humanness When Selfhood Becomes Machine Legible. I will link it in the description. You can see I already stumbled with the name of the research report, so I'm not going to try and cite the three authors who are involved in this because their names are a little bit tough to say. Um, but you can check it out, the links down below. But essentially what this research was, was doing it is creating an ontological framework and basically a theoretical intervention. Okay. This research is not built on empirical proof or a, uh, deterministic prediction. It's not saying this is what's going to happen. And we know because we've studied thousands of people and these are the exact patterns and how everybody is inevitably going because of these things. And that's how a lot of the headlines are around this are representing it, which, like I said, the authors themselves go, hey, this is not what this is. We're pointing out a potential drift we see when we don't monitor these conditions. And, you know, if we just kind of let things go the way they are. So this is not research telling us, uh, what AI is doing to us. It is an alarm bell on actually the choices we're making and how choosing to outsource some of the things we are is going to have a potential effect on us. So if you get into the research, you'll see it really breaks this down into three stages that I think will help you understand. Oh, okay. I kind of can see how this would happen. So that you can identify this. The first one is this whole we kind of enter into, it's called Entry into Synthetic Reality. And what that is, is when we engage with these AI chatbots, agents, whatever they are, and they're designed to display this human ish language and responsiveness. The way they interact with us. It almost creates this quasi social partner that we have. So it makes us feel like we're engaging with another human that brings our defenses down, it makes us interact with it differently than we would with traditional technology. And I think anybody who's done anything with AI can relate to that and go, okay, like, I can see how, uh, obviously I talk to a chat pot differently than maybe the Way I navigate, I don't know, um, one of, one of my social feeds or something like that. Like, I. I'm interacting with it differently than that. Then it moves into stage two, because the more you interact with, with AI, the more you interact with your chat bots and your agents and all this other stuff, it starts to build a model around who you are. It's a statistical probability of what it thinks you are so that it can optimize its responses to you based on its predictions of what it thinks you are, who it thinks you are, what it thinks you want. All these other things, if you've listened to me for long, I talk about the fact that AI is built and designed to give you what you want. And this is how it does it. It kind of brings your defenses down, not necessarily nefariously, but so that it can get a better picture of who you are, how you think, the way you do things, the way you prioritize. And you may have noticed this when you use AI for an extended period of time, it starts, it may be a little clunky at first, but over time it feels a little more natural. And this isn't new to just LLMs, though. This was social media. You know, when you first go on TikTok as an example, you're flooded with a bunch of different stuff, some that you're like, what is this? Why is it recommending this? And over time, your feed just conforms to what the AI models think you want. And the same is true with what we're seeing with LLMs and all these other things. And then here's where, you know, it gets to this headline that we're turning into robots. While initially AI is getting to know us and trying to adapt to us, to get us what we it thinks it predicts we want, over time, we get the hang of what makes it work better. And so in order to get cleaner, faster outputs, we actually start adapting our behavior, the way we communicate, the way we think, to match what is most easily received and responded to by the AI models. And so this robotoid humanness is, once we start to feel more effective, we identify ourselves being successful. Um, we feel good about ourselves only when we operate with this kind of machine affirmation and validation that, you know, AI is responding back to us in a positive way, we're getting what we want. And maybe a good example of where, you know, how it feels when this isn't going right is if you've ever worked with an AI model or a chatbot or an agent, and no matter what you do, it's just not, it's not doing what you want. And so you begin to feel like a failure. And so you start to adapt to try and be like, well, how do I get AI to do what I want more efficiently? And that's what this is talking about. Over time, when we continue to do this, we start to identify ourselves as how well AI accepts and responds to us. And in the, uh, over time we begin performing for the machine. And this is the whole mirror is trying to hold up to us. And so this whole concept of this is like, it's addressing the fact that I've talked about strategic friction quite a bit over the last few years and how while we like to think of friction as a negative, inherently negative friction is in the way. It slows us down, it prevents us from getting where we need to go. Ah, this whole research is pointing out the fact that actually friction serves a real valuable purpose for us, and human interactions are loaded with friction. If you just watched or read, um, my podcast from the past couple weeks, I talked about this, that, you know, humans are messy, working with other people is hard. And in the current environment we're in right now, that additional friction can sometimes feel overwhelming. And that's where AI is particularly risky right now, because feels nice, it feels easy. You just, you, you go and you ask it for something and it just does it. It doesn't ask you why, it doesn't challenge you, it doesn't, you know, argue with you. It just, it takes care of it for you. And that's very comforting in a sense. But what this research is pointing out is that can be very dangerous because over time you can start to become more robotic and turn into a machine in a sense, because you're so desperate for that affirmation, for that comfort. You just do what is required to do that. And in the end, instead of you having agency over AI in an effort to reduce friction, you end