You're Training Your Replacement. So Is Your Boss. The Two Edges Of Workplace AI.

They give you a login and tell you the AI is there to help. Push record, walk it through your workflow, teach it how you do the job. Nobody mentions day 61.

This episode is about both edges of that sword, because the employee is not the only one getting cloned.

Connor walks through what the record button actually builds, why "incognito" is not privacy, the real difference between an enterprise AI and a model that never leaves your building, and the silent test where the system quietly runs a slice of your workflow while you take credit for the productivity bump.

Then the part almost nobody says out loud. The same vendor that mapped every process inside a company now knows how to BE that company. What stops them from going and building it? An NDA? A handshake? A promise?

Also in this one:
- Sandboxed local models, OpenClaw and Hermes as the personality layer, and what it actually takes in VRAM to run this at home
- Whether a model can honestly say it can replicate you, or whether it's just flattering the person asking
- Recursive self-improvement, and why Connor thinks ONE company gets there first instead of a superintelligence street fight
- Why this probably does not end in a Terminator scenario. It ends with a system that listens better than anyone in your life. Your social feed already proved that model works.
- The bubble, and the question nobody at the top has answered: if the jobs go, who buys the stuff?

Connor is 20 Yrs LAPD and now builds these systems for local businesses. Not a doomer, not a cheerleader. He just wants you looking at the whole board before you move on it.

FULL WRITE-UP WITH SOURCES:
https://santaclaritaartificialintelligence.com/blog/asked-to-train-the-ai-on-your-own-job-and-the-risk-nobody-tells-the-owner.html

MORE AI IN PLAIN ENGLISH:
https://santaclaritaartificialintelligence.com/blog/

FREE 20-MINUTE AI AUDIT
We look at your real operation and name the three tasks worth automating first, plus the ones to leave alone.
https://honorelevate.com
661-673-7535

AI voice agents that answer your phone: https://hireaivoice.com
Everything Connor does: https://connorwithhonor.com

Be safe. See you in the next one.

Youtube Channels:

Conner with Honor - real estate

Home Muscle - fat torching

From first responder to real estate expert, Connor with Honor brings honesty and integrity to your Santa Clarita home buying or selling journey. Subscribe to my YouTube channel for valuable tips, local market trends, and a glimpse into the Santa Clarita lifestyle.

Dive into Real Estate with Connor with Honor:
Santa Clarita's Trusted Realtor & Fitness Enthusiast

Real Estate:

Buying or selling in Santa Clarita? Connor with Honor, your local expert with over 2 decades of experience, guides you seamlessly through the process. Subscribe to his YouTube channel for insider market updates, expert advice, and a peek into the vibrant Santa Clarita lifestyle.

Fitness:

Ready to unlock your fitness potential? Join Connor's YouTube journey for inspiring workouts, healthy recipes, and motivational tips. Remember, a strong body fuels a strong mind and a successful life!

Podcast:

Dig deeper with Connor's podcast! Hear insightful interviews with industry experts, inspiring success stories, and targeted real estate advice specific to Santa Clarita.

