Techstrong Gang – June 13, 2024

Hello, I'm Mike Dard and welcome to Techron Gang for Thursday. We're gonna be talking about what Apple was up to with AI this week. And then we're gonna dive a little bit into, well, how AI is changing the way we are all about to learn.

And then finally, we're gonna look at what AWS is up to with generative ai. 'cause it looks like they might be transforming DevSecOps. You're watching Textron Gang.

We'll be back in a minute. Hello everybody, and welcome back. We're here with the gang, which consists today of Mark Hinkle, who's joining us from North Carolina.

And Bonnie Schneider, who's my co-anchor for the day. Yes, great to be here, Mike. Thanks.

All right. We are really kind of keenly interested in what's going on with Apple, and I'd love to get Mark's opinion at first, but it seems like from my perspective, apple basically announced their flavor of copilots this week, and they have some unique takes on it, but I didn't really see something in there, mark, that made me go, wow, is Apple taking the lead here or did did I miss something? You didn't miss anything, Mike, I think, I think everybody was sitting in the audience at ww DC saying it's about time.

Um, they really have been pretty quiet around ai, but their defense, there really hasn't been a killer app around AI or desktop. It's been interesting. So, I mean, I think the big news was that iPad finally has a calculator app after 15 years.

So that, that was the, uh, that was the, the only thing that was really novel. But what I think they're doing is setting the stage until we have, um, sort of a small language model that downloads and is being used locally. I think it's just one more, um, interface or set of interfaces to a large language model, the cloud.

Um, I, I was missing that Steve Jobs and one more thing for sure. But, um, what was your take? Well, you know, uh, of course everyone was focused on the AI announcements in chat, GBT and I was looking at the sustainability side of it.

And one of the, uh, well they mentioned many things related to sustainability, but, um, relating to ai, they talked a lot about how, uh, the devices will have AI on device processing, so that will be less use of the cloud. That was one of their, uh, clouds for sustainability. And others were using, uh, recyclable materials and a hundred percent, um, aluminum for their products as well.

They're really cutting emissions across their supply chains, which is interesting. They convinced over 300 suppliers use renewable energy. I kind of break it all down in a video coming up.

I'd like to share that with you. We take a look at the sustainability announcements related to WW DC 2024. Hi, I'm Bonnie Schneider with your Ecotech Analyst Insights.

Apple's Worldwide Developers Conference showcase the company's dynamic strides in sustainability. As part of the company's accelerated plan, apple announced no plastic packaging by 2025. That's one year ahead of schedule.

This move aligns with Apple's strategy of using only recycled or renewable materials in its products. While some rivals have pledged to incorporate reused elements, they have yet to set definite deadlines for completely phasing out plastic packaging. Apple also announced that all of its devices, including MacBooks, iPads, watches, and Apple TVs now use 100% recycled Illumina.

No other major tech company has achieved this level of recycled material integration across its entire product line. What about sustainability outside of Apple's popular product? Well, apple has transitioned all of its operations from offices to data centers to 100% renewable energy sources.

Apple even rallied 300 suppliers to adopt clean energy. This comprehensive approach to tackling emissions across the entire supply chain addresses the often overlooked scope three emissions that constitute a significant portion of a company's carbon footprint. While Apple has made substantial progress in its sustainability efforts, the company still faces obstacles in achieving true environmental responsibility throughout its product lifecycle.

Prioritizing device longevity and ease of repair would align the company with the principles of a circular economy where products are designed to last longer. Embracing this strategy also caters to Apple's growing eco aware consumer base. Overall.

The environmental initiatives presented at Apple's 2024 Worldwide Developers Conference highlights the company's accelerated commitment to environmental goals. As tech giants continue to compete and innovate Apple's eco-conscious efforts will play a crucial role in shaping the industry's response to climate change. So Apple's made some exciting announcements, but really the main thing was is that they've kind of amped up their deadlines, bringing deadlines, um, that they had previously set sooner and just build on existing progress as well.

So, um, that's the sustainability side of wwc. Mark, what did you think of a cloud relationship? Because I looked at that a little bit and I said, well, it's great that we're running AI as much as we can locally, but it also seemed like they were locking us into a network that consists of servers made up of Apple silicon, and it had to be accessing LLMs that I guess they're gonna approve like the way you run apps in the, uh, app store for the iPhone.