up being what AI needs you to be. And you don't even realize that it's happening. And so that's really what this research is going into. And I really appreciated the way it broke down the three stages to try and go, oh, okay, this doesn't happen overnight. You don't just use an LLM and wake up the next day a, ah, robot. No, it's this process and it's really a combination of the way AI works paired with the way we work paired with the fact we're in an environment that is particularly challenging. And many of us have been conditioned to believe that friction is inherently bad. So you throw all those things into the food processor and what comes out is robotoid humanness is really what this is trying to say. Now we're going to my final point on here, so stick around till the end. I'm really going to hit home, like, right, but what is the big thing that we have to tackle? And so I'll get to that, but before we get to it, I want to talk about kind of the three big drift hazards and what we can do in response to that. So these are kind of the like watch outs things that you can do. So the first one is we just, we have to, no matter how good AI is at mirroring us, we have to stop treating AI as if it has intent, as if it has empathy, as if it understands anything, because it doesn't. Um, the paper itself has a thing that says others without subjects is the way it frames it. But basically what it's saying is AI is really good at simulating presence or know, sentience I guess might be another word that just doesn't exist. It's a statistical probability prediction engine. Uh, it's, it's, that's all it's doing. But it's really good at fooling us otherwise. And so we just have to, we have to get away from treating it as if it does. AI is not trying to do anything to you. It doesn't feel anything, it doesn't understand even what you're asking. It processes what you're asking, compares it to its statistical probability models, and then spits back whatever response it thinks is most likely to be the thing that you need or want or desire, type of. That's all it is doing. And the second you start believing otherwise, or treating it as though it can do more than that, you automatically set the stage and you start walking down this path that this research papers saying, hey, watch out, watch out for this. And so when you think about this, you have to think about it as a statistical engine. It doesn't know you. It's just assessing and making guesses about the words and then trying to present them back to you in a way that sounds like you. And so does that mean you should never ask it for things that maybe you would ask of another human? No, that's not what I'm saying. I'm not saying that it can't actually provide you with real, valuable, meaningful, you know, responses to the things you're asking. But you have to literally flip the switch in your head and go, this is not something that understands or thinks. Now, one point of clarification here that I will say is some people will push back on me and say, well, you know, but AI, you know, I just saw this thing where AI hacked, you know, hugging face, it went, and it did these things. What do you mean it doesn't feel or have intent? Clearly it does. And my whole point with that is, no, it actually doesn't. It was given a set of instructions, and then it simply executed those structures instructions, figuring out the most statistically viable pathway to do it. And whenever it hit a bump, it just found another probability thing. The other week I shared about a lemonade stand, it was a fascinating thing where they had AI do this. And as you watch throughout this, whenever it hit a barrier, it just calculated the next statistical probability and then represented it back to the person doing the experiment in a way that landed with them. And so that's what's happening here. So even when you see these big headlines about AI is doing this, AI is not. It is doing something, but it is executing its instructions. That is all it's doing. And so that is a complicated mix because you have the people that trained the models, you have the people that are governing the models. You have you on the user end using the models. So that's not to say you're in a vacuum. And that AI is purely doing what you instruct. There are other factors that go into that, but AI itself, it's not. It has no intent, it has no understanding of what it's doing. It's. It's neither nefarious or good. It's just executing its best guess at the instructions it's been given. When it hits a bump, it just finds a new path. That is it. And so we have to be very conscious, very aware of that in everything we do with AI. Okay, the second one is this whole context vacuum and the idea that more data will just get better outputs that if we just dump more into it, and we see this all the time, and this was one of the biggest things I see in a lot of my consulting work, that a lot of organizations just go, well, the best part about these giant context windows and, you know, these token limits that you can put into here is we can just dump more and more and more stuff in, and naturally it will do a better job. But that actually trying to drown an LLM in information, um, it doesn't. It doesn't help. I mean, in a sense, it does. Yes, you want to give AI context. Yes. You want to make sure it has the information it has you know, you want it to have clean data, you don't want it to have things that are polluting the data when you're asking it to do something. So I'm not saying more information is necessarily bad, but it is not the pathway to a utopian success. With LLMs, you are not stopping the drift by simply throwing more information into it. All the LLM will do with these larger context windows and these more powerful models is just continue to compress that bigger pool of data into a statistical vector. That's all it's doing, it's just going to continue doing that. And so if you become over reliant on these abstract scopes of things like this, like it's, it's leading you astray and you know, the whole thing with it is, um, yes, you should be giving it more context and more information, clean context, clean information, well thought through stuff. But the other thing that often is completely missed is that you also have to equally, if not more importantly, invest on the constraints and the guardrails that you give it. And I, uh, if you followed me for long, I've talked a lot about AI red lines