Workforce dynamics in an AI-assisted world

Welcome to the Practical AI podcast, where we break down the real world applications of artificial intelligence and how it's shaping the way we live, work and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X or Bluesky to stay up to date with episode drops, behind the scenes content and AI insights.

You can learn more at practicalai dot fm. Now onto the show.

Welcome to another episode of the Practical AI Podcast. In this fully connected episode, where it's just Chris and I. We'll, uh, dig a little bit into the latest news related to AI. Hopefully share a few of our opinions and and maybe some little nuggets that will help you level up your machine learning or AI game. I'm Daniel Whitenack, I'm CEO at Prediction Guard, and I'm joined as always by my co-host Chris Benson, who is a principal AI research engineer at Lockheed Martin.

How you doing, Chris?

I am doing splendid today, Daniel. How's it going with you?

Uh, it's, uh, it's going going great. It's been a productive week. We've had a had a customer visit and, uh, lots of lots of interesting development stuff and last week did a workshop. So lots of hearing all the different kind of archetypal use cases that people are working on, lots of interesting questions. I think we're also in a hiring phase right now, which means I am thinking a lot about kind of what positions to hire in.

Especially in light of all the AI stuff that is happening and everyone talking about, oh, you know, maybe you should consider AI could do this task before you hire someone in to do that. Uh, what are those discussions happening in your circles as well?

Oh, I've I. They are, and, uh, they. I hear that a lot of people are asking me about questions like this, you know, increasingly, and it's funny that you bring that up. I was, uh, I was doing another podcast, uh, as a guest, and some of those questions came up in that podcast, uh, asking about kind of how we saw that. I think things are getting so fa- you know it. We've gone through, uh, several years of big LLM experience and now the agentic world, followed rapidly by the physical world.

And people are really like, paying attention to how their lives are starting to be impacted by this. Uh, with the with jobs being right at the top of that list. And the, you know, concerns on what it is and what should I do and will I be replaced? And it's a mixture of excitement and fear out there.

I know that you can't share things necessarily from your immediate actual, uh, j- data, day to day job, um, always. But let's say Chris Benson in a fictitious world. And you were well, I guess you also have the nonprofit side of what you're doing. But as you're thinking about, as you would be thinking about in this context, like the world we live in now, knowing not the promise of what AI might do next year, but just based on capabilities now, the tools that you're seeing out there.

And you are building maybe a small team for a new company, w- What do you think is kind of on the, you know, maybe positions or things that might be on on AI, which may have traditionally been, uh, a full time or a part time hire? What? What is your thought?

Yeah, the the, you know, the the. The sense I'm getting from, you know, from my interactions with people and and what I'm observing is is is still using AI mostly as a tool to empower people. Though, that might mean that there are fewer people in a given function, you know, than there might have been before. And you know, like, y- there's some really obvious examples that we've all seen, like like marketing material and using AI to to generate that and stuff.

And so, you know, I think I'm not seeing AI like completely replace lots of positions right now. Um, there's probably some here and there, but I think, uh, I think that companies are looking for efficiencies. And you know, if a marketing department could be two people with AI tools instead of five people in a company, I think that's a really common approach. That that you're seeing right now, you know, today and and over the last couple of years in terms, and I think there are many, many instances of that.

You know that that small efficiencies, small functions that are being, you know, are task level things that are being offloaded, but which by doing that reduces total headcount, you know, to some degree.

And I guess there's things that maybe people can do faster. So, to your point, there's maybe fewer of a particular type of position that is needed. Maybe it's marketing or or even development or or other things. I also find it interesting that there may be things that people could do that in a certain position that they couldn't maybe do before. And might have to hire, like an outside agency or services organization to do, which is an intriguing part of this.

So just thinking of, you know, advertising, PR, market research, tho- those, of course, are things that that you mentioned. But also there's kind of, like, prototyping new software things, you know, proof of concept that maybe you would have to engage an external, uh, consulting firm or software dev shop, right, to to do before that. Maybe you, even with, you know, fewer years of development experience or less exposure, could knock out internally. What? What's your thought on that kind of interplay with, uh, service providers and how, I guess this this world may impact the service provider side of things?

so yeah, with service providers, I think it's, you know, it's an interesting story. I don't bring it up very often in our episodes. But before I was in the defense world, I spent a dozen years in the digital marketing world. And, you know, as a as a CTO or other technical positions and stuff. And I think that it's really tough with these new AI tools to be in that industry at this point because there's so much capability.