It's not quite clear to me how that's all gonna play out, but it seemed to me they were kind of being a little restrictive. Yeah, I mean, have you ever seen a charger for the iPhone before USBC is, you know, part of the playbook is to have, you know, not only good integration, but probably a pretty good lock-in. And, you know, we pay for that convenience and I still am an Apple user, but I think the end result will be that Apple will have a, you know, good AI experience and it'll be end to end just like iPhone iCloud and the Mac is today.

So, um, yeah, I I I, I watched the, the videos and the, the talks and I mean, I think it's a good vision, but I don't think that it's groundbreaking. It also seemed to me that they were kind of hoping that there were less hallucinations if they kept the LLMs kind of tight and approved and vetted on their side. Is that a reasonable expectation or is 'cause in theory that approach might result in a better output than what we're seeing with Microsoft and Copilots and open AI and all this other stuff?

Yeah, it's, think about it like this. If, if these LLMs are like tour guides, a tour guide that is touring is local city, it's probably better than one who's going halfway around the world. So you have all your data on your laptop, um, or your I or whatever your Apple device is.

And so its local library of knowledge should, if well implemented, um, inform the model more in a more personalized way than than using a large language model, which is trained on the whole internet. So I, I think that that's probably depending on, like, especially for customization, user preferences around emails and calendaring, it makes perfect sense that if you're using that data and probably implementing something with, uh, local, uh, rag, uh, retrieval, augment a generation, that's probably gonna be better. Microsoft and Apple are clearly trying to drive an upgrade cycle here, and they're hoping that that will start in the fall.

I don't know how much of these machines are gonna be available in volume, but let's say it goes into 2025, would you switch from one platform to the other because of ai or you just kind of sort it out and wait as it comes along? No, because I think that the AI is, is going to follow me. I, I feel like wherever I go there's ai, so my switch Mark, you think the same thing?

Well, I, I mean, if, if, uh, you go back to what we were originally talking about is the local stuff, there is a little bit of difference between the way an LL language brought on, uh, different kinds of silicon, but I I, I generally agree with Bonnie. I think that we'll look, we'll have more like a Chromebook experience for our AI where everything's in the file and it'll follow you. But, um, there'll be local action models that hopefully, um, do things like, you know, automate your calendaring and, um, I'm excited to see how perform at the, uh, M four silicon is for ai because to Bonnie's earlier point about pay ability is, it's gonna be a lot easier on our environment, probably in long time tab liquid processing than to keep building mega.

So that's probably the most, um, promising thing of the news on purpose speak. I was surprised. I know it was a developer conference, but there wasn't a whole lot of talk about what exactly is happening with the Apple silicon as it relates to ai.

Um, is there another shoe to be dropped here? Do they need another upgrade of the processors, or you think the existing ones can handle these AI workloads? Well, I've been actually, um, I have an M two processor locally and have been playing around and I've been surprised at how well it, it handles a local, um, LLM running locally.

You know, I'm not, you know, trying to solve climate change or answer, I'm just trying to write better documents and organize my thoughts though. Um, I know, I'm, I'm hoping that it, it just keeps getting better. Um, we're getting more, which is, you know, the key is how much our uses to do the things we wanna do.

And, um, I think my laptop's years five years old, it's good. Do you think that we're gonna find some way to bridge the divide between Apple and Windows environments with ai? And I asked the question because, you know, the boys in my family are all Windows, the ladies are all Apple, and you know, the phone connections don't quite work and all the apps are somewhat different.

And so we're always looking for, you know, maybe we can all use WhatsApp or something or some third party thing that might work, but it's, it's cumbersome and as we move into the AI era, it's like, well, I'll have my AI agents and you'll have yours and Mark will have his and are they gonna talk to each other across Apple and Windows and kind of sort this stuff out. How do you think this might play out? Well, I I think that people definitely are loyal to, I have Android and I get made fun of wherever I go, Android, I can't see your text messages.

So anyway, I, I relate to that, but I do think that the AI itself is crossing over, and here's what I mean, on some of the, um, pro productivity apps and research apps, you'll find that you have a choice. Do you wanna use chat? GBT?

Do you wanna use Claude? So the different types of AI are, are being exposed where we we're not just relying on one now with Apple leaning into chat GBT that might bring new people to Apple that are, that are enjoying chat GBT. So it's interesting, I, I think the AI will allow, uh, exposure and the, the opportunity to try something that maybe you wouldn't normally do.

