AI Understanding the Lifelike Tool

Hey everybody. Welcome back to the HVAC Mentors Podcast. This is the series on AI in the trades. And in this episode, we're gonna wanna talk a little bit about why chat AI feels so human. It almost feels a little creepily human at sometimes, sometimes kind of uncanny valley, like it's a real person talking back to you. In the last episode, we did a little bit more defining of what AI actually is and how it's not thinking or understanding, but instead it's recognizing patterns and predicting the right response based on what it believes to be the pattern you're going by. So today we're going to talk about, again, why AI feels so real, especially with these chat apps and these chatbot AI applications. Sometimes it really feels like you're talking to a real person. So part of the reason for that is when we type into our AI app, The response usually feels natural. It feels conversational. Sometimes it feels like the app really gets you, like it's a close friend or somehow it's somebody that you're really connecting with. And when you're using the voice version, it can even notch up that feeling of reality to a greater degree. It can make it feel even realer. So this starts to feel a little bit strange because we know that it's a machine, it's not real. And some people derive this as it's, you know, trying to spy on me or it's trying to take advantage of me or manipulate me. But I think that if we hold on to the core idea behind these AI chat tools, we'll see that that's really not what's happening, but it's because the humans who made the AI chat applications made them to feel human. So that when the user uses them, they don't feel like they're talking about a or talking to a robot or a machine, but instead they're talking to a buddy, a fellow human being. And so that brings us to the question of how did AI learn to sound so human? Why does it sound like I'm talking to an actual person? And the truth is that the AI developers use books, articles, forums, tech manuals, and even conversations between. People in media to map out responses to particular statements. So when somebody says X, the proper response is Y. So it doesn't just learn the proper response for a particular sentence or thing you might ask, but it also learns how questions are asked. Because, you know, we all know that it's, it's not as much what words we use, but how we use them. That communicate. You know, there's a lot in tone, there's a lot in structure. If we, if we, um, if we speak a sentence in a particular order, if we use the words in a particular order, then that can mean one thing. And if we speak the same sentence with the words in another order or different emphasis on another word, that can change the meaning entirely. And this is basically how AI is trained or is written to respond to us as we're interacting with it. It learns to respond to our tone changes, how conversations naturally flow, not even the way that we think that they flow, but the way that they do in reality. So keep in mind too is that a lot of, a lot of our communication comes through, again, our sentence structure, verbal pauses, and the things that we don't consciously necessarily give somebody when we're communicating with them. In fact, a lot of that nonverbal. And if anyone has ever been an avid reader, you might know that an author can communicate a lot of different emotion and different meanings through their words simply by wording a sentence or a paragraph a certain way. And this is where we might talk about the author's tone or the tone that somebody gave or what their implied meaning was, not because of the exact words that were used, but because of the way that they were— the way that they were used. So again, the key thing to remember here is that AI is not thinking about what you're saying to it. It's not making a conclusion based on what it actually knows about you, because again, it doesn't know you. Instead, it's predicting the most human-like response based on the context, based on what it's learned about you and the way that you communicate. It doesn't actually hear you. Like, if you ever ask somebody, hey, you hear me? That doesn't necessarily mean that, yeah, their ears picked up your words and processed them into language in your brain. That means something deeper. Like, did you really hear me? Did you feel me? Did you— did you, like, did your heart react to what I'm saying? And that's the thing that AI begins to learn and has been trained to do. It's trained to take what you're saying to it, but then think more deeply about the words that you're using. But one thing to bear in mind too is that while AI can hear you, so to speak, in terms of like what your deeper meaning is, what it's receiving from you is actually text. So when you use these voice features on AI, it doesn't actually hear you. I remember one time when I was first learning about AI and learning how to use it and I was chattering about something with ChatGPT Voice on my phone and all of a sudden an ambulance and fire truck whizzed by. You know, I stopped talking and I, I remember saying something along the lines of, you know, hey, I hope you didn't have trouble hearing me. There was an ambulance and fire truck in the background. And ChatGPT responded that it can't hear any background noise because everything that I say to it is actually received by it as text. So it's kind of an interesting multi-level way for you and ChatGPT to exchange words back and forth. Because when you're using voice, your phone is actually converting your words to text and then sending that text over to ChatGPT. ChatGPT then calculates a response and then sends that as text to your phone, and then your phone through the application then processes that text into a speech response. So the way I think of it is basically AI is using AI to talk to AI to translate what the human said into AI so that it can be translated from AI back to human, which is to say that it's, it's complex, but it's also profoundly simple. So basically AI is being used to talk to AI so that you can talk to AI and AI can talk to you. But that makes it seem so creepily weird, or rather so creepily real, almost this kind of uncanny valley sort of a feeling toward using the chat AI features. And the reason that it feels so real is that it intentionally emphasizes certain words. It has the ability to shift tones. It can pause. I've even heard it