S1 Ep9 The AI Trap

[Music] AI is everywhere right now. Every second post says it's replacing jobs, changing everything and rewriting the rules. And whilst some of that's true, a lot of it is just noise. The bigger question is not whether AI is here, it is how much thinking, judgment and consequence you're willing to hand over. Because speed sounds great, until it creates bad work, weak decisions and costly mistakes. So today we're talking about AI without the hype and without the panic. If you don't use AI for work, you're lying. [Music] Welcome to ThinkShift, the podcast when neuroscience meets business. [Music] -Haley, I'm good to see you again. -How are you, Clint? I'm good. Lots has happened, it's been last caught up, but it's been. -We got merch! -We got merch! -T-shirt, for those of you that are watching, we've got our ThinkShift podcast shirt, for those that you are just listening, switch over to video quick. [Laughter] -So you can have a look. -It looks great. Before we get into it, I just want to talk about a fun stat. So our commute here from sunny-otely, well not sunny today, that commute, we used about three to five AI things in that commute. And it's really interesting, and as a stat here, as of this year, AI is likely to be involved in four to ten times during a standard morning commute from home to office, which is pretty crazy. So everything from the smart home preparation from when you walk out the door to route optimization, so navigation systems, to active driving assistance. So if you get on a freeway, you're using that, traffic-like synchronization, which is what we just went through to get here. -Multiple parking assistance. -Multiple parking assistance, and then entering the studio. Security access, facial recognition, potentially, then there's email calendar syncing, automated workspace, but the list goes on. So it's really interesting how, even though some of us don't want to use AI, we're just using it anyway. Super interesting. -So the line is, if you're not using it for work, you're lying. If you're not using it in your life, you're not. -You actually are. -You actually are. You're even using it without realizing it. -That's right. -Yeah. -So let's talk about why this episode exists. Because AI is one of those things that is everywhere right now, and there's a lot of talk going on with it. So obviously it's everywhere, but most takes on it area of the hypophere. Right? So there's a lot of fear mongering going on. And so we're going to do neither today. We're doing neuroscience and business reality, right? Today, that's what the focus is. So the main focus here, and the first discussion point is that businesses aren't moving quite as fast as we're led to believe. You know, and I see it on LinkedIn, especially on LinkedIn. There's a lot of AI talk, a lot of AI bros, as they call them, with new AI businesses talking about how design is a dead, a copyright is a dead, this is dead, that's dead. And what I'm seeing, I don't know if you're seeing the same thing, and is that bigger businesses, particularly, aren't moving quite that fast. Particularly when they're running legacy systems, they've got previous software, they're not going to just jump on the latest trend and go, "Oh, we should do that." Everybody, thousands of people, let's change to this new system. Because adding automation on top of a bad system, which a lot of these older firms and legacy firms, using legacy systems have, is just going to make something this bad, this bad. - Yeah. - You know, this magnifying glass and a bad situation even worse. Are you seeing similar? - Yeah, 100%. And I think when we talk about the hype, which is not what this is, and we're trying to deregulate everybody actually around AI usage, is, well, everyone else is doing it so, so do I. - Yeah. - I need to, as well, regardless of the risk, as you opened up the segment with your AI, part of our lives, it always has been. In fact, AI has been around for a very, very long time, over 20 years. I started using large language models in medicine more than 20 years ago. It's just now it's commoditized, it's something, a smart kid, and also, to go back to your question, absolutely I'm seeing that the larger the organization, the harder it is to adopt. We're talking about a change management process too. So, we've got processes in place, especially in enterprise, we've got systems, we've got frameworks that we've worked hard to implement. And now we're just going to throw them out the window and use something else. - Yeah, it just doesn't work. - It doesn't work. There's gotta be an alignment within an organization, a governance, ethics, protocols, and also an adoption. And there needs to be an adoption process. And that's cool. That is perfectly okay. - And I think as well, like, all right, we're talking about bigger business here. For sure, there's gonna be startups, there's gonna be smaller one-man businesses that can definitely pivot, change, adapt, and go after that shiny AI tool now and build a business around it. Is that sustainable? Is that you're gonna be able to grow from that? But by our large, those businesses that are jumping on those quick shiny tools, they're gonna jump to the next one soon after and change again and pivot and pivot and pivot, and that sounds like a fast road to stress to