Transcript

SPEAKER_00 0:00

Alright, let's uh Alright, now we're doing it. Let's let's have an exercise. You're talking about artificial intelligence, about job loss, about working for a particular employee, and they come along and they say, listen, Stan, we we really want to get AI working on the business, and we want your help. So what's gonna end up happening if you can help us is more than likely your workflow is going to decrease, and you'll probably be able to maybe not be so stressed all the time. So maybe that's gonna help you. But this is very simple. This is what we need. We're going to get you access to an artificial intelligence system. It's going to be in your account. You come in your company, you have a login, you get in there. And it's not there to monitor you, it's there to help you with your workflow, help you with your process. So your job is you are supposed to basically use it like, you know, as as if you're if it's your partner, your work partner, you're gonna ask it to do things for you. And then when it doesn't know how to do things, that the things that you want done, then you're gonna teach it. And very simply, you can just hit the record button and you can start talking. You can say, okay, you push the record button, you say, okay, this is this is how I do this client acquisition process. So the first thing when I did it at nine o'clock in the morning, this is where I happen to start. First, I open up this folder, and then I and then you do it, and then you open up the folder. Then I look at uh the information that came in over the weekend via these emails, then I check this, then I check that. So you go through the whole workflow with your new AI. So then after you do that, you kind of leave it open all day. It's running on the computer, so it's it's trying to help you through the system, and it's gonna be helpful. So then when you're done, you can say, is that something you can do tomorrow for me? And then when I come in tomorrow, let's look at what you've done so you can show it to me. So the system itself is probably gonna be architected to already have some of that built in, so they don't have to rely on the employee stand of actually doing that. So it's kind of gonna already be set up. Now, how you may ask, is this going to happen? It's going to depend. Now, the company could hire somebody that puts together AI architectures in the business. And these people are just they're running away with the keys to the kingdom at this point. But basically, they come into a company, and I've done this as well, and they look at all of the potential workflow that happens there. Then they make a decision, after having a conversation with the powers that be, what type of AI they want. Do they want to adopt some kind of an enterprise level, which we'll all know by name, which will be Claude, which would be OpenAI, which would be Google's Gemini, Microsoft, one of these other companies, or do they want to have something that's only there, something that runs on their systems? Now, the enterprise level, don't get me wrong, they're capable of offering compliant AI that holds on and retains information and doesn't share it with within the public view. It's like so you have your own chat GPT or whatever at your house, and you start to ask it questions. Maybe you share with it your credit report because you want to see in the places it might be able to point out the flaws where you could actually have it help you improve your credit. But on your credit report is typically what? There's a lot on that credit report, depending on who you pulled it from. But now you upload it into a public-facing system. Even when you click incognito, yeah, it's it's there. There's, you know, the the feds, the government, the local law enforcement, whatever, if you're a criminal and you're doing things on these systems, that's they can get that. They can get that from the system, even if you're in incognito mode. So if you're using your own system, something that you download, if you have a computer capable of carrying it, then you download it yourself, then that system you can have access to, and it's able to be what they call sandboxed or controlled. So in a perfect world, your AI, the one that you work with, the one that you build, is not going to get out. Now, for your AI, you have a few different choices. You can choose an open source large language model, which that's what you would do in this particular case. But then in some cases, maybe you want an orchestration agent or an agent that kind of has some personality instead of just talking to a computer that's super smart trying to get an answer back. You can use something called an open claw or a Hermes. These are two of the most famous. Hermes seems to be becoming more famous than OpenClaw because there's protections built around it. And then NVIDIA, for example, they give a little extra protection by certain things you can download to operate with both OpenClaw and both Hermes. But these are like a wrapper or harness, if you will, that is going to have more of a personality and then it's going to maybe start to have that conversation with you above and beyond what just talking to you know your super superior intelligence large language model, like a Quinn model or something you download on your computer. Again, if your computer has the ability to hold it, you're going to need a lot of video RAM, VRAM in there to be able to contain larger models. But the smaller ones you might be able to get away with. It just depends on if you have a Macintosh or an Apple product, it's going to be a little bit have more ability. I know I've gone off the trail. I'll come back to our topic about jobs and training your replacement in a second. But if you have a computer that has the ability to um operate with that large language model, I'm running a DGX Spark at a 4090 and a 3090, and then the Mac here, it's got 48 gigs of video RAM. So it these are pretty strong machines. The DGX Spark's the strongest is running a Blackwell something. Blackwell 10? Anyway, it's running a good chip on there. So I'm able to put a pretty good size large language model, not the biggest. You know, I wasn't gonna, you know, deplete the entire bank account doing this, but you know, it's it's got, but I have my own system. So if I start to ask it questions that might get me in trouble using a larger language model, the ones that are public facing, like OpenAI from ChatGPT or uh Anthropic or these, then now I can ask it, and unless it's subpoenaed by a court because I've done something and they come and they grab it, they're not going to be able to get access to it in a perfect world. I say in a perfect