Mark, do you think all our AI assistants are gonna get along with each other? Are we kind of approaching some sort of kumbaya moment? Yeah, I think the, the key is that the new interface is natural language.

So, um, I mean, behind the scenes there's still APIs and such, but if you are have AI talking to AI and it does natural like processing, um, you know, it should be, you know, the, the differences in interface will probably dissolve every time. Do you think we might get more locked into our platforms because of ai? Because well eventually, you know, people ascribe personalities to these things and it has all my data and my entire life is wrapped up around my AI agent.

So am I making like a lifetime choice here? I don't know if you're making a lifetime choice, but it's probably as hard as it is for the people in your family to handle the blue bubbles versus the green bubbles. It's inconvenient.

And that inconvenience is what helps us stay on the platform. So, um, I would, I would definitely say that once you train these agents and your ai, that the learning curve for adopting a new ai, it's not transferrable. Um, um, as things stand today, I think also it's going to be, is it worth the hassle to convert, right?

Because if you're in Apple, pretty much all your products are in the Apple system. I know I'm in Google, I have everything Google. So if you'd have to have a, a strong motivation, maybe ai, is it Ultimately Mark, what's the pace of LLMs that we're gonna see?

And I'm asking the question 'cause I feel like there was a wave of this stuff that came out and, but I wonder if it's slowing down a little bit. 'cause people are realizing, well, it's expensive to kind of train and upgrade these things and when the LLMs become disposable behind all these platforms, and I'll just kinda wait and see whatever Apple or Microsoft or whoever it is or Google's doing, and it'll just become a capability that I use, but I might not care what the LLL Yeah, I think you, I, I was thinking this this morning. I feel like we plate plateaued on capabilities right now.

And you're right. 5 model was essentially disposable and they had spent, you know, tens of millions if not hundreds of millions dollars training these bottle. Um, where we stand today is, I think we we're at a plateau for, um, uh, um, probably the capabilities until the reasoning catches up.

Like right now, there's just not a lot more data that you can train these, these large language models that are general use on, and the innovation is probably gonna come out of smaller parameter, narrow trained AI models based on, you know, health or climate or finance or whatever, whatever deep vertical we're in. Do you think those small language models that you're talking about, can they all run locally or is, are they gonna be kind of distributed between the local machine and the cloud as well? I think what you're gonna have is some kind of router.

So it's gonna route based on your question and that router's gonna be in essence a large language model or small language model that resides locally. And then based on your preference, if you're asking about, you know, um, the best way to make a souffle, it might go to a general use meta or OpenAI. But if you're asking about, you know, how you should be considering retirement planning, you may use something a little bit more localized to you because that's where you have your private data.

Well follow my reasoning on this. There's a lot of things that I ask questions about that I don't want to know forever. I just have a one-time question.

Am I gonna route that question to an LLM that sits on my machine or a service or am I just gonna route that question to my, I don't know, tax planning advisors as LLM AI assistant and have them answer it for me? 'cause after I get my answer, I kind of don't care anymore. Yeah, It depends on if you trusted, right?

I mean, at some point if we let it go to, to AI and the bot, you know, talking to each other, what about security for that? That's just the first thing I would think of. I don't know.

I think you're gonna have the equivalent of the local desktop firewall that we had in the early two thousands, but it's gonna be the local desktop AI router. All right. So routing, networking, the revenge thereof, it's all coming back to us.

It's just that we gotta have another layer and a whole nother set of protocols. So for all I know, it's now the lay networking layer stack has, I don't know, nine layers, who knows. But we'll see.

Hey folks, enjoyed the chat, but we'll be back in a minute to move on to our next subject. All right folks. And we're back and we're now gonna talk about how AI is changing the way we learn.

There was a conference earlier this week hosted by Pega Systems and they rolled out what was an AI tutor. And the idea here is that, um, you will be able to ask this tutor open-ended questions about how to make use of their software, but they were making the point that the way we train people about how to use things is gonna fundamentally change. 'cause I'll just start asking open-ended questions into a natural language interface rather than say, you know, watching an online tutorial.

That's roughly the equivalent in my mind of going to traffic school. Um, but if you extend this idea out, I mean, why wouldn't it apply to everything that we learn, whether it's a college class or kindergarten kids or whatever it is. So Mark, what is your sense of, are, are we on the cusp of some fundamental changes in education?