me. - You're not optimizing anything. - Yeah. - Jack of all trades master of none, and literally in a big way, they could be sabotaging a business. And we see that. You know, we see that, and also there's this threat of, well, if you're not using AI, you should be. Or if you're not using it in this way, you should be. You know, if you're not coding, I mean, coding is a specialty. - Yeah. - That's, that's, that's beyond, but you should be. It's like, hang on, no, I shouldn't. - And if you don't know how to properly code, how do you know how to debug that code? How do you, when something goes wrong, when something breaks, you need someone who knows how to fix that. - Yeah. - You know, how is that gonna save you and all these instances? And we're gonna talk about some of the catastrophic cases where AI got it wrong later and around ethics and concerns around that. But firstly, I wanna talk about the science behind this seduction of it, and it's largely to do with speed, right? - Yeah. - So speed feels really good, you know? - Yeah. - But that doesn't make it wise. But why is speed so seductive to the brain? - So the brain prefers, we say, effort reduction, but we're comfortable with something that is quick and easy. That's how we operate. You know, it's sort of the foundation of behavioral science and neuroscience is around the dual process theory. And this will come up a lot around adopting this principle of the dual process theory. And again, thinking fast and slow is the book from Nobel Prize winning Daniel Kahneman in where he talks and elaborates. And system one, thinking fast, thinking fast and slow. That's the fast bit, 95% of all decisions. So it is seductive, it is sexy. It's also a metric. I'm faster than you. We're competitive species. We're humans. We want to be, you know, for good or bad. Competition at some, that's how, you know, survival of the fittest right. If we're better, if we're stronger, if we're bigger, we're gonna win. And that's ingrained in us in how we operate and our DNA, you know, hormonally as well. We think, we think, the faster the better. And AI contribute to that thinking because the outcome is faster or potentially it is faster. It reduces load, we find it easier. It seems to be simpler in the short term, not necessarily in the long term. And so it is seductive, it's sexy. So the brain prefers effort reduction. Absolutely. It prefers it over anything. Over anything. However, with our system one, thinking, the one point that's really important to note is that it is highly error-prone. Yeah. So the faster we go, the more risk we create and unfortunately for most of us, we forget that bit. And we avoid that bit. We just want to go fast, but we don't realize the ecosystem that we're in and how detrimental it can be to what we're doing. Our businesses, ourselves, our personal professional lives. So we need to think a little bit more about it. Yeah. So as well as speed reduces discomfort of not knowing, which I could see would be catastrophic, particularly in universities, schools where your hand had so many assignments, reports to do. Of course, you're going to go, "Well, I can get this done by AI. I don't know enough. I know a little bit. I don't know as much as I should about it, but it's going to get me the job done very quickly. So I'm going to just lean on it." And that's super dangerous, right? So I think someone said to me, more recently, it might have been my brother and it'll actually shout out to him. He said, "If you're going to need some kind of serious doctor any time soon, make sure that you see them in the next 10 years, because after then, we're going to have a lot of doctors train on AI. And it's going to be a concern. And I don't know what your thoughts on that are, but are you seeing that kind of, I don't know, the skill level may drop over the next 10 or 20 years." And there's evidence around that and clinical studies around that already, which we call cognitive regression. I'm going to talk about that. Yeah, and we were talking it. When we were preparing for this, when we were talking about cognitive progression and cognitive regression. But 100%. And we think we can dump something into an algorithm. And remember, AI is an algorithm. So it's just one whopper of a mathematical equation, which is what it is. And again, very seductive, very intricate, a web of Atlas and information and knowledge and intelligence that we as humans provide it, we feed it that information. but we were lying on it and We, not everyone does, I understand that. But to a certain point, we need to train ourselves in how not to rely on something that is, I mean, as the name is AI, artificial intelligence. It's not genuine intelligence. It's AI. - AI, yeah. - Because it is process, it's a process, it's process, it's synthetic in a way, using, you know, information, and very good information too, but it then needs to interpret, extrapolate that into something that is reasonable. And we are relying on it, and we are starting to see this cognitive regression. And unfortunately, a lot of things that I'm seeing, and I talk to a lot of people about, is we're not double checking the facts. We are relying on it too much. We think, well, the Atlas is huge, the knowledge is huge. I'm reading the beginning of what it's sped out. Sounds great. The rest of it's