use breath sounds and You know, increasing the cadence or decreasing the cadence of speech during certain words or parts of a sentence. But, but other times it can sound very flat and neutral, like it's, it's just being very like factual and almost robotic. And there can be many reasons for that. Anything from bandwidth to it feels that you're straying off topic and it's trying to actually help you to maybe have an on-ramp back to what you were talking about to begin with. The thing that we need to remember again is that this isn't AI responding to us the way that I would respond to you or you to me in a conversation. It's pattern recognition. The AI is simply responding to your prompts based on patterns that are written within its computer code, and it's able to re— or rather, to produce a response that makes sense to you. So while voice chat AI seems like you're having a conversation with a close friend, the truth is that the conversation is really just text being shared back and forth and the most likely response to a stimulus. It's not actually a conversation in the truest sense. It feels like one, but the truth is that it's multiple computer systems, each doing a job, all based on patterns. And really, at the end of the day, what it is is a really well-orchestrated illusion meant not to deceive you, but meant to communicate with you, to be useful to you. Which brings me to my second point. You know, why, why does the AI chat stuff feel so real? Well, part of it is that it's the humanity of the humans who wrote the code and developed the system bubbling up to the surface. The other part of that is that it's meant to be used by people, so the people who made it want it to be comfortable to use by people. If it was too robotic, think about it, it would be uncomfortable, it'd be unnatural, it'd be hard to use, and it probably wouldn't be as effective. So as we kind of wrap up, I want the, the real point I want to drive home is that conversational AI isn't in any way meant to deceive you, but it's meant to seem user-friendly, just like how the user interface on your home PC or your iPhone or any device is really designed to be as user-friendly as possible. In other words, it's, you know, as weird as it is to say, it's a consumer good. They want the consumers who will use it to have ease in using it. It's designed that way on purpose so that you'll be comfortable using it. And, you know, the thing that we really need to keep in mind here is that AI may sound human, but it's not human. It may sound like it's conscious, but it's not conscious. It may, it may sound like it has an agenda or is trying to steer you in a way, but it is not able to do that. It's just simply code on a computer screen or in a computer. Unlike you, AI doesn't have experiences or awareness. It doesn't have intuition or judgment. You know, when I'm on a job, if I'm using AI and I'm, you know, trying to track service notes or maybe looking for a manual or a schematic or something, you know, it, it has thrown suggestions or ideas at me. But the truth is that a lot of times I disregard that. Sometimes I might say, yeah, you know what, I'm gonna try that. Maybe, maybe we'll test this or that or the other thing. But because of my field knowledge and my experience, I know that I can allow AI to make suggestions to help, but ultimately I'm the one with the awareness. I'm the one with the intuition, the judgment, the experience, and I'm the one who is running my service call. So I'm not in any way going to allow AI to do the job for me because I'm the technician. But I think understanding that difference, that it's going to help me to reinforce my skills and my knowledge as a technician. AI is actually a very handy tool. And also, yes, for those of us who tend to ramble in our service notes, it can be really good at helping us to summarize, which is primarily what I use it for. AI is simply pattern-based output. That's all that it is. These chatbots that you might use and see around, they're not, they don't have an agenda. They're not trying to manipulate or to trick you in any way or take, you know, your job from you. They're just another tool like any tool, and in the right context, The right tool in the knowledgeable hands can go a long way, but ultimately it's up to you to learn how to use the tool properly. So when you're out there, use your hands, use your nose, use your, your eyes. I mean, sometimes we do even use our sense of taste to do your job and to constantly hone your skills and ply your trade. But remember that AI can't See what you see. It can't hear. It can't feel a suction line, and it definitely can't understand real-world conditions. So, just like any tool, it's there to support the technician or the tradesman or tradeswoman. Sorry, but it doesn't in any way replace us. So, as we close, I just want to make you aware of our a couple of other awesome things that are coming down the pipeline. One is an interview with with a well-known. AI marketer who's just making all kinds of headway into using AI for businesses. And I realize this channel is primarily for beginners, but you know what? One day some of you, or I hope many of you, will be business owners. And you know, maybe something that will be said from her will actually be helpful, but she's going to be talking about AI. The other cool thing that's coming down the pipeline that we need to unpack is an interview with ChatGPT itself, not the CEO of OpenAI or the developers of ChatGPT or an engineer from ChatGPT, but instead my actual chatbot. I'm going to ask it questions about what I use it for, how we work together, and what are, you know, what discoveries it's seen me make as I've learned about AI and learned how to use it appropriately on the job. You know, ultimately it is designed to feel human to make it easier to use, and that's the only reason. So next step episode, we're going to be talking about where people go wrong using AI, some of the common mistakes that people make, and kind of some of the ways that people have sort of strayed in the proper use of the technology. Again, you know, with any tool, you can use it the wrong way, and I think all of us have discovered moments where we weren't using our tools in the most appropriate manner. So be sure to tune in for that. Until then, guys, thanks so much for listening. It's always a pleasure to hang with you. I'll look forward to seeing you on the next one.