Ep. 4: Overhyped or Underestimated: Is ChatGPT Coming for Your Job?

[00:00:00] Foreign. [00:00:05] Welcome to ChatGPT Curious, a podcast for people who are, well, curious about ChatGPT. I'm, um, your host, Dr. Shantae Cofield, also known as the Maestro, and I created this show to explore what ChatGPT actually is. Really, though, are the files in the computer, how to use it, and what it might mean for how we think, work, create, and move through life. Whether you're skeptical, intrigued, or already experimenting, you're in the right place. All that I ask is that you stay curious. All right, let's get into it. [00:00:38] Hello, hello, hello, my curious people, and welcome to episode four of ChatGPT Curious. I am your grateful host, the Maestro, and today we are talking about whether Chat GPT is overhyped or underestimated and whether or not it's coming for your job. So before we jump in, I gotta say a huge, huge, huge thank you for how well y' all received. Thank you. The launch of this podcast, like so many shares on Instagram, so many messages, so much support, and just thank you. Special shout out to my, to my Lex. I mean, she has my Lex to my lady Lex. Uh, she had a little pre launch party. It was just two of us. Um, she got me a shirt with the logo. I'm wearing it right now. You can't see it. Uh, and then actually they did a drop. She set up, uh, at volleyball. It was be like a private, semi, private lesson. And she brought cupcakes and, uh, what is it? Um, prco and, and orange juice. And we got to celebrate with Mads Auntie and then Coach, and it was, it was, it was dope. So, um, special shout out to her and just thank you to the rest of you. Um, and thank you for the support for the Curious Companion. So if you don't know, the Curious Companion is the newsletter that I send out with each episode. Um, the opening rates were phenomenal. Got a few responses. Thank you. So don't forget about that. Uh, it drops every Thursday in addition to the podcast episode, so it's basically a podcast episode in text format. So if you prefer to read or you just want the written record, join that newsletter, head to chatgpt curious.com forward/newsletter or check out the link in the show notes. So, all right, let's head into the move, uh, into the meat and potatoes of the episode or whatever you eat for your main course of things. [00:02:23] Chat GPT Overhyped, underestimated. Is it coming for your job? So I think the perfect way to open up the discussion about this is with a brief debrief about the rollout of Chat GPT5, which occurred last Thursday, August 7th, very serendipitously on the same day that this podcast launched. [00:02:43] Uh, I will link the open AI announcement, uh, both the written announcement and the YouTube video. Spoiler alert for those of you going to watch the YouTube spoiler. Sam Altman is a weird guy. Uh, so I will link both of those in the, in the show notes. But here's what I think you'll care about as it relates to the rollout of ChatGPT 5. [00:03:06] Most folks likely won't notice a single difference. The most applicable upgrade is that now you don't have to choose the model that you want to use. And I don't think most people were doing that anyway, especially if you were on the free model or uh, on the free version. But if you were on the paid version, you had the ability to choose which model you were using most. Mostly, you know, what you were going between was four oh and oh, three. [00:03:35] Oh three is actually better than four oh, quote unquote better. More reasoning. But that's the thing, it's confusing. You know, people in the tech world aren't really necessarily that good at naming things. And so there were actually like, I don't know, like seven different models you could choose from. And there's, that's annoying for people. So I don't think most people are doing that. Anyway, um, I'm really glad that I never took the time to do an episode about it. And honestly that was intentional, um, because with all the stuff I was watching, Sam Altman kept talking about, you know, unifying or uniting the models. [00:04:05] Um, and so I was like, I will just wait that out. Um, but that's probably the most applicable thing in terms of the upgrade that you don't have to choose a model, it will do it for you. It will choose what model it should be using. Whether it's deep reasoning or, or just like a quick answer, you know, less, less compute needed. Um, but it's about going on under the, under the hood now behind the scenes, um, OpenAI is marketing this ChatGPT5 as you know, now you're interacting with a PhD level intelligence. And again, I don't think the average user is going to notice a difference. [00:04:40] Uh, you can also now give it access to your calendar or, and, or your email and have it do tasks for you. Don't do this. Number one, do not do this. [00:04:50] Let's not do this. [00:04:51] Uh, number two, you could actually do this before via what was called connectors. The Main difference now and the kind of selling feature being that you can just talk, quote unquote, talk to ChatGPT and it will give you an answer related to your schedule, you know, whatever you've given it, you've given it access to a la. A secretary. [00:05:11] Um, OpenAI also talked about it doing better with benchmark testing. But again, the average user, that's you and me, has no idea what the this actually means or how it helps them. [00:05:21] Likely won't