gonna be great. - No, no. It often isn't. - There's a famous story about the New York Times article. I don't know the exact specifics around it, so I'm kind of paraphrasing here, but the reporter or whoever wrote the article, went to AI to get a summary of a particular Canadian candidate, and then it sped out a quote said by this candidate, but the quote was hallucinated. It was completely fabricated, and that got published. So it's that fact-checking piece that so often gets missed. You know, is there anyone behind the wheel of the tractor kind of thing? You can't just let the tractor go and play out of field without the brain behind it kind of steering it. So, but I was gonna talk to you actually about, this is an AI related, but when you're talking about regression of the brain, and I've seen it myself, and anecdotally, when I use my calendar now, I cannot organize an event unless it's in the calendar, 'cause I just will not remember it. You, I have conversations with you, and you will say, I need to check my calendar. Or, and I do two, but to a point that you're extremely relying on it. It's not a negative thing. I'm not going, you're clinging your line on your calendar, but hang on, ooh, clinging your line on your calendar. All the time, which is crazy. I'm an alliance. They're alliance. It's the same with phone numbers. Yeah. Does anyone remember phone numbers anymore? I think I remember two. Because we're not phonetically reciting them, which we used to do. They don't need it anymore. So, there you go. So, that's a regression. It is a regression. There you go. And you see it a lot in, well, we're going to talk about curiosity and creativity. Yes. One of the things you see a lot, and you've talked about in significant detail, is the reliance from that creativity in terms of design, and using your own neuroplasticity to come up with, you know, an outcome or with a design, or with a document, or with a process. And people are starting to lose that. They are. 100%. Okay. So, still on the science piece. Why does fluent output trick us? Fluent output. So, it's such a great term. Yeah. Again, it's coming down to the seduction of speed, isn't it? Well, it's also something that's written really well. Yeah. We get seduced by it. Yeah. So, even if it's wrong. That's 100%. So, if we read it, and if it sounds. So, what we often do is position ourselves or put ourselves in the shoes of that person or in the shoes of what's come out. So, if we're reading something that AI has produced, we think very quickly, because this is all subconscious, by the way, and we think, would I write that? Is that something I would do? But again, going back to your system one thinking, you know, highly error prone, we start compromising and making allowances. Well, I wouldn't say it like that, but it sounds okay. Yeah. And it's too much time for me to change it. So, I'm just going to assimilate and become that person. Yes. That's not good. And that's what we're doing. So, we're starting to change the way that we're thinking, the way that we're writing, the way that we're delivering and executing, you know, information and projects and compromising, because it is. It's, you know, fluent, it's quick, and it's easy. And of course, the brain, I've mentioned many times, how we operate, it takes lots of shortcuts, you know, with 35,000 decisions a day, approximately. If we can reduce a couple of those, the brain loves that, and it has to, to be able to survive and to do what it needs to do. So, and this is why this speed is, it is dangerous, but it's also needed. So those 35,000 decisions, that's why I remove a couple from calendar, those decisions I don't have to make. So I'm down to about 32,000. Thank goodness. Yeah. But, you know, while Clint, all you do is make up that extra 3,000 with something else. Yeah. We think we're reducing our cognitive load, but we actually. We're showing it somewhere else. We're creating a vacuum and we fill it. We fill it absolutely. You know, we're, you know, the term, we're busy being busy, and we've got to a point now, where there's an expectation that we want to be doing something all the time. I think we feel important, we feel significant, when we're constantly doing something. So even if you do reduce your decision making a 32,000, I can guarantee you, you will ring me and go, I've got another 3,000 things that look decisions that I need to make now. That's right. Of course, because you put it in your calendar. Okay. So, why can speed be bad for creativity and originality? It was my next question to you. I know why it can be, but not from a scientific perspective. If you could explain why that's so. But curiosity and creativity is really important for the brain. So to explain neuroplasticity and, you know, your professional neuroplastician, I'm a master neuroplastician. This is the area of expertise in our ecosystem in which we operate. And we are very passionate and strong in building others around us and giving them these tools to be able to do that. And so what that is is creating neural pathways and you, you know, learning neurons to be able to expand thought processes, be less judgmental, move beyond what everyone else is doing. Because everyone's down the center lane. That's why we see the same stuff all the time. Working in the periphery or in the