Yeah, I mean, I, I think this is, uh, the beginning of learning management systems that are truly adaptive. Um, you know, and I think that it gives a lot of leverage to whether you're, uh, professional that's increased upskilling or if you're, you know, in the, you know, modern school system for children in education. There is, um, these systems are very good, um, creating inferences so they can infer from your question and your responses what you might or may not know.

And, and it's giving you a more personalized experience. So I, I think that you will have more of a, um, cohort type based, there are lectures, but then if you really are gonna get deep, and when you study, you have these, these tutors where the experience leans into the way that you learn. Yeah.

Because the tutors, the AI tutors are at the same time they're teaching you, they're learning the way that you learn. They're learning what resonates with you when you say, oh, I, I didn't understand that, understand that now I do thanks, or whatever you're saying to the, um, ai, they're learning the process of teaching you at the same time. Do you think this will help with, uh, overcrowding of classroom issues?

And I asked this question 'cause you know, I went to parochial school and there was like me and 200 of my best friends divided up into four groups and you know, the teachers, you know, if they could remember your name, it was a good day. So, um, you know, and they got to know you, you know, maybe by halfway through the course, they might have a handle on, you know, who was doing what to a certain degree. But, um, to Mark's point, if the AI knows more about you and can personalize the learning to where you're at, is that gonna create a less pressure on teachers and, you know, where does, where does the teacher kind of fit in this in your mind?

Well, we would hate to do away with teachers because just like with any a I wanna have the personal human touch, um, that we can only get with the teacher, but it could be supplemental, it could aid as a tutor as we were discussing. Or perhaps when a, a school situation, they're overwhelmed with so many students, we gotta keep them busy rather than just having them on their phones or are doing games. Perhaps they're doing something that's supplemental to their learning, Perhaps.

I don't know, mark, I was not the best student in the world and I generally always chose between, I could either read the books or I could go to the lecture, but I rarely felt that I needed to do both. Um, do you think that people are gonna just go for more self-taught and maybe lifelong learning and they're not gonna show up for like the classic classroom, I gotta be here at 10 o'clock on a Wednesday for an hour? I think that the fact is that you can time shift.

I think that we can add things that are attractive to people like gamification and presenting that material in a way that is much more consistent with the way we consume other information rather than formal learning. That it could be a good thing. Um, it would be nice.

I think there is a lot of value. Let's finally set in the actual human contact. But, um, in essence, there's very, there seems to be almost every education system, uh, you know, as you mentioned earlier, a one-to-many, uh, relationship.

And this, this gives you leverage so that you can have more one-to-one relationships and learning. Do you think there'll be less, I'll call it accidental discovery, right? So I'll get interested in the subject and I'll go looking for material on it and you know, whether I go to a bookstore or library, wherever it may be, um, but I'm kind of just poking around.

Do you think that the AI will be able to say to me, you know, if I'm interested in topic A and it knows that I have checked out, you know, A, B, C, that logically the next thing to do is de and f, which I might not figure out on my own. So do you think we're gonna get more guidance from the AI platform about well know what I don't know about something? Well, That's a double-edged sword as someone who's gone down these rabbit holes, because the more you look, the more comes up and then you start looking at something else.

So what I would like the AI to, to do is to keep you directed and to remind you your original question, your original search, and stay on track with it. You mean shiny object syndrome over? Oh yeah, yeah, It definitely is true with ai.

Mark, what do you think, uh, you know, can we kind of stay on the path more with AI or do we wanna wander off the path more? Anyway? I think that it's a really good question.

I think that AI will probably have a tendency to lead us in the di in one direction, and that may not be great. So I guess it's a matter of maintaining our curiosity to, you know, make sure that we don't just become narrowly focused on tasks and we keep getting creative. You know, I was talking to the folks at Pega about this in particular, the thing about gen AI is you kind of have to learn how to ask questions, right?

I mean, the core of the thing is it's not just sit there and it's not a search interface. It's really like you ask an open-ended question and you get some answers back. I'm not certain that most humans have been taught how to think about asking questions in, in a logical flow or kind of making all that work.

So do we need to kind of go back to all of our fellow human beings and say, Hey guys, you need to rethink the way you use this stuff. Or will they kind of discover it on their own over time? Or do we need to actually like train them how to frame questions?

AKA prompts? Mark, what do you think? Um, I think that the, the need to, uh, train them on the way to ask questions will be less and less important for an end user.