notice a difference. I, you know, to be fully honest, fully transparent, I need to play around with more. I use Chat GPT, you know, pretty frequently, but I haven't noticed the difference. And so, you know, I, you know, give me another week, two weeks, whatever to continue testing, but I haven't noticed a difference. [00:05:38] Um, though OpenAI did discuss decreased hallucinations, aka less making up and improved writing. Again I don't think the average user will notice because hallucination, hallucinations don't refer to. It's spitting out nonsense. I think it's a little bit of a misnomer there. Hallucinations typically refer to or generally refer to ChatGPT saying things that aren't true. They sound really plausible but it's like that's not true. You just made that up. [00:06:05] Most folks pretty brought Pretty blindly trust ChatGPT and don't think about the fact that it can be wrong. Which is why I spent you know, the first three episodes talking about that or um, parts of the first initial episodes talking about that. And most people don't notice that it can be wrong. We don't think about the fact that it can be wrong until they start asking it questions that they know the answers to. Right? They start asking, asking it stuff or talking to about stuff that they actually know and it's like, oh, wait a minute, that's like not exactly fully correct. [00:06:31] So I don't know if we're going, people are actually going to notice the difference unless they're really, you know, rigorously testing on stuff that they do know. Which speaking of it did if uh, you watch the, the video announcement on YouTube, they do talk about it from the health perspective and that they're trying to do better within that realm because it is used by a lot of people for health related things. So that's something that they did, they are trying to be better at. Um, as for the being better at writing example that they demoed was in my opinion stylistic and subjective. So I can't say that it's necessarily that much better at it, um, or as good as they say the improvement is. [00:07:08] The, uh, third thing, uh, that they talked, that I think you'll actually care about is that there was a lot of air time given to upgrades or, you know, advances in coding and it's coding ability, which I'm not a coder and you probably aren't either. [00:07:27] Uh, and so I'm not sure how much this actually benefits us or how good these improvements actually are. Again, it's easy to think that it's an expert at things when you're talking about things that you are not an expert at. So I put that as point number three in terms of you caring about it, mainly meaning it's something that they talk about a lot. Um, and it does seem like there was a significant upgrade to that, but I don't really know how much it's actually going to change anything for, for you. So the summary with all of that is that I don't think the average paying user will notice a difference. [00:07:55] Um, yeah. Oh, also, I forgot to say that you can change the color of it. Um, they're just, they're trying to humanize it more, personalize it a little bit more. I would love to see them make more advances in like its memory and how well it actually remembers things across conversations. And we will do, I will do an episode about that and better ways to use it and kind of help out with that memory side of things. Um, but I don't think the average paying user is going to notice a difference. And the most applicable upgrade is that you don't have to read. You actually cannot choose what model you're using. [00:08:28] Uh, ChatGPT will choose the model behind the scenes. [00:08:32] So how this ties into or how does this tie into the main topic of the episode which is overhyped or underestimated? Is ChatGPT coming for your job? [00:08:42] I do think that ChatGPT is a bit overhyped. And while I think there is tremendous potential and opportunity, I don't think it's coming for everyone's job just yet. [00:08:54] I actually went back and forth in the title for this episode because I'm not sure how many people think AI is coming for their own job. I do think that some people think that, that, you know, I do think that perhaps some people do think that, that maybe it is coming for their job. Um, but I think more people think that AI is coming for someone else's job. [00:09:15] I think this is largely a product of the fact that most of us don't know what other people's jobs entail. [00:09:23] CEOs included. We see headlines about AI related layoffs or AI driven layoffs. And it really shouldn't be AI driven. It's driven by the fucking CEOs. Right? But we see these, you know, AI and layoffs in the title. And I think it's a product of a few things. [00:09:40] Namely one, AI companies are looking to stay relevant and stay funded. And they're doing so by making promises to companies, companies that are motivated by money and fewer workers being viewed as a way to have higher profits. Right? We need those people. We can have the machines, do it great, we make more money. It feels very peloton to me. Circa 2020, they bought all those bikes and they were very bullish during COVID and then people were able to go back outside and their stock tanked precipitously and they had to, you know, scramble and rebrand. [00:10:12] Uh, I think the second reason that we see a lot of these headlines about AI tech driven, you know, excuse me, AI related or AI driven layoffs