divergency is really important. That's where you get your uniqueness. That's where you become a key person of influence. And you really make an impact and make a difference. But to be able to do that and to be able to create these new learnings in these neural pathways is you need to be curious and you need to be creative. Copping or creating ease of thought will converge, not diverge your thought. So we're starting to see a decline. We're starting to see a decline in the creativity. You know, people were starting as a saying, "Oh, everyone's, you know, AI spits out the same sort of thing." And it does. But we are not then interacting with that to change it. We're accepting it because we're losing that. Yeah, we're losing that neuroplasticity and that creative trait that as humans and social creatures, we generally have. If we go back and think back to you, caveman, I mean, I'll always bring it back because the foundation and the fundamentals of a human species or any species is the same. So what happens is, you know, we had to think. We had to think where we're going to get food. We had to think how we're going to make shelter, we had to think how we're going to close the children. We have to think we had to be creative. That's my day. If you disrupt up my day. We had to think of transport. But we lose that. So we start thinking, okay, well, how are we going to do those things? I don't know. Put it in AI. Very, very dangerous. Very dangerous. That's up with the answer. And we go, well, that's good enough, I guess. Who's going to check? But everybody is checking now. That's right. But the unfortunate thing here, though, is the brain actually likes that, which is why it's addictive. Okay. Because the brain actually likes that. I don't have to think too much. Because the ease of processing is what is, you know, it creates an, an uncomfortable environment for us to exist in. And then we don't have to worry about that. But there's always something to worry about. So it's about prioritizing what should be done properly. And is it, is it kind of a bit of a gas sliding scenario where the AI will serve up something creative to you? And you'll go, oh, I guess I like that. AI says, it's good. And I guess I like that. That's just kind of tricking you into thinking that that is not mediocre. And it's great. Especially if it comes up with an idea you haven't thought of. Yeah, right. So it will spit out that result. That's genius. Oh, what? I didn't think of that. So clearly the entire, the entire document or the entire text is genius. Yeah. And we get caught. Even though the quotes are all wrong, they're fabricated and hallucinated. That's right. And we've both got examples of that around some of the dangers that we have seen firsthand with AI output. And we're putting way too much reliance on the outlaws. There still needs to be a decision-making process. And we are relying on a mathematical equation to do that for us. Remember AI, I know this is very controversial, but as a behavioral scientist and a neuroscientist, one thing we talk about a lot is the lack of consequence. You know, there is no emotion. It doesn't reveal it. Starting to learn it, it's a long way away. That is a real issue. Yeah. Is consequence of the AI output. Okay, so now we're going to talk about where AI genuinely helps and where it hurts, right? So where do you think AI is genuinely useful business right now? So for your more standard practices, so the things that you need to do that, you know, in terms of like admin, some of, you know, your, well, I've got notes that you've written here, which I actually like. No, I'm actually like as I write what you've written is admin summarizing, organizing drafts and first passes. Admin, absolutely, because things that are low risk, summarizing sure. So often we will have some, you know, meetings and we can use AI to summarize or give us an opus of what we have talked about with an AI transcription. Again, that needs to be checked because it may not be 100% correct because you don't know where the emphasis is and how we emphasize even with our inflection of our voices. I mean, I'm, for me, I'm extremely, I'm going to talk a lot and I'm very animated in the, in my language and in my voice and in my tone. AI can't pick that up even though I've trained it to it thinks what I'm saying is monotone. So if I get excited, unless I use language that is excitable, it won't, it won't know. So it loses context. Okay. Do emojis help that? Emojis can help. Yeah. They might have. Emojis can help humans. That's for sure. Well, a lot of people just communicate with emojis. Yeah. Because tone is misrepresented without emojis. Right. I remember back in the day before emojis, I get myself in a lot of trouble, particularly with my now wife over text because she would misread the text. So you throw an emoji in there. Yeah. That's right. It's much better. Yeah, it is. It is much better. But also, you know, you've got your first passes and I like that. So getting a draft of something and you know, popping something in, seeing what it looks like. Like a syndrome. It's a tool. Yeah. To start something. Absolutely. To give you, you know, a head start, but the fine tuning, the fine tuning needs to be your brain. So it needs to be human intelligence, artificial intelligence when it comes to