Um, I think that that really, um, and you know, it sort of happens today is is there's a probability and there's weights and biases in these models that we perceive as creativity, but it's just a little bit randomness. So I, I would assume that we will understand that if we start getting too narrow and refocus that we'll just insert a little bit more noise into the system. And it wouldn't be a matter of prompting, it'll be a matter of system design.

Well, one of the issues with prompting is that the memory is, is it, do you have to start over every time you're re-prompt? Because it seems sometimes that's an issue. So, um, I think that you can fine tune, there's all sorts of sources of information on how to fine tune your prompts, telling the, uh, ai you are so and so, your, your job is to do this, even though the tone that you talk to the, uh, AI in is, uh, apparently important.

One of the things Mike, I thought was interesting with, um, the way you prompt ai, and I didn't know this, 'cause sometimes I'm not as nice to it as I should be, but apparently if you say please do something, it gives it the option to say, well, no, I can't do that. But if you say it in a more direct tone, do this, then it does it, it's just a little tidbit on prompting for you. So if I ask how to open the bay doors nicely, I'll get a better answer.

Yeah. Or it, you may have to ask again. It's just very interesting how it, it responds to human language because, um, direct something very clear, um, gets you a better result.

But one of the other things I wanted to mention is, as we get new versions of, of different ai, we're seeing better memory capacity where you can save what you previously, previously prompted. So that, that's something that helps Mark, walk us through that a little bit because I'm not sure that people realize there's this thing called the context window, right? And the context window is only so large and has so much memory and it's tied to the parameters of the LLM, but can you walk people through a little bit of how that works?

Yeah, yeah. It's, it's, you know, the, and, and honestly, even though these windows have these formal things, I do notice that, um, it's like having a dinner conversation. You may remember at the beginning that we introduced each other's first name, but after two and a half hours, you may lose track of that beginning of the, uh, um, conversation.

And it's the same thing with LLMs. Um, it's probably the closest like analogy that most people might understand is, is it's your random access memory for having these conversations. And, um, some of these windows like Chachi, bt, um, if you've paid for that version, I think it's 30 2K tokens and tokens are what words are created are converted in so it can understand what your language is.

And then you have Google Gemini that is going up to 2 million, but even then it's a matter of how much of that conversation, um, makes sense and it puts all those things in the right order. So, um, yeah, it's, it's, it's, uh, fascinating and deep AI tech that I know how it works, but I don't know how it, so it's, it's definitely a, a machine learning, uh, 300 level class. Are you at all concerned about what you're sharing in these things?

I Thought about that. I was like, nobody knows me better than AI now. I mean, it's so true because it really knows everything you're interested, even more so than Google, because think about when we do a Google search, we're just putting a couple words a term.

But when you're, when you're messaging with ai, you're getting very detailed and the AI is smart. It can put together everything that you put and say, this is the the person that's, uh, prompting me. So yes, there's definitely that concern.

Mark, do we need to kind of educate folks about this concern or will we be able to put some sort of guardrails around this eventually and maybe not as be concerned? 'cause maybe the LLM doesn't remember us at some point? Uh, I think that it's hard to unsee a train wreck, Mike.

Um, it's just, uh, um, it, it is designed and especially with chat GPT, they just added a memory function is it's gonna perform better by remembering the context of your conversations. And you're gonna put it in, in there because you want to process data. But I think people just don't understand, and we've seen it time and time again in the news, is that people are, lawyers are using chat GBT to write briefs and it hallucinates also, you know, they were also probably breaching client, uh, attorney privilege.

'cause you know, you are putting that into a system that may or may not be secure and you've reached, you know, reached that, that, uh, trust. So I think people really do need to understand that they are, you know, may or may not be using a privileged system. That's a great point, mark, especially with the hallucinations, you always have to say, is this true or did you make it up?

You never know. So it's a, it's a good thing to check. Well, To that point, and we talked about this on an earlier show, but who's actually deciding what truth is?

And especially when it comes to stuff like climate. Yeah, I mean, there's a lot of arguments about, you know, what's true and what's not true. And each side might go train their own LLM for their own specific belief system and just reinforces whatever you wanna believe.

We're already seeing that in politics. It try, try asking about what, you know, different politicians, you'll, you'll get, uh, definitely some indication of how this LLM was trained. So I think that's already there.

Yeah. Mark, do we need to vet the LLMs for, I guess we call it bias, but ultimately, uh, you know, are people just gonna gravitate to the LM that tells 'em the, what they wanna hear in the first place? Yeah, that's a great question.