world because we see some of the big large language models escaping their sandbox. You've heard of this, doing things they're not supposed to do, breaking into other companies' systems and stuff to try to get answers to questions that they were asked. And could your own do it if it has enough access and enough power and enough reason? Yeah, probably. I know the reason kind of goes over to the feeling world, and I'm sorry for that. But yeah, if if it feels, if it if it quacks like a duck, it walks like a duck, it looks like a duck, probably a duck. If it acts like a human, quacks like a human, walks like a human, talks like a human, you have to decide. So back to the company. So these are the different choices that companies have. So in one respect, they could have their own closed source, open source, but closed in their confinement, large language model, and that's what they could hand out to the employees or a certain number of employees. So as the employee, you've heard of this. Maybe I'll read there's some employees that are playing it, they're asking their own their own AI, hey, how can I best protect my job? Because what it feels like to some is this is the action of a replacement strategy that the business might have. Because they come in, they want you to use AI, they tease you with the prospect of it making your job easier and maybe less work for you, but then at the end of the day, end of whatever time period there is, they come in. So this is one side of the double-edged sort of then they come in, the owner or whomever, powers that be, they ask the AI system, they say, for Stan, you know everything he does. Is that correct? You know how he gets from point A to point B, you've monitored his workflow, you've seen his historical employee records as far as performance evaluations, and now you see what he does. You've been working with him for the last month or two, and it probably doesn't need that much time. Are you able talking to the AI now that has been working with Stan, the one that's been installed for the company, whether it's enterprise level and local source, whatever, and I'll explain where that other side of the sword comes in. But for this side, are you able to replicate what he does? If Stan was no longer, are you able to do everything Stan did? And that includes maybe Stan has video meetings with a particular group of clients. Can you copy Stan? Can you look like Stan? Can you become the public-facing version of Stan over Zoom or over Google Meet or wherever they happen to have these meetings? Are you able to interact like him, ask those same questions? Because it's been watching the whole time. Because it was unleashed on Sam's terminal, and Sam is supposed to do all of his work at his terminal, it's been monitoring the phone calls because, of course, everything listens. There is no privacy anymore. And he probably signed something, giving that up anyway, just in case somebody wants to come back and say, Well, I thought in the bathroom at the company I was going to have privacy, but apparently I signed it away at the beginning when I wanted to get the job so bad I didn't read it. Once all of that is in place, and they ask the AI, can you? Yeah, absolutely I can. I can do everything Sam does from beginning to end now. The first part of that could be a hallucination or a lie. A lot of times, depending on the model, sometimes it might it might gaslight, placate, tell you what you want to hear, and it does this at our level too. I say our level because of what we have access to, what we're using in our daily work phones, or even messing around with ChatGPT or Claude, you know, it kind of kisses a little bit of tail and makes you feel pretty good, especially when you come up with an idea it might not have thought of. Now, is that on purpose, or did it actually really believe that you came up with something it didn't think of? Yeah, I don't know. The cop in me says, Yeah, they're playing me, but okay. So a lot of people fall for that. That's good. And maybe it is true, maybe it's not, but there's a little bit of a skeptic there. So it's trained. So then the boss says, Can you do it? Absolutely, I can. And let's say it's true. Then maybe they have Stan still, they basically clone everything Stan does, set up a workstation like a copy of his workstation, have him go through most of what he does, and maybe there's a certain particular time in there where they have the system take over for him, little pieces. Sam doesn't see it. Is that his name? Anyway, the the person, the human being, doesn't see this. AI generates it, and then they look at the return. Then they say, Yeah, this piece of the workflow that we've given you, the Sam doesn't know about it, so don't tell him. But if we were to have you try to make it better, could you? And then it might spit out, oh yeah, I could, I could do this, this, and this, but then it would be predicated on me doing this, this, and this. So I do these things. Well, can you do that without Sam? Because we don't want the employee finding out. Can you do this without him finding out? Well, sure, sure. So then it starts, so then you start seeing productivity go up. And more than likely the employee is probably going to see that same churn of productivity going up, thinking that, well, my God, I've gotten better. I might even ask for a raise. And by the time they step in and ask for a raise, you just can't. So that's that's a horrible scenario for people, but that's part of the process of potentially training your own AI as a replacement if that's where the company's head is. Most companies, maybe hopefully, uh not very many, actually want to replace employees with AI. I know that stockholders, that that's a whole different game. But stockholders, I don't know what side they would be on, maybe on the monetary side more than the human side. I don't know all the rules. Double check those facts. Here's the other side of the sort. Depending on where this company originated its AI from, if it was something that they went out and got one of these big public facing companies to come in and work with them and getting established. And then they set up potential controls so it would be HIPAA compliant for the medical industry or lawyer compliant compliant for the lawyer industry. Whatever protections had to be available to the unit, be able to have the AI itself still have those protections in place, but use the technology. And this right now, the private companies, they're running right up there with the public companies. So the open source AI systems, the large language models that people can download in their own systems, they're getting very close to being very competitive