You know, we've commented many times that, you know, it used we. Once upon a time, we would say AI would take care of all the grunge work and and leaving humans for the creative work. But we've actually kind of seen the opposite, uh, in some cases. And I think the the creative aspect of some service industries like that, you know, compared to AI tools, which may or may not be able to do it at the same level, but they might be able to do it good enough.

You know, we, you know, in software development, we often talk about good enough. And you mentioned prototyping a moment ago, uh, and and I think, you know, with the notion of vibe coding, making at least simple coding efforts, you know, for things like prototyping, accessible to people who normally, uh, historically, would not have been able to do that. Is also a bit of a game changer. And that means that, um, those are. That's another area where hiring externally or hiring that function into your organization is also reducing.

Because you don't need that necessarily just to get a prototype and that, you know, it may be different for writing production systems. And even then, AI tools a, as we know, have have had a had an impact, though. I think I think the impact on that can be debated on whether it's positive or negative, and depending on who you talk to. But yeah, it the these tools are definitely changing many industries right now. And and they may not be completely knocking out an industry, but they are impacting it in terms of its viability from a revenue standpoint and such.

So, um, yeah, we're, I mean, it's, it's a concern. And I think most people start off thinking, well, most people that are kind of, you know, AI cognizant on a day to day basis, start off thinking, what tools do I have available with these new AI services and the models that I might host ourselves at our company? Can I do now that maybe I couldn't have done before, and how can that change my cost structure?

Yeah, I, I'm really intrigued by this, uh, service provider side of things, actually. Because on the one side, there's maybe traditional things that people would have gone out and hired an agency to do. So, maybe that's a new website for your business, right? Uh, or a, you know, simple website for your business, or some type of marketing assets. Or maybe a prototype of a project, like a SaaS product, or I don't know, a a bunch of different things.

Those things seem to be good targets for kind of the either vibe coding or kind of wrapper type platforms around certain AI systems. So, you know, for the very first, I think the very first one of our websites, um, that we made for Prediction Guard, I just used one of those sites. Where it's like, create your website with a prompt, you just go in and prompt the thing like, I want a website for this thing, and it kind of generates it.

And you move around a couple things and there there you go. So I think that those types of things, the sort of non-AI things that are generated ba- by AI are very much maybe tho- those are. There's less of a need to go out and hire an agency or a services company for those things. At the same time, I do think it's interest interesting that if you look at building AI things for your company. So let's say you are a mid or large size company and you want to create this AI tool for your company, right?

The talent is so scarce for that sort of thing that really, I I mean, if you look at things as a whole, OpenAI or companies like that are definitely not the companies making the most money off of AI. You know, OpenAI is losing a lot of money. The companies that are making a lot of money off of AI are Deloitte, Accenture, McKinsey, right? Those those are the company. It's absolutely raking it in off of, off of AI, right?

Because indeed, there's all of these companies that say, like, well, we wanna, you know, transform our company with AI. So you need sort of change management and you need to create, like, you know, these prototypes, and maybe these AI tools that you're gonna deploy a across your company. You need to take care of the security, privacy concerns, you need to have a strategy around it, a roadmap, like all of this stuff, and that is all very much tailored to what these services organizations do really well, right?

And I think even smaller MSP type of, uh, companies that have a book of business in healthcare or other industries where the impact of AI is clear, they have a big win to have in in this area too. So I find it really interesting on the services side, where some things maybe are getting a little bit eaten up by AI, but the actual AI stuff, the AI transformation is actually very ripe for services organizations. It's it's interesting,

it is, I mean, uh, uh, you know, I guess, uh. People are looking for guidance they don't understand how. And there is, of course, this whole industry of consultants who are, uh, happy to tell you they have it all figured out. So, um, and I, I did a, and a brief

and they may a and I'm sure that m much of that. They're very much helping people, but they are well positioned to do that, and that is kind of their business model, right?

indeed it is. I did a brief, uh, stint at Accenture as well. Uh, so I, I, I can, I can, I can attest to that.

yeah, yeah. So maybe it is a bit of a mixed. Well, if you're an MSP out there, maybe you're, uh, you're. You're thinking that some things are are going away, but there is maybe a big opportunity for you to go into this sort of set of offerings around AI services.