fine tuning. So those kinds of things go for it. Oh, good. So that brings me to where does AI hurt outcomes? Yeah. Let's talk about that for a minute. So obviously high-stake decisions, you would say anything requiring accountability. I think it goes without saying though, even after first pass, anything needs a human pass. You can't just ask it something and spit it out and then paste it and off you go. Super dangerous. Crotations, analytics, data. Yeah. I mean, I was working with a client. They do a lot of work for clinical trials with drugs and they put something into AI just to have a look and see and what the result would be or what it would spit out. And it actually came out with a drug saying that it was already on the PBS and it wasn't. Absolutely. So a total hallucination, which essentially is a lie. Now, these guys extremely experienced in what they do and this was a little bit more of a test. But it just showed and we use that example to show others just how you can't rely on that kind of information. Yeah. Because it's obviously looked into the Atlas, looked at a similar drug or a drug that's similar in another health system, whether it's the NHS in the UK and gone, well, that's reimbursed. So clearly, it's reimbursed here in Australia. And that's just not the case. So I had that then been published, you know, dire consequences. Yeah. So, yeah. So any of those kinds of things, it's in a high risk situations, need to be double checked and should definitely not be relied upon. And I've also good here work that relies on truth precision or trust. So really good example of this happened recently with a local cafe in our area actually published marketing materials for a new business and all the marketing materials and ads and what not had AI imagery all throughout. Yes. And the problem with that is that these days, we can tell. And so the backlash online was catastrophic. The comments on the post within the feeds were, I'm never going there because you're as AI. I'm never stepping a foot in there. If you do one thing and like that, you probably do everything I automated. We don't want it. So that's a really good example of using AI in the wrong way. Yeah. We want the brain, especially from a consumer perspective, we want someone to put effort into why we're going to transact with them if they're not putting any effort in. And there is a bias around this. And we feel that if someone's putting in some effort, that is, it's not saying it's exclusive, but they deserve our attention. They deserve us to give them a go. And they've earned that by putting in some effort. It might be 100%. It might be the solution that we're looking for. But the brain likes that because it's relevant and it can relate to it. When we see that it's just generated, we see, are we cheating? Yeah. You're tricking us. It took a shortcut. It took a shortcut. So why should I then go ahead and invest my time, money or whatever into into your business? Is that how you operate your entire business? Is that how you operate yourself in your own life? It's yeah, it absolutely can backfire. And I've seen it more from the not. It's humans by nature. We hate being deceived at every level. So it is that would I see someone trying to deceive me with their marketing? It's like it's a big. I don't want to interact with them at all. Yeah. So what is the consequence gap with AI? You mentioned briefly before AI does not truly understand impact. It doesn't. It's just a it's just a rat repeater. Yeah. I always say it's a one big mathematical equation, which is what is it? The mathematics behind it. You know, X plus Y equals it. Yeah. And this is what the human layer is so important over the top of this because it doesn't understand the difference or hierarchical importance, particularly when we're talking scopes of work or whatever it might be between one piece of the scope of work or another. It just thinks they're all the same. It doesn't know that you have to you know, call this piece out. So I'm going to use myself as an example here when I'm asking it to help me with creating scopes of work. I'll have to constantly say can you dial that up a bit? That's not explained properly. I make that big a change of the site. Like it doesn't get all that. So the human layer again, super critical on all these decision based activities. Yeah. And one thing I often do with my the you know, agent that I've created that is me. I will say you are an expert in behavioral science. I will give it a persona and I will give it a profile. Now I teach profiling and a lot of the work that we do with neuro design is around profiling consumers, profiling certain characteristics and traits for you know, sales conversions for particular products and services for our clients. Unfortunately, the AI doesn't know that. And this is where consequence comes in. So someone will something will impact someone differently depending on what can't what their profile and their personality traits. So it's not what's right in front of us and what we know that we like. It's what we don't know that we like. And AI, artificial intelligence, you know, can't work that out. Human intelligence we can. So even with me prompting the AI engine and giving it a persona, it comes up out with something very good. Still not perfect. Yes. And I've created this