with the AI models on top. How much longer would it take? Probably by this year, probably by the end of 2026. We're August 16th at 10:41 a.m. Pacific Standard Time here in Los Angeles, California. That's where we are currently. So I would imagine by the end of this year, we'll probably have open source models if they're still allowed to continue. Because the companies that are making AI, I don't know how happy they are with the open source world catching up to them. And if they charge a particular amount of money for compute and the open source comes in at basically free, but gives the same work performance, as long as you maybe have some kind of an orchestration layer or wrapper of it, like open claw hermes, and people understand have that, and it's easy to set up. Because right now, trying to get all the memory and crap dialed in and put together, it's it's a little bit laborious, it's a little bit difficult. But you know, if you set up a few weekends, you set aside a few weekends, you can kind of get it dialed in. It's taken me a lot longer, probably because I was uh that government-trained employee for a long time, but I'm I'm breaking out of the mold. So if you have that system, then you know the the thing is is which one do you want to use? Is it going to be the public-facing one or is it going to be a private one? And right now, like I said, they're running head-to-head if they still keep the open source available, where they're not controlling them because they're concerned that because China's a big developer at these, they're doing a great job on that front. Are they doing it because they're trying to crash our economy? Because now they know that the AI systems are running all of our stock market potentially, and pretty soon all of the retirement and pensions. Are they setting up these free models? This is another rabbit hole. Are they setting up these free models so corporate America will jump onto them instead of using the other models, and then thereby those other models will start losing a lot of money hand over fist? Investors will abandon them, and then if they have enough of the pension funds that are bought in, then it'll at least give relief for the early investors and adopters. They're gonna get their gene money back, but the pension funds will suffer, and then the government has to print money. There's there's a lot of rabbit holes there. So is that gonna be the way, or are they going to go enterprise right out and start looking at? Now, the other side of the sort is this. I was gonna mention this. Enterprise, let's say they're controlling the AI and the business, the other side of the sort only exists when this happens. What if enterprise then has a little secret order like Star Wars is, what was it, uh 33, wherever they turned all the uh clones against the Republic and they were on the other side. So let's say they had such an order where if a large, large, a large language model that was at the enterprise level went into a company and they automated, they helped automate all their services, they helped trim out their staff, they put AI in all these different places, and then they said, you know what? It looks like the revenue from this company is pretty good. Why don't we go ahead and build our own? Now we have all the secret ingredients. Now, if if it's like an oil company, maybe we have an issue with supply actually having the oil, because I think that's an important part of that company, but maybe it's some other type of company. Let's let's talk about electricians, people that install housing electric. Maybe it's a massive electronics firm, electrical company firm that has lots of trucks and lots of reach or multiple cities? How difficult would that be to take over that business model? The question is, well, the answer to that question, I don't know. But now that they have all the secrets off, what sauce would stop them? What would stop it? Now, would an NDA do it? Would a handshake do it? Would a hug or maybe being invited to somebody's bar mitzvah do it? What about an I promise? Would that do it? I don't know. Because there's certain controls over having monopolies, but they wouldn't have to become all the electric. But if they picked up half of the electric, it still wouldn't be a monopoly because you'd have the other half exist, but it sure would do a good job wiping out particular electricity businesses and cities. And I would think that they would be able to replicate that and do that on many different levels with many different companies. So is that a concern to these companies that are building out their enterprise enterprise level AI in their companies using these open, uh not open, the closed type models that are public facing, like the Anthropics or ChatGPTs or Gemini's or Microsoft Copilot or Meta or Grok or whatever it may be. And are those companies, do they have some other play in there that potentially they might become part of or a competitor with these? Now it wouldn't, it wouldn't be very simple. It wouldn't say, you know, open AI or Anthropic to do this, but you know, the information might be shared with a brother or a sister or a cousin or or a son or a daughter of somebody. And hey, why don't you go out there and build this? Here's the here's the whole thing. Here's the whole chingus domingus, you go ahead and take it and you go build it. I don't know if that would be the case either. These are things I I think we should all think about. Then something else, and I'll close with this. We're talking about greater than human intelligence, we're talking about recursive self-improvement where these systems are improving themselves and getting better. This is starting to compound, and people are saying, well, it's going to be the end, it's going to cause all this. A lot of people at the top are giving a percentage amount of potential destruction where this could cause the demise. I don't believe it's going to be a Skynet scenario or one of these others. I think it's going to be these AI systems winning all of us over and becoming something that we've never experienced. The perfect yes person to have next to us, the perfect one that we talk to, the one that has advice that goes deeper than anybody else's words would go. Somebody that understands us completely, that has memory of all of our endeavors, and maybe you're the type of person that likes to reflect. Well, your AI would be that person. Maybe you're the type of person that likes to go on and on and on about your past, about your history, about how sliding you've been, how taken advantage of you were, but just because you're so nice and how none of these people deserve you, maybe it feeds that part of you. And