Well, friends, if you wanna build the future of multi-agent software, check out Agency. That's A-G-N-T-C-Y. It's an open source collective building the Internet of agents. It is a global collaboration layer where AI agents can discover, connect and work across frameworks. That means better tools for you standardized agent discovery, seamless inter-agent communication, and modular components to scale your multi-agent workflows. And they're teaming up with Crew AI, LangChain, Cisco and many more, dropping real code specs and services with no strings attached.

Start building alongside engineers who care about high quality, multi-agent software. Learn more at agency dot org. Again, that's a-g-n-t-c-y dot org, that's agency dot org.

Well, Chris, we started talking a little bit about influence of agents and that sort of thing on workforce. There was a a very interesting story that came out that intrigued me as related to this. Because basically, everyone's saying, well, you know, all, uh, 60% or eighty, or whoever's giving the percentage will say, you know, the majority of development work is now gonna go to AI systems. And it was very interesting because there was a story that came out, the one, I mean, it was reported by multiple folks, I think.

But um, this one I'm looking at is from Fortune. S, and the the title is An AI powered coding tool wiped out a software company's database, then apologized for catastrophic failure on my part. Um, and this was, uh, I believe it was Replit that, uh, is is the the culprit here, which is a really a amazing tool. And, uh, you know, they have, I think there's been apologies, and you know, of course, they they're handling this in one way or another.

But I, I'm sure this is not the only instance of this across these kinds of tools. But also it's a it's maybe a bit of a wake up call for some people, um, to to understand that maybe these tools are moving at a little bit of a faster pace than what might be reasonable for companies to support just operationally and permissions wise, and and at scale, et cetera, et cetera. Any thoughts?

My 1st thought was, oh, was, you know, lack of guardrails, lack of of, uh, consideration on that? And I don't know the details, uh, any more than you do, you know, other than what was in the article. And maybe part of my perspective is biased because of the industry I'm in. Being in defense, we're really, really cognizant of safety measures and guardrails and such like that in our industry. Because obviously it can turn into a problem if you didn't have that.

But I think other industries sometimes may need to to stop and think. And I, I get that the marketplace is moving really fast and there are competitors, uh, that will eat your lunch quickly and you wanna get to market with the best thing. But kind of at this point, you know, we're still fairly early stages in agentic AI, you know, a lot of frameworks are still under development. Uh, this is, and and so I think having a, you know, kind of.

It's a good thing to ask yourself what happens when things don't work out per your expectations. Um, what are the guardrails for when things really go bad? And maybe think kind of a little bit of, maybe, uh, a little bit old fashioned in terms of, you know, what can you do for, uh, data redundancy and backup and things to where? If your, uh, AI experiment or effort does go off guardrails, that you have, uh, uh, a guarantee that you don't have real damage done in a in a large sense, uh, after that, something that literally has no permissions to.

So there's probably some mitigating factors there that I'm not aware of. Uh, so I'm not trying to to to lay in on on the company. But I would just urge people to to think about worst case scenarios and and maybe architect, spend a little time architecting for that, uh, and then and then go do some cool stuff. But make sure you can fall back without blowing your company up.

Yeah, good point. Uh, what, what do you think? Well, one of the things I've been wondering around this is there's all of these tools coming out, you know, vibe coding tools, vibe marketing tools, vibe, you know, design tools, or even HR, hiring, recruiting, et cetera. Na- name any function. There's a way to do it with prompts or these sorts of tools. Now, what? How do you think this influences the type of both education and professional development that people need to have coming into a job?

Because obviously there's, there was maybe, and and I again, you know, we only know, uh, what we know about this one instance of the database being dropped or or that sort of thing. But I'm guessing that, maybe, and you know, people can correct us on social media if I'm saying something wrong here. But I think in a scenario like this, regardless, there's probably some of this that's on the company using the tool. And like you say, the the the trust and the guardrails that are put in place around kind of how things are pushing to production and how they're iterating on their platform or code or that sort of thing.

And there's part of it that's probably on the platform itself and how it's architected and failure modes and safeguards and and that sort of thing. But part of it, I think, is well, how? How should we train our new workforce such that they're ready to kind of team with AI systems in any role, one or the, you know, from not technical to non-technical roles? And how do people in existing roles that maybe didn't haven't spent their career using these tools?

How do they kind of? How should they professionally develop themselves to be proficient in this new space? Any thoughts?