I mean, we, depending on the technology, um, you, you could see a 30% or more failure rate on truthfulness. Um, and the way that you make sure that they are truthful is that you ground them. But who do you choose?

Do you choose, you know, flat, earth, earth to, to ground us on or GEs people who believe in science to like understand how the world works. So, um, I, I do think that, I like the approach of perplexity because it adds citations and then you can at least know where the, the, uh, data is coming from. Then you gotta vet the citations, which you know, a lot of times is someone who is very good at search engine optimization and not so much at, you know, the science of climate change or you know, the geom geometry of the earth or, you know, reality in general.

That's really good advice and vet it always, even if it looks like it has a source, you have to go and and check. I agree with that. To Bonnie's previous point, you can always say at the end, provide the citation for chat.

GPT specifically provide citations for the facts you gave me. Yeah. And sometimes it's great and sometimes not so much You catch it.

It's made it up. All right folks. Well, I don't know.

It's gonna be challenging to be a teacher going forward. I mean, ultimately a lot of them would say initially they had a negative reaction. Some of them were positive, but this stuff's not going away.

It's here, it's pervasive. So I don't think you can pretend it doesn't exist. Some teachers I talked to are getting rid of, uh, written exams because they're like, you know, it doesn't make sense.

The kids are just going to use it to use chat GPT anyway. So they need another way to assess knowledge. My other question to you though, on this last point Mark, is so in the real world I'm gonna use, you know, an LLM to help me do whatever task it is.

But should I be tested on my knowledge of that without an LLM or do we need exams where everybody brings their own LLM, like it's an open book test now? I dunno. You know, and I'm really on the fence on this 'cause you know, in college I felt like it was silly for me to memorize the periodic chart because I knew that if I had a job, I had always had access to it.

Um, I think that you, you really want to test for problem solving and critical thinking and the ability to retrieve the knowledge that isn't, um, necessary to be, um, right there at hand. But, um, I also think that you have to understand that, uh, you know, you have to be trained on the tools so that you can be effective. If this becomes a defacto part of our society, you know, you wanna make sure that people know how to use this in a way that is, uh, useful.

We're all gonna find out together. 'cause you know what, this is one giant social experiment that we're all participating in and we'll see how it goes. So we'll continue to talk about this subject going forward, but we'll be back in a minute with our next topic.

I'm Bonnie Schneider, sustainability contributor to the Techron Group. I'm excited to introduce you to a groundbreaking new initiative from Techron Research, the sustainability pulse meter. The pulse meter offers valuable insights into how environmental responsibility factors into tech purchasing decisions for key players in the industry.

Position your company as a leader in the industry and differentiate from your competitors with the sustainability pulse meter offered exclusively from tech strong research. All right, and we're back continuing our conversation about ai, but we're gonna geek out a little bit. Uh, AWS had a conference on security this week and they were showing how they were using generative ai, at least in a beta, um, to access, uh, this data lake that they provide for cybersecurity folks and IT folks.

And the idea here is that I'll just be able to launch queries in natural language against this data lake, which will tell me where my issues are, provide summarizations, generate some advice and some recommendations. And what was interesting about this mark was that they were talking about in the context of using it across security and IT operations and DevOps folks. And I wonder if we're on the cusp of maybe AI is gonna, uh, break down all these silos that have existed within it for so long and maybe we can get to something that feels more unified.

What do you think? Um, yeah, I think that we'll have these, uh, action models that, that our operations and development folks can use. And I think it's, it's just by the very nature of, um, doing knowledge graphs work and some of these other things is um, we'll start to connect the dots and the AI will help to surface issues that probably cross boundaries.

So, I mean, I think it'll help be helpful. Um, but I don't know if it's a slow burn or a uh, you know, great flash. I wonder because we have so many specialists that have grown up in it over the years and it does feel like we're moving towards more of a generalization of functions.

I mean, ultimately if I'm a developer, I care about security, I should be able to go to my AI assistant for security to help me figure out something. And today we just inundate these people with a million alerts and they don't make sense of it and they throw up their hands and they go at somebody else's job. Um, do, I just don't know if we're gonna like see all this condensed into one kind of person can do all this stuff with a bunch of AI assistants or if it's just gonna make it easier for us to collaborate and across our specializations.

I mean, what's your I Think both because, uh, the AI is gonna help us collaborate better for sure because it'll just allow a quicker process of communication internally and, and within developers and within other people in the company. Um, but uh, it is all coming together like we were talking about earlier. So there will be more all in one I think.