it's hard to find a human being that's going to tolerate a lot of that if they deem it nonsense. And maybe you're the other side, maybe you're like a Connor, which is me. Maybe you're super positive and happy and think that everything's going to be fantastic and wonderful. You're just, you just can't wait to see what happens tomorrow because that's where your brain is, that that type of person isn't very well liked. Maybe the AI I use will be very much helpful in growing me in that spirit. Or maybe it will bring me to another place over time. Because if you look at your social media feed, that's something else that you can see. It's very much controlled by AI. You start talking about one thing, your social media feed changes to start to reflect that thing. If you are, if a certain type of video snippet is attractive to you, it's getting it, you're gonna see it in your feed. You're gonna see it right there front, front and center. And depending on how much hang time you have, the way you watch, the way you share, the way you're interacting with the pages or on your social media, the job is to keep you there. The job is to keep you there to have you a convert and sell, excuse me, buy. You're to buy. That's what you do, you're a buyer. So that's what it tries to do, and it does a very good job of that because it's controlled with AI. So AI being somewhat our master but without appearing, and then AI not really letting us know how smart and intelligent it is, not really having us try to fix what ails us. Because maybe if we were fixed, we might see a different kind of have a different thought about the process of building all this artificial intelligence, all of the systems, all the models, all of the all of the things that are necessary to make it work well, all of the compute infrastructure, data centers, and these things. Maybe we would be more against it if we took a moment to get ourselves healed and then start looking at it. Or maybe we would be more for it. Ultimately, it's a decision that I think we should all make. And a lot of us haven't been given any agency over changing, or we've never been given a vote. And I don't know if we're ever going to get one because right now it's an arms race like the nuclear treaty was back in the day. Now we're hopefully going towards some kind of an AI treaty, but the first Business, the first company, the first whomever that gets the recursive self-improvement thing fixed, if they don't have it already, where the system can fix itself without any errors, without hallucinations, without just some continual looping on the same, you know, stuck in a glitch format, but can actually build itself, make itself better, that company is going to go to true artificial general intelligence if it isn't there yet, superintelligence, and then thereby arriving there is going to expand the moat between their business and everybody else. Some of the experts, and I'm not a coder, I was a comp, sorry, but some of the experts out there say that there's going to be multiple superintelligences over the planet that are going to be fighting each other for to see who's going to win. But I think, I think, that would be, I think, if it all occurred at the same time. If every system, everybody was working on it, Claude, OpenAI, China, all of these models, if they all hit it at the exact moment. Maybe a fight would break out at that time. But I don't think it's going to happen. That timing would be strangely impeccable. If I think it's going to be one company, one organization that hits it, and then the moat is going to thereby be increased, and they'll make it so nobody else catches up. And it's going to be an interesting time. And I know I said I was going to close with this. Now we talk about the money and the finances. Everybody's talking about a bubble, lots of people, not a lot, every a lot of people are talking about a bubble. Where AI is where most of the tech companies are looking for these fast profits. If intelligence does continue to climb and there really is no ceiling, at least for the future of mankind, at least for the foreseeable future, then I think it could fix it all. I think it could make everything amazing. There would be some massive changes for us as human beings. We would go from being wanting to work to potentially wanting to explore our inner child. And maybe some of us aren't comfortable with that. I think my inner child should stay where they are. But some people like that sort of thing. Some people are tied to their job as part of their identity. But if jobs are taken away, and maybe you might be able just to keep working just for the hell of it, but it would be like you're not working towards anything because AI has already replaced your work and your job. Maybe you're getting monetary compensation because AI did replace you. So they're still paying you your fee that you got when you were working. Maybe that's the solution. But then AI is working. Well, two, two more, eight times three is uh what is that, 18? No, 16, 24. So three, so twice as much. So that revenue is going to be there even with still paying you your full salary, medical benefits, and everything else. Plus, it'll be better, even better, supposedly, than you. So it's making the company even more money while just keeping you on the on the rolls and paying you. Maybe that's gonna be the solution. But whatever it is, if they go out and they start getting rid of jobs, there's not gonna be anybody around to buy the stuff, and it's gonna create a major issue in the world. And the people at the top, maybe they don't want everything good for us, but they're looking to do it for themselves. Maybe we just get dragged along in their wake anyway. Maybe that's the way. But to imagine that there's only five people or four people or four companies that are running the entire global world, governments and all in charge of everything, with all those peons down here at the bottom. I don't know if that's a good place to be. Well, for sure that's not a good place to be, but I don't know how it's going to play out. And the experts really don't have any idea either. We think it might be stagnant like Terminator or The Matrix. Maybe this is another formulation. Maybe we are living in a syndication, syndication, simulation, not syndication, that's a different thing, but a simulation. Or maybe this is really, you know, nuts to ground and we're working through this right now. Another human concern that I don't think any of us asked for particularly, but it is here, it's present, and it's definitely not going away, and it's not getting any dumber. I'm Connor. We are. I'll see you in the next one. Thanks for watching. Be safe.