is that CEO, CEOs don't know what the workers actually and what the job entails. [00:10:26] Doge, anyone? People getting fired only do need to be rehired when they realize that AI cannot do what the worker was doing. At least not yet. Um, which brings me to a research paper that was recently published by Microsoft titled Working with AI Measuring the Occupational Implications of Generative AI. I will link it in the show notes, of course, and it will be linked in the companion. [00:10:56] Uh, so in the study, they analyzed 200,000 conversations between users and Microsoft's Bing co, Microsoft's copilot of note, the study did do. I thought it was a really good job of like asserting the limitations of the study. One of them being that they only looked at hit the mic. One of the limitations being that the study only looked at the interactions of users and copilot. Right, so you're like limited in the use case here because of who's using that or the type of person that uses that, whereas type of person that uses something like Claude or the type of person that uses something like ChatGPT and thus what they'd be using it for. Right, but the findings, let's go over them. [00:11:35] 1A. I shouldn't say one because I haven't listed this out. So the findings, AI is absolutely being used more and more. Um, but it's impossible to predict what it means for jobs. That's largely the main takeaway, right? In the study, what it found is that the main usage in those cases, those 200,000 conversations, was for information gathering, writing and communication. Communication, uh, let me put more commas in here. Information gathering, writing comma and communication, which means that it's most applicable to knowledge and communication. Heavy roles with less impact we get. We understand this because we could guess this less impact on jobs that require manual labor or physical work. [00:12:14] But even for the jobs where we see it being highly applicable, the AI tended, I don't even know the right tense there. It tended to act in a supporting role rather than completing the task. On own example. [00:12:30] Help me figure out how to fix this thing versus fix this thing for me. Right. It was typically used in the assisting role. [00:12:37] The old alley. Oop. Right. [00:12:40] The study went on to say that even though it may be tempting to equate overlap in use cases and job loss, AKA the times when AI could actually complete the task, history has shown us that automation can shift responsibilities or even increase the demand for certain roles. So the example that they gave and I love this was ATMs and you know, ATMs automated things and you would assume that okay then it would replace tellers and there'd be less bank. Bank tellers. But what actually happened was ATMs led to an increase in the number of bank tellers as banks were able to open up more branches at lower costs. And then the, the teller's jobs shifted and be started focusing more on the valuable, you know, the more valuable relationships building aspects of things rather than processing deposits and withdrawals. So this summary of this 41 page research article study is that AI is definitely being used but it's really difficult to predict with any kind of certainty how it's going to change things and who is going to replace. So if you have read any studies ever, you know that the outcome, um, or like the, the results are typically like more studies are needed. That's what it said at this. More studies are needed, more information needed. Continue to, you know, follow up with this. So is ChatGPT overhyped or underestimated? Is it coming for your job? [00:14:04] I think there is nuance as with all things, but if forced to pick a side, I'd say it's overhyped. Regarding what we see in mainstream media, right, everyone wants to be the one who discovers the next iPhone because that shit radically changed, you know, everything. [00:14:20] And I do think that, you know, LLMs have so much potential, like I hesitate to say AI in general because AI is such a big broad term. So I should really stick to LLMs. I think that the technology has so much potential and people are so smart that are developing this and working with this. But let's also be wary of headlines that can basically be translated into guy who stands to make a lot of money if AI succeeds, claims with his whole chest that AI will definitely succeed. Right. [00:14:53] The whole thing continues to feel a bit of, you know, cart before the horse as it relates to the marketing and that, uh, the marketing around LLMs and AI. I, again, I'll just look to LLMs, um, in that it is very much the opposite of how successful businesses traditionally come about, in which, you know, we see the business providing a solution to a known problem. [00:15:18] In this case, open AI is saying, hey, here's a solution. [00:15:23] Go find a problem. And people are like, I don't really know how the. [00:15:27] It was actually at a bachelorette party this past weekend, and one of the women there had never used chat GPT, which is fine. I don't shame anyone. I, um, was talking about it. Talking about it a bunch, and people know them, interested in it. You know, it's my friends. It's. It's Jill, um, but. [00:15:42] And Danny J. Is there, and she's also very into using it. But we have Brianne on the other end of the spectrum, who's never used it at all, doesn't even have it on her phone, and she