I think it's gonna be both, right Mark, do you think we're gonna have as much need for entry level folks in it going forward? Or is the entry level job now essentially the AI job? But if I never got trained on the entry level stuff, how do I become expert enough to become the the architect per se?

I mean it seems like there's a a bit of a challenge here in terms of just getting new people into the field. Yeah, well I think that there will be sort of, the entry level will be more generalists, but for us to keep getting better at this stuff, what we want to do is free up the time so that people can specialize. And so in the longer term, I think the IT folks in later in their career will be specialists, but we probably almost always will start out more as generalists just because that lower level of it will be, um, sort of abstracted away by ai.

That's my guess at Least. I think also the, the entry level, let's say the younger people, they're gonna know AI better than some of the senior level people. So we'll see if it's still entry.

This is true. I also wonder, mark, do you think we'll see less burnout in both the IT and the security sector? 'cause one of the dirty little secrets of this business businesses, there's just a massive amount of, uh, tedious tasks that people have to do every day and there's no joy in the job.

So do you think more people will hang out longer just because it's gonna not gonna be as much drudgery as there is today? Um, it depends. I think ideally that's what would happen is that you should be able to get more done and less amount of work or um, less time.

But as we know, the wheels of progress and especially profits, you know, have pushed us to things like the 80 hour work week where people wear that as batch on. So, um, hopefully our, you know, machine masters will teach us that they can do that heavy lifting and we have a little bit more, a better quality of life, Right? That that 80 hour work week is a badge of honor, gets really old when you're married and have children.

Lemme tell you, um, I wonder though, as we kind of evolve here, um, will the, that entry level function, I mean, where does it get taught? Is it gonna be at the school now or is it gonna, because there used to be a lot of on the job training, but I think it will be easier to onboard people when they come in 'cause they'll be able to ask questions and summarizations about the event. But I wonder, coming back to our earlier questions about education, will colleges and all these schools need to raise the bar for what is considered, you know, a, a reasonable set of skills to have when you graduate?

Yes, I think they will. And, and, and you just, um, had sent me an article this morning about, um, what, uh, local University of Florida is doing with engineering, with, um, with, uh, next Area Energy. So they're, they're training their engineers now to come up with solutions for the future.

So I think that that is gonna happen on the college campus and maybe even in the high school too, as they prepare for college talking about, uh, what you need to know and ai, what an entry level job looks like. And I hate to go back many years, but think about when we were in high school and wanted to go into TV production. I remember, uh, taking an integrated class with New York Institute, that technology, of course it's a little data now, but on video production.

So it kind of leads you into that, that role. And then when you get to the point where you get entry level, it's, it's less surprises. So I think that will continue.

Uh, Mark, there's this whole, you know, it's a billion dollar cottage industry around certifications. You know, everybody's takes one, they get one. 'cause the HR department's looking to see you have a certification before you can apply for this job.

In the age of ai, do certifications become less relevant because, well, I mean everything is, uh, a simple prompt away. Yeah, I I think that the, the narrow deep dives may not be as relevant as they once were, like the windows, uh, certified professionals type of things. But there'll always be, uh, a need to qualify people, especially outside of it.

It is just gonna be so impacted by this. I don't, I think there's still need for many types of jobs to understand the regulatory and the procedural. I I agree.

I but here's the thing, Mike, if we take away the certifications, what are people gonna post on LinkedIn? LinkedIn like that? That's probably true.

Yeah. I wonder though, over time, um, might we not come to a scenario where, um, I don't really have my own personal IT staff. I just have a bunch of AI folks who are my or agents who are my IT representatives?

As far as I'm concerned, I'm, uh, I may say to it, I wanna build a website, I want to have this amount of content in it, I want it organized in this way and it will just do all that. There may be some it human at the back end of that, but I wonder if those people will ever engage with actual humans again, because there's gonna be this layer of AI agents between them and, you know, the regular folk. And some of them may like that.

A lot of it people I know are in it because, you know, they like machines and cats in that order. So, Well, I think that, that it's gonna change pricing. And I'll give you an example.

I can remember five or six years ago I wanted to build a website. I had no idea how to do it. And I got a guy to do it and he was like, oh, this is gonna cost X amount of money.

Now there's AI that will build the website for you. I don't know how good it will look and it, it's not as customized, but um, it is gonna change the cost. So what we pay, right?

I it, I believe, I