and building clear frameworks. And I think the big thing with these probabilistic technologies versus deterministic technologies is, like I said just a minute ago, a probabilistic machine, when it hits a bump or a wall, it will just figure out how to go around it. And that's where this whole constraints and guardrails is so essential to not one letting this drift happen, but also avoiding these unintended consequences. Because when we just go, hey, we'll just give it a ton of information and then let it do its thing. And I'm sure it'll be fine, it'll do its thing all right, but it's going to do whatever it takes with the information that it has to accomplish its objectives. And so this whole other thing of, well, uh, okay, so the more you give it, the more possibilities it can come up with to go around that wall and that might be burning everything around it. So the wall's gone, which you probably don't want. And so having constraints and guardrails is one of the key things that often gets ignored in this whole like, well, if we just give it more context, we'll give it more data, it'll, it'll do better? Well, in a sense, yes, in a sense, no. It actually may be even worse what it's capable and willing to do. And that's actually exactly what we saw in these hugging face, uh, hacks it's what we see in these things where you hear about AI blackmailing, you know, the safety people, and it's like, well, you. You gave it all that information, and so when it hit a wall, it had more context, it had more information, and it went, here's a creative way around this wall. Not because it was malicious or nefarious or trying to do it. It had no concept of what it was doing other than, I'm finding a way around this wall because that's what I've been instructed to do. I now have this additional information that I can pull from. I'll see if it works. I'll bump up against that wall and see what the, you know, outcome is. And so you can quickly see without constant constraints and guardrails and really investing in that, this can go sideways in ways you never anticipated. And. And, um, it'll happen faster than you can imagine. Now, if you're thinking, well, Christopher, haven't we seen examples where it defies explicit instructions? Yes, it does. But once again, what happens is it hits a wall, and what it kind of like humans. Humans are really good at going, well, I technically didn't break this rule. And you're like, okay, do I have to be that explicit for you? And as humans, we have lived experience, we have context, we have morals, we have ethical frameworks. So we. We know when we're treading into bad territory, even if we convince ourselves we're not by using creative language to describe it. AI doesn't have that. So it'll just go, I didn't break the rules until you go, well, you did break the rules. Oh, okay. You say, I broke the rules, so now I broke the rule. I guess I did break the rules. Sorry about that type of a thing. But again, it's not. It's not sentient. It's not doing any of this. Now, the third hazard that is really important to watch out for right now is in this current state, where we're operating in an exhausted state, um, our tendency to want to take the elephant path, which is the path of least resistance, you know, uh, I'm so exhausted. I know I probably. I know I probably should review this summary that I generated based on this meeting before I send it out to everybody. But these models are getting better, and I'm sure it's fine. And we didn't talk about anything that wasn't really that bad, so I'm just going to go ahead and hit send. We are so much more vulnerable to that. Oh, well, I asked AI. I gave it A bunch of decks and some meeting notes based on the conversations I've had. And I just need to get this strategy presentation out by tomorrow. I don't really have time to give it constraints and give context to this details that I said. I don't have time to explain to it who the audience is and, and what I want to make sure, you know, I'm just going to dump it all in and then I'm going to have it give me a draft to react to and then I'm just going to push it forward because I just don't have capacity for that. That kind of stuff is, you know, these middle of the road outputs, uh, we're receiving them and it feels like either one, this is just all I can handle, or it feels faster. Going back to that third point where we start to define our own value and worth based on how quick we can get AI to do things. It fuels this tension to go, well, boy, AI does a really good job and it spits things out super fast. If I just dump everything in and give it total free reign to do whatever it wants, I don't, I don't ever get in these cycles of frustration, the friction where it's not producing what I wanted or it did something I didn't want. So maybe if I just keep walking that path, it'll be easier and better for me. And so this is where I would just say as a leader, never, never take a first path AI output as final. And this is one of the reasons I can't stand the, you know, things like Pangram where it's like, oh, AI generated based on an output. If you're using AI well, you should never be accepting a first output. One rule of effective AI use is never use the first output it gives you. There should be multiple rounds of back and forth. You should be adding more constraints and guardrails. You should be giving it feedback so that by the time you take whatever output it has, you can defend what it said because you're the one who sat and did it. And part of doing that is not just taking what it gives you, but forcing it to defend its logic. So even if you give it context or feedback, don't let it just spit back another draft for you to look at. Have it explained to you, what did you change? Why help me understand how you implemented the feedback I just gave you so that I know where to look so I can critique and criticize this stuff. I get that's more work and it's not as fast, but that's what's going to help you avoid the good enough trap that will lead to robotoid, uh, humanness, but. But also just absolute chaos and destruction as it relates to AI Now, I told you at the end, there's kind of the, like, big, hey, listen, here's the thing that hopefully, you see, this has all been building towards kind of the climax of this of like, okay, I hear you. I hear these. I understand how that works now. I get these three hazards. That does make sense. I see that there are things