Yeah, I, it's something I think about quite a lot on several factors. Um, there's, there's my own career, you know, as we're, as we're working through things and things are changing constantly. And I also I have a a 13 year old daughter, you know, who is, uh, who's going into 8th grade. She'll be in high school soon and then into college. And you know, we are. We are living in an age of experimentation right now because these technologies are coming so fast that the nature of what it means to use it from one year to the next is very different.

You know, um, 2023 to 2024 to this year have all been distinctive in terms of how these tools, um, are able to affect our jobs. So I think the 1st thing is, I would say, kind of accept it, accept the change. And I see, you know, I see a lot of people who, uh, are highly technical developers, things like that, who probably should know better, um, kind of resisting and all that. And so I think the 1st thing is to say this is happening and it's not going away.

And then the 2nd thing is on on individual roles, on individual jobs. To try to imagine where things might go, given the tools that you're seeing today, what do you think might change incrementally tomorrow? And what kind of trends are you seeing in your particular industry? And go ahead and get ahead of that curve, uh, and and I think that's really valuable. I think it's hard to do, so it's it's very easy for me to go make a a recommendation like that, but seeing the future is not easy.

Um, and being able to to do that. So I think my recommendation is given that this is a. This technology is only going to continue to grow and change things. Uh, accept that your job is changing, accept that that that the the vision that you had yesterday is not gonna be what happens tomorrow. Um, and so, uh. Try to track on what you're doing, embrace the tools and figure out what's changing in your job. Um, how can you make it a net positive instead of a net negative?

And those are very generic, but I think it's a very job specific, um, you know, set of tasks that you have to do in terms of that. And I think that starts with education. Um, it's one thing to be doing it on the job, like we are in the middle of a career, but um, how do you train today's youth to to do that? They're, they're really good with these technologies, but we're, uh, th- they're not the ones setting the curriculum, it's us older people that are setting the curriculum.

And I don't think we're doing them any favors right now. Uh, in school curriculums, they're, they're definitely not set up for success yet.

How far behind, in general, are we, do you think on that, on that front? In terms of, let's just say, college grads coming out into the workforce, in terms of what, maybe, you know, take any field, uh, you know, accounting, marketing, communications, business, whatever it is. How far off do you think we are in terms of the realities of what day-to-day work might look like with these tools versus how that's being represented in the in the classroom?

Any insights?

Yeah, I, I think, I think, look to how people are doing it in a hobby sense. Um, you know, because people who are interested in adopting these technologies. Um, uh, we've seen this for, I mean, this is not new, this is for many, many decades, you know, d uh. Companies tend to adopt what their developers go home at night to their home and play with and are interested in because it's the cool thing to do. And I think you can, you could expand that across, uh, a lot of different fields and industries.

You know, w- what, what do people do when they're doing it on their own time? Because there's this new thing out there? And if companies were smart, they'd pay attention to that, if universities were smart, they'd pay attention to that and track that fairly early. Kind of what the maker space or the hobby space is doing and build curriculum around that. Because right now, I mean, as we sit here today, most schools, if if students go in, the, uh, the professors are trying to find ways to keep students from using models that are publicly available.

Uh, to do their homework and stuff. And when I hear, and and and every time I hear an academic, uh, say something like, how do I keep my students from using ChatGPT to cheat on their homework? I'm like, Well, you're teaching the wrong thing then, because that, that that model is never going away. So if you're teaching something that that's causing you a problem, then you're doing y- y- your whole curriculum is off track. And I realize that may not be the Per, you know, a specific teacher's ability to change overall curriculum, but it's certainly, uh, university and school leadership, uh, school system leadership that should be doing that.

So, um, there's a big correction that needs to happen in education. Um, that companies would be very smart if they could also track that very early and go ahead and make adjustments earlier rather than later.

Yeah, and maybe part of the struggle here, which I maybe just playing Devil's advocate a little bit on the Edu- well, there's a lot of, I think, as everyone would recognize, there's only so fast that education can move, just because of bureaucracy and other things. But uh, just putting those things out of the uh, I I mean, this is also true, probably for companies where you look out a around the landscape of AI tools that can affect my job.