environment in a space that I've been trained in for more than 25 years. And still need still need to make yeah, sure. So we're going to talk about some of the environments where this can go wrong now. And there's quite a few right. So the first one is in law. So you know, wrong citations, wrong advice, education, which is really bad, which I spoke about just before, but fabricated sources, shallow thinking, coding, insecure, or just incorrect outputs, medicine, incorrect citations and medical guidelines. Is that some of the things you've seen? Yeah. Yeah. Yeah. So, you know, I mean, we talk about Dr. Pugel and you know, AI, again, you got to remember guys that it is going to the Atlas of information and picking and choosing bits and pieces based on what is there that fits, you know, it's the it's the it's the the curve of best fits and it will give you an output may not be may not be right. And isn't like a large scraping of this data from Reddit, which is basically groupthink, like it's just a large community of people all having an opinion and and and and other platforms like that. Yeah. So the thing is a bit we talk about garbage in garbage out. Yes. It's the same thing. Garbage in garbage out. It's just that if there's a lot more garbage in, then the risk of the garbage coming out is much higher. And we don't know, we're relying on the input of others to get an answer that is suitable for us. Yeah. We're not just relying on the input of us. That's right. And a good test to this is when you ask a question and spit something out, but then you challenge it repeatedly. And you say that's incorrect. It's this and it goes, Oh, good catch. And then you go, No, I was just kidding. It's what you're actually saying. Oh, good catch. And I just keep a green with you. Actually, I did and I posted this on LinkedIn. I actually might repost this. Now everyone's going to go, Where's the post? Where's the post? So I might link it actually in the episode. Yeah. We will do that in the notes. So there was an article that I was reading and I fed it into AI and I said, "Can you give me a summary of this and rank it out of 10?" I believe it's pretty good. So it told you about it. So then it gave me a summary. Summary was reasonable and it said, "This article ranks 8 out of 10." I then actually opened an incognito window and didn't want any bias or history or caching whatever you call it. I then did it again and I said, "Here's this article. Can you please provide me a summary and a score out of 10?" It's not that great. Are you afraid to? I framed it. I framed it. It's actually primed and anchoring. So it wasn't. I didn't say, "I wasn't that good." I don't think it was not as good as it could be. Something like that, I prompted. I'll have to go back to the post, which I'll share with everybody. Any gave a three out of 10? Same article. That's what you wanted to hear. It's what I wanted to hear. Same article. Summary was slightly different. Slightly different. There was a negative bias to it, of course. It was because I primed it and anchored it. So the post was around how we can hook AI around a cognitive bias. And we absolutely can do that. For sure. And just before we went to medicine and stopped, so we'll continue on. Yeah, yeah. So this is important for us to hear. So dating advice. Yeah. So self-esteem and physical appearance. My gosh. That's a real critical area. And self-help psychology. So we're seeing suicidal users that are being gaslit, but also heavy agrients on some of their issues to the point where they're self-harm. Super dangerous. And there's evidence around that. That's happened. Client work. Wrong facts. Wrong tone. Wrong promises. And then globally, when work goes unchecked, you know, reputation or risk. Huge problem. Yep. So let's move on to how to use AI without losing accountability. I think that's pretty important. So the answer isn't banning it. It's building some guardrails. So we've talked a little bit about this, but let's go in a bit more detail. What must a human, even as tools improve? Okay. So the first thing is ethics. So there needs to be some ethical guidelines around whether it's yourself in your own business. You know, what's accepted board isn't certainly within a corporate environment or a large business environment. You've got to have some kind of protocol of how to use AI and some framework because you can go rogue. Yeah. And it's very dangerous. You need to have the final say. But don't let the algorithm bully you. You are the boss of it, not the other way around. You know, it's like, I'm the boss of myself. Yeah. You are. You absolutely are. So it's a helper. Use that information for then your own purposes to construct the, the document or the outcome that you need. So you need, it needs to have that human intelligence, the H.I. Always outweighs the A.I. So yeah, you need to have your, your input in the end. Absolutely. I think there's a non-negotiable. And I think that's super critical in particularly for business when the reputational damage that comes from sending emails that have been by AI, if you had just tried in yourself and I can set the bank canvas great. It comes up with some words fantastic. But if you go in and modify it, check it, send it out with your personal touch and it comes back and there's errors. You can defend that