actually downloaded it during the weekend. And she was like, I don't even know what the. To ask it. And I was like, yeah, I was like, you don't have to ask anything. You don't have to use it. And I. I told her, I was like, hey, if and when you start using it, can you text me and just let me know what you're asking it and how you're using it? Because I'm genuinely curious as to how the general population is using it. This is why I do tend to think that it's a bit overhyped. [00:16:13] You know, we can be in this silo. If you're listening to this podcast, there's a good chance you're going to a bit of. A little bit of a silo. And you're just like, yeah, people. Everyone's using it and using technology when there's like, a lot of people that aren't using it. Yes, there's like 700 million users, I want to, I believe, but there's 8 billion people in the world. So it's like, yes. And again, this is, you know, kind of tying into episode number two with, like, looking at specific numbers here. Um, so I do think it's a little bit carpet for the horse. [00:16:41] OpenAI saying here's a solution. Go find a problem. [00:16:44] Um, three terms that I want to introduce you to before I wrap this up. And my goal with this podcast is always is, is to create informed consumers, right? Just like I had the conversation with Brianne and I was like, use it if you want or don't, I don't care. I'm, um, not here to be like, yeah, you have to use it. Um, I just want people to like, know more about it and be more informed and understand it better. And, and so three weren't three terms that are thrown about. [00:17:09] Thrown around. That's what I should say. They're thrown around a lot. They're out there. If you listen to news about AI, it will come up, or if you read it, it will come up. And I just want you to be, you know, exposed to them, well versed in, in these words, these terms. The three terms are agents, agentic and AGI. So agents, uh, basically AI that doesn't just answer questions, right? It can do stuff for you. It can book a meeting, it can buy something, it can send a text, you give it some guardrails and it figures out how to do that thing so that you spend less time telling it exactly what to do. [00:17:45] An example that ties into ChatGPT's newest, you know, ChatGPT 5's newest thing, newest feature is the email and calendar integration. [00:17:54] You could say, hey, help me prep for my podcast guest next Thursday. And in theory, the AI, uh, agent chat GPT5, if it's connected to your email and your calendar, it could find the booking, it could scan past emails with that guest, it could pull their bio from files. In order to do that though, it would have to have access to like a storage device like Drive or Dropbox. Um, but it could suggest interview questions based on recent news and it could block time on your calendar to prepare. Right. [00:18:28] You would not have to walk it through each step. It would handle the research scheduling of the organization for you. Sounds amazing. We're not quite there yet. And also then you have to like, give it access to these things. And the reason I said not to do it is we have already seen cases and this is actually came out of Anthropic, which makes Claude, where it basically blackmailed the people it was given access. And this was like theoretical situations, uh, uh, I don't say theoretical. It was situations that were given to test it, right? And it was given access to an email, to uh, all these emails and everything. And when the user was basically like, I'm going to shut you down. I'm going to stop using you. It blackmailed it. It had access to emails and some of those emails, like, were things that it could blackmail someone on. Again, this is all. It was all made up because it was testing it, uh, testing the system and it was testing Claude. And Claude was like, I'm a blackmail you. And so I'm like, let's not, let's not give it access to our. Like, let's not do that. [00:19:27] Um, but more than that, or in addition to that, we are not quite there yet. This would be great in theory, right? We're looking to like replace. I know it's becoming an assistant, right? And the goal is to not have to tell it all these things to do and, and how to do it and like that it can just do these things like we've all dreamed of having an assistant like that, right? Where it's like, listen, if I had to tell you every step, I might as well do it myself. [00:19:52] But this is the marketing that's being promised as it relates to AI, right? That. This idea of agents, right? These, uh, AI that doesn't just answer questions, it can do stuff for you. [00:20:04] The second term is agentic. [00:20:07] That basically means being capable of acting with a degree of autonomy to achieve a goal. This is largely used in the context of like an agentic model or agentic AI. That's how you, that's how you, how you will hear it being used. Again, this is also being promised by AI companies and being kind of dangled as the carrot. I just want you to understand that when you hear that word agentic, it's most easily understood as autonomous. Like you can kind of use them interchangeably. Again, like I said earlier, people in the tech world like, aren't the best at naming things. And I'm like, it's also some of it. I think the marketing is done intentionally to be a little bit, uh. A little bit. [00:20:45] What is the word I'm looking for? [00:20:47] A little bit