And you look at the snapshot today and then you look at the snapshot next week and it almost seems like it's totally different. There's so many things coming out and I can put in the investment to learn this thing, and then next week it's total. It seems totally useless, or, or it could seem totally useless. Um, how, how do you? how? how does one out there that's maybe listening to this navigate that kind of, uh, whirlwind of releases?

so so I, I don't plant an anchor in any one thing as a long-term bet. Um, I think I think to win these days, you have to, uh, I think you have to embrace that learning mindset, and it's very cliche to say that, but I think you have to do that in your, in your day-to-day actions. Um, you know, I, I'm cert- I'm at an age, I'm 54 right now, closing in on 55. And um, a lot of my peers, uh, in in terms of age, have already just kind of given up.

They've already decided what they think their worldview is, and they've stopped and planted right there. And I I think it's exactly the wrong thing to do. I think, um, I. The world is changing fast, and to continue to be successful and happy in such a world, you have to roll with the world on a day-to-day basis. And you may invest in something today based on what you think tomorrow looks like. And you may realize in a few months that that's not the way things are going and you should be ready to pivot again.

And so there is a never-ending series of pivots or or adjustments that you need to be making on an ongoing basis for the rest of your life. All the way till the day you're dying day. And if you ever quit doing that, you're putting yourself in a worse position. So that learning mindset is a is an action-based, uh uh activity. It's something that you're, you're literally doing all the time. And I think people who can do that will tend to be more successful and happier in what they do than people who just resist and resist and plant flags.

So that that would be at at 54, that's what I think I've learned.

well, I d I definitely. So there's a whole element of this that really has nothing to do with AI, which is, uh, also kind of based in neuroscience. And you know, I, I've read a lot about, like, uh, dementia and these sorts of things, as you know, the sort of, uh, that's occurred in in my immediate family. And and that's one of the things that a lot of neuroscience neuroscientists talk about. Is that just act of always trying to learn a new thing is one of the best things that you can do.

That's tied towards, you know, po- Not that it's a guarantee, but it's definitely, you know, positively correlated with good outcomes long term in terms of preventing dementia or, you know, other sorts of brain type of conditions. So maybe that's a maybe that's a good sign, uh, regardless, uh, of whether it's AI or not.

That's right, a little bit of, I mean, it's. There's a mental health benefit to doing it, and there's a career benefit. So just love learning and pick lots of different stuff to try out. And some of it's little things like puzzles, and some of it is learning a whole new thing that you've never touched before.

Well, Chris, we we talked a little bit, I guess, more generally, about the impact of AI on the workforce and maybe professional development education. There's an interesting side of this too, which is the actual workforce that is developing the AI things. So, you know, AI, and I think, you know, we've kind of started to adopt this term AI engineer, you know, I think in a lot of, uh, uh, a lot, a lot of thanks to, uh, Leighton Space and our friends over there, swix and and others from the AI engineer, uh, summit and and those types of things that have put this term out there.

But whether it's AI engineer or data scientist, or machine learning ops, or what, whatever it is, this sort of workforce that's actually supporting the build out of the AI, both infrastructure and application layer. That is driving a lot of these transformative tools. I think we would all recognize, well, maybe our listeners would recognize, but certainly, and I hear this, I, I know this directly because I I talk to our customers, who many of which are trying to hire in this sort of talent into their companies.

It's very difficult right now, there's a lot of positions available there. There's kind of, uh, it almost reminds me a lot of back in the day when people were trying to find the unicorn data scientists, right? That sort of didn't exist, maybe back, kind of 2011 to 2015 type of time period. Now it's like it's expanded, cause the impact of this technology is just s so much more. But everybody's looking for AI engineers and talent that is related to kind of this wave of generative AI technologies and having trouble finding those.

Um, I, yeah, I don't know, is is this a also a trend that that you're observing?

Yeah, I think so, I think. And you know, not only, not only is it a trend, but, you know, it kind of points to the inequ- the inequality in the larger space. In terms of we- the value, in terms of what compensation would be for that, the value in different, uh, industries. And so, uh, AI engineers, and, you know, um, specifically, you know, research scientists that are developing new model architectures and stuff like that. You can get some awfully big compensation for that.

And yet there are many, there are many fields out there. Uh, if you're not able to do those kinds of skills or whatever that are, you know, as we talked about at the beginning of the conversation, which are decreasing and stuff. And I think that's another, I think that that's another, uh, thing that people really need to consider is, uh, you know, is the track you're on going to be in a place? Do you do you?