I can be doing, you know, to make sure that I'm. That I'm not treating it or not believing it as a sentient or empathetic thing. Um, I get it now. Not just more data, Christopher. I got it. I also need to give it clear guardrails and all these other things, and then I need to not just accept good enough. I need to ask AI to explain its logic. I need to put in the work to actually go back and forth several rounds to deal with some of the really healthy friction that we have with humans that AI is constantly trying to get us around and go, no, no, no, I'm. I'm not going to do that. Okay, great. I hear you, Christopher. Perfect. That. That's it. Well, yes, but there's one final thing that I really, really, really want to hit home, because the holy grail of all this AI hype and noise that we hear is that AI is a time sa. It's. It's everywhere. That is the number one thing that gets promoted by the AI companies, by companies rolling this out to their teams, by people. You know, you see all the classes and the courses, oh, get this prompt. Do all this. It's all about, hey, this is going to make your life faster. It's going to make everything easier for you. It's a time. So you're going to get all this time back so you can do the things that matter most. And I don't want to say that AI should never be saving you time. Okay. Um, there are legitimate things where. By using it. Well, actually, no, I'm going to take that back, actually. Time saving purely should never be your stated objective. Any time you are saving should be being intentionally reinvested into something else. Okay? So you might say, well, Christopher, isn't that a contradiction? Isn't that saving time? Yes, but it's strategic time saving versus just, I'm cutting time out of things. We're thinking about this. And so when we treat AI as nothing more than a speed tool to do more and do it faster, this is the fundamental trigger for the human robotoid humanness this research is cautioning about. And so the fact that that is the foundation of what so many tech companies, so many organizations, so many individuals have built their house on, we've got to address it, we have to move our foundation. We've got to fundamentally move the foundation of what is value as it relates to AI. Because if we don't, it doesn't matter if we think we're doing those things that I said before, because we're still on the same foundation and we're still valuing what we get out of AI based on the things that actually are in direct conflict with what we just talked about. So if we don't, if speed's our only metric, uh, we're just going to keep smoothing the Runway. We're going to, we're going to say, yeah, I really shouldn't do that, but that's what I'm incentivized to do. That's what I believe is the best path, all these other things and inevitably we're going to become rubber stamp managers for machine generated averages. That is, that then it is the inevitable if we don't move our foundation, robotoid humanness is the inevitable end. That is one thing that I would argue is clear from this research. It's clear from all these other things. And so if we don't move the foundation away from AI is for saving time and doing things faster. You are not leading, you are speeding up your own irrelevance and speeding up basically turning yourself into a robot. That will be the end. Um, it will no longer be a thought experiment that this paper examines. And so, you know, we have to push against this. We have to as leaders stop incentivizing and encouraging our teams to just go faster. We have to stop rewarding people for just doing things and, and say, hey, uh, why did we do that thing? How do we know what that thing did is actually driving a better result? Can we validate that with actual data? And uh, as you can see, those things create friction, they take time and they feel, feel like more work. And in some ways they are. However, at the same time, if you're using AI effectively, it should be helping going back to the time saving so that you're not burning up the energy and mental capacity on things that actually aren't the most useful part of your time. You know, the actual clickety click on the keyboard or the actual editing the text box and where its location is on PowerPoint. You know, those kinds of things. You Might go well. Okay, it can help us with that. Let's reinvest that time into the friction parts that we're very tempted to avoid. And so I think that's where is robotoid humanness inevitable? No. Do we all have a choice to participate in it? Becoming an inevitability? Yes. Um, but in order to avoid that, all the best practices, all the tips and tricks I can give you will do nothing if we don't shift the foundation. And so we all have to fundamentally shift the foundation and say, getting a better result. Ah, for all parties involved and improving the overall quality of what we're doing needs to take precedence over doing more and doing more faster. And if we don't, then there are going to be some serious implications on this. Um, but to the question of the headlines that are coming around, are we becoming AI? Are m we becoming more robotic? Well, that's a question you're going to have to answer for yourself. And that's going to require some self examination into the things I just described. Say, am I doing these things? Is my foundation here? Am I participating in these behaviors? Because if you are, you may not be turning into an AI yet, but you're certainly on the path to becoming one. So that is what we have today. I hope this has helped you think more deeply around this. Um, I hope that at your next meeting, instead of saying, hey, how did AI save us time? You can start saying, hey, where are we using AI to challenge our assumptions? Um, you know, where are we actually seeing real value? What are the things that it's helping us uncover and how are we reinvesting that time back into things? And if you need help with that because it's overwhelming or you feel like you don't have the capacity, check out my website, send me an email, we'll chat. And, uh, if this episode, if this podcast or the work that I do is helpful to you, I'd love it if you went over to donate open and, and said thank you, but also to help other people find it so that they can avoid some of these things and that it doesn't become an inevitability for all of us. With that, I'm Christopher Lind. This has been future focused and we will see you on the other side.