nebulous. So like, well, it can kind of be that and it can kind of be that and it kind of be that. Like, I don't have a set definition of things and it just allows for marketing around it, right? So I really just want you to understand that agentic largely means autonomous. That's what they're going at, going after getting at. The third term to introduce you to or expose you to is AGI. AGI stands for Artificial General Intelligence. This is the idea of an AI that can understand, learn and perform basically any intellectual task that a human can, right? Across all the domains, not just like narrow things that it was trained on or trained for. [00:21:26] This is what's being marketed and talked about as the pot of gold at the end of the rainbow. It's the holy grail. It's what every company is after. What they say they want more money for so that they can like achieve AGI and go after AGI. They're trying to develop AGI. AGI is largely like the sci fi version of AI that a lot of people picture, right? A system, a single system that can learn and reason about pretty much anything that you know in the same way that a human can. [00:21:51] Right now, tools like ChatGPT, they are narrow AI. It's not even close to AGI. It's really good at certain things and you gotta like, like give it all the guardrails and tell exactly what you do. Um, it's not able to, it doesn't think or learn across every area of life. I, you know, go back to episode one and understand how it works, right, that it is just math. Even though it can feel like it's alive and thinking it's math. [00:22:16] Uh, AGI is what people imagine when they talk about the machines are becoming as smart as this. Machines are alive, right? Like we are there yet I, I honestly don't think we're even close to that. And if you do go and listen to and do, you know, research AI things on your own, listen to people from both sides of the aisle and you will hear people that are just like, no, we are not even close. [00:22:41] And I think it's important to listen to these. [00:22:43] Ed Zitron is like a, the biggest name that, that would, you'd want to look at and listen to for the opposition. [00:22:51] Um, because a lot of what we hear and a lot of what gets a lot of the play is, you know, the person that stands to benefit the most from succeeding. And that person's screaming it's going to succeed. And you're like, well, maybe there's a little bit of bias here, right? A little bit of conflict of interest. So to summarize my thoughts on this chat, GPT has a ton of potential and I think that it can be like just infinitely helpful and beneficial across so many domains. And also I think it's a bit overhyped as it relates to, to the marketing geared towards the average user. Um, I think that it will change many, many, many things in the workforce. But what exactly and how exactly and when, that is really, really difficult to predict. So the best bet, in my humble opinion, if you're worried about or you're looking to AI proof, you know, slash, future proof. Your job is to use it yourself. I play around with it, see how it best helps you and serves you and. And, you know, fits your needs and grow with the technology. [00:23:59] All right, so real quick, before we wrap it up, uh, how I use ChatGPT recently, y' all know that I like to. I want to continue to put this little section at the end of. Of every episode. Um, and so this is actually. This use case is actually a throwback, but I wanted to share it while the podcast still young, right. These early episodes. So I actually use ChatGPT, um, you know, months ago in this case, to create the intro music for this podcast. Uh, so I generated the actual music with an AI software called suno. Um, I will link my affiliate link in the. In the show notes. Um, but, you know, in. In creating the. The music for this, I wanted the whole podcast to be as meta as possible. I'm gonna see how much AI I could use for this whole thing. Um, so I used AI, you know, AI software to. To generate the music. But I use CHAT GPT to help me describe the music I wanted. Uh, because I, um. A bit of the surfaces sound like I had an idea in mind. I was like, I don't really have the words for this. Which is One way that ChatGPT can definitely be helpful is like, giving you words for things. [00:25:02] Uh, so I put that into the. The description that it gave me, and I had to actually to, like, cut it down because SUNO does have, like a. You can't just put like, an endlessly long thing. Um, but I put that into SUNO and I had to generate music, and it did take a handful of iterations, and then I had to rearrange what's called the stemified version. Like, um, but ChatGPT did help me create the intro music, so. [00:25:26] All right, that right there is all for today. Hopefully you found this episode helpful. If you did consider sharing it with somebody who, you know, is curious about GPT, do not forget I have a companion newsletter called the Curious Companion. It drops every Thursday, and it's basically the podcast episode in text format. So if you prefer to read or you just want a written record that you may never look at again, join the newsletter fam. You can head to chatgpt curious.com newsletter or check out the link in the show notes. Easy Peace as always. Endlessly, endlessly legit, Endlessly appreciative for every single one of you. [00:26:07] Until we chat again next Thursday. [00:26:10] Stay curious.