Do you expect your track to be in a good position two, five years out from now, based on how you see things today? And and that's another reason, uh, even even in a late career stage, maybe consider making changes, maybe maybe, uh, maybe it's time to make adjustments and stuff. Because, yeah, AI engineers are able to basically write their own paychecks in a lot of cases. Uh, a lot of other industries are seeing, uh, decreasing revenue, uh, to the for those particular jobs.

Yeah, one one data point on this, which is kind of interesting, is news that came out a little bit earlier in July, where Apple's kind one of Apple's top AI folks, Roaming Pang, was poached by Meta. And I don't know how accurate this is, I haven't done the but the articles are saying sort of total compensation package over x number of years of like 200 million or something. Is sort of like, I, I guess how you could think about this is like these top AI people being swapped around, almost like top athletes in a professional sport.

With big contracts are the best F1 drivers going from this team to that team and signing a big contract or whatever it it. It almost seems like we're living in that world. And, That's a good point, uh, and yeah, I mean, and we are yeah, yeah. I'm very happy to be building what, what we're building, you know, with with my team. But it also makes me think in the back of my mind, it's like, whoa, like, you know, we've got some great talent here.

Like, I, I hope, I hope no one offers them 200 million, cause I can't, I can't, I can't match it there. There's like a market where, uh, you know, you can, just I. I also wonder about retention of some of these folks, where you finally find an AI engineer that's working well for you. But they could go out and demand twice as much somewhere else. And yeah,

they they could, uh, so, so there's two points there. One is is this story doesn't really stand alone. There's been a number of of very similar stories with Meta. Um, Meta has been has been hiring

I think Netflix too, uh, and some other companies, yeah.

yep, but but Meta, in particular, has been known the last, you know, couple of months. Uh, for pulling people from OpenAI, Anthropic, Google, Apple, now, and that's a GR- the sports analogy made was right on. They, they're basically weaponizing kind of that hiring process to try to get the right talent from. And and they're gambling a lot of, a lot of money that they're going to be able to put together a superstar team that will outpace the market.

I think, but that's that's also at that extreme high end area that we're seeing that I think in the short term, I think we're seeing compensation, uh, go up for these, for these skills. As people are, because we're still in the very early stages of people setting up, you know, truly mature AI infrastructure and stuff. But I think over time, just like we've seen in software development over the decades and other industries, and, you know, with technology.

Uh, because there's more compensation there, more people will gravitate to those skills. Um, it will. You will have a bell curve of capability in terms of the re of the candidates that are out there. And generally, compensation will fall over time because they'll become, uh, you know, much more commonly available, because that's where the money is.

And then people will get a lot of money for being AGI engineers.

Yeah, you know, that's coming, we're gonna see by 2026, 2027, we're gonna be talking about AI, uh, AGI research engineers and stuff like that. That'll be the new title coming. Folks, you hear, you heard it here 1st, right out of Daniel's mouth.

I'm I'm gonna, I'm gonna start a AGI engineer summit and just There you go, start the the trend.

I mean, you'll notice that that's what that's all Sam Altman at OpenAI talks about all the time, and and so is Mark Zuckerberg with Meta. They're just talking AGI, AGI, AGI, but no one's doing that. Yeah, like, they're, they may be doing research, but they haven't gotten there yet. But it's great marketing and it, and it's driving the compensation spectrum. But that will, that will mature, and the AI tools themselves will become better suited for creating infrastructure and setting up stuff.

Just as we were talking about earlier with other areas, so it will, you know, we're in, we're in a bit of a bubble in that area. It it will, it will eventually, um, pop.

o- one one interesting we we need to have. Um, we had a previous guest, uh, one of my friends from the Intel Ignite program, uh, Ramin, on he, he's, uh. He also has a job as a AI engineer, but also teaches at North Northeastern. And I was just talking to him. Just this sort of interesting data point, we we need to have him back on the show to talk more about this. But he, he was mentioning that he, he teaches a couple different courses.

One of them is kind of more the sort of theory behind generative AI, more of the, I guess, the transformers architecture, you know it, just the the the bones and underpinnings of all of those things. And then a different one that's like machine learning ops, or like AI ops type of course. And w- he's clearly seen a shift over the last a few times that he's taught this, where in the beginning it was everyone wanted to be in the generative AI theory course, and that's basically flipped to everyone wanting to be in the ops course.