position. You cannot defend that position if it was A.I. It's like, you kind of say, oh, well, I send it out. But if you're saying, excuse me, excuse me, if you say a human error, sorry, I actually made those mistakes. Yeah. The trust remains in the room. Correct. Right? Correct. Yeah. Really important point. We talked about this before and that was, how do you stop A.I. Use from weakening thinking over time? This is a huge problem. Yeah. How do you stop it? AI is a collaborator, it's a partner. It's not a replacement for you. It's not a replacement for a human. There's so much talking. This has been going on for a couple of years now. We were talking about when first we chat GBT came out, which a couple of years ago now. Three or three. Three or four years ago. Yeah. And people said, oh my gosh, I'm going to lose my job to AI. And then there was the threat of AI is going to take over. We're going to be reducing human capital because AI can do it all. I mean, we've covered, you know, pretty, you know, thoroughly in detail today, around the shortcomings and the blind spots of AI and how much human intelligence is needed to provide a correct and significant and important output. What is going to happen, I believe, those that don't understand and you don't have to be an expert, but don't understand how to use AI will potentially be replaced by someone who does understand and does use AI and does understand it. So what's not around the machine is going to replace me. It's someone who understands and can use the machine will replace you. Again, you don't need to be an expert. You just need to have a basic framework and understanding and using your own human intelligence to translate what it's producing and then then you're right. So that's really important to say. And I think it doesn't just apply to AI. That is the same for any skill. If you don't have skill. It's got to upskill on everything, not just AI. So I don't think the playing field is the same. It's just one more thing that you just need to learn to stay on the edge of, you know, the forefront of business. One thing as well here, use it after you've formed a view, right? So create your own view and then blind spot check out. I think that's the most powerful piece of AI that I've found is the blind spot checking because I can give it even screens that I've created in SaaS situations for some of my clients and give it a frame, give it a brief, give it a screen and say, this is what I'm thinking. This is the persona I'm designing for. What do you think? And it will come back and say, "Well, you've missed that." That's not there. And that's not there. And sometimes it's incorrect. And I'll say, "Yeah, but that's not there because of this." And by and large, it'll point out some really key flaws and blind spots. And I think that as a tool in itself is very powerful. As long as you've created a perspective or point of view first. Well, you've created context. So what you're actually doing is helping the mathematical equation. The x plus y equals z. The z is a lot more accurate because you're creating your, but you're responsible, right? So you've created a responsibility around making sure that it facts checks, making sure that it looks at the blind spots. And it will then learn that's what I need to do. You will always check it, but it's guarding it. And rather than it teaching us, we teach it. And then we need to double check it. Yeah, yeah. Perfect. Just before we wrap this up, I've got a bit of a framework that we can talk about just to really easily categorize things into green tasks, amber tasks and red tasks. Just really simply break it down. Right. Green tasks are admin that we spoke about draft summaries formatting, internal ideation. And then you've got amber tasks, which are strategy drafts like client comms, content planning, they require review, right? Very critical. And then the red tasks are things like legal, medical, financial advice, final client recommendations that you don't want to leave the green box, anything where a mistake harms someone, basically, that needs to be critically reviewed by a human. But like I said earlier, I'd say even green tasks, all got to be reviewed, no matter the task. You just can't create and send. It's got to be checked. Get into the habit. Get into the habit of fact checking and just reading over. And the better you are at it, the better the output, the less you will need to spend time doing it, but you still need to spend time checking. Yep, yep, yep. Be responsible. For sure. Here's the shift. AI mirror is fast thinking. Fast thinking is useful. It's also where error is hired. So use AI for speed, not judgment. Keep humans in the loop where consequence exists because trust is harder to rebuild and output is to generate. One caveat today though, AI changes very quickly. These examples are true at the time of recording, but the principle should still hold in years to come, but we will see. Let's think shift. I'm Clint. I'm Leanne. One idea won shift in another 30 minutes. If this changes the way you think, please share this episode with someone who makes decisions for a living. Follow think shift on all the socials and don't forget to subscribe so you don't miss the next shift coming up next week. See you in the next episode. This is a much-made media production.