And his, at least his quick reflection on that, was that people sort of assume now that, well, you don't really need to know the theoretical underpinnings of any of these things. It's all about how you operationalize the technology and put together all the plumbing and integrate the data and and scale it up and and that sort of. So, I'm not saying, I'm not saying that's, uh, the correct perception necessarily, but I. But I think it represents an interesting shift in how people are thinking about, like, what is what is an AI engineer, or what is an AI position in in, maybe an AI, uh, AI or not AI company.

There may be a trend, and that's only one data point, so I can't speak more generally, but there may be a trend where people like when we say AI engineer, kind of like when we said data scientist. There's sort of an assumption that you kinda know statistics at some level, you know, kind of some of the underpinnings of these models, maybe how to do evaluations, this sort of thing, the more quote technical kind of underpinnings of that.

Where, you know, maybe some data scientists were just really good at using gradient boosting machines and, you know, using all the tools, right? And so there may be this kind of interesting dynamic now, where there is such good tooling around AI. That it's almost like there's this engineering of the tools, not engineering of the AI that people are interested in. So not not the actual model or the underpinnings, or how how to use it, or the theoretical understanding, but in a sense, how to engineer at this more abstract layer and connect up all the plumbing and ops around it is very interesting.

Yeah, I I think I, I, and I can understand that. And there is a bell curve of skill and capability that people will bring to such positions. I think, uh, I would encourage people to do both, uh, don't do one or the other, do both,

Take both of Ramin's courses, please.

there you go it, it, it'll make, it'll make you better, never stop learning. So, um, yeah, uh, that's when I hear people, you know, skipping over all the theoretical. Um, I get that, you know, you might have a practical intent in mind, but, um, there's a point where a certain amount of theoretical knowledge helps you do a practical job better than you could otherwise do it. So, um, yeah,

yeah, I, I always like this idea of, uh, having a mental model or having int- at least intuition around how these things work under the hood. We've talked, you know, multiple times on this show, just understanding. I think there's a lot of people who use these tools that don't, even they. They don't understand that. When text is streamed onto a screen, right? Or you're getting a streamed response back from an LLM, that is actually every token.

It's an operation of the model which produces that output, and then it cycles around and you run the model again. And that actually has a number of implications that help you understand a lot of things. Well, that makes sense why, you know, closed model providers charge by the token output, because that's connected to compute, right? Or if you host your own model, it means that throughput or or streaming, or the time that you need for a response.

It is tied to how much text you're you're outputting. And that has implications for how you quote AI engineer your solution and put it together, especially if you're concerned about latency or user experience or other thing. So there's just so many like trickle down things, even just from that one kind of example of how mechanically a model works, right?

I I think that's, I think, your point on uh, spot on right there, so, uh, that's I. I think that was well said.

Yeah, yeah, well, well, what? Uh, I guess. We've talked about workforce, we've talked about education, we've talked about hiring, how jobs are are shifting. How do you feel about, as we close out here? How do you feel about your your own, your own position? And what? Uh, what are you thinking about, Chris, as you go to the next phase of of what you're involved with?

So, uh, I, I am trying very hard to follow the uh. My advice from a few minutes ago is I try not to let a single day go by where I don't do a little bit of self analysis, a little bit of introspection. Look at what I'm being asked to do and how I might, uh, change that. That almost every month, I make a little insight into what I need to do to adjust. Agentic AI here in 2025 is changing, um, how I'm putting solutions together, uh, and how I did it in 2024.

Um, and so, and it will be different again in 2026, and I already know that. And so I think that that never, you can't ever stop, uh, evaluating what your present and future look like, and never put a stake in the ground on how you're doing things, uh, that might hold you up tomorrow. And so I think that's, I think that that can. I think that's general enough to say that that could be applied across almost any industry.

And so don't find yourself stuck in the mud, don't be, don't be the old codger that I am always trying to keep from being myself. Stay young in your mind and stay agile in your thinking. That's what I would think.

That's awesome. I think that's a great way to end this discussion. Thanks for thanks for digging in with me, Chris. Enjoy the evening and whatever you're gonna learn tonight.

Sounds good you too, Daniel, take care.

All right, that's our show for this week. If you haven't checked out our website, head to practicalai dot fm and be sure to connect with us on LinkedIn, X or Bluesky. You'll see us posting insights related to the latest AI developments, and we would love for you to join the conversation. Thanks to our partner Prediction Guard for providing operational support for the show. Check them out at predictionguard dot com. Also, thanks to Breakmaster Cylinder for the beats and to you for listening.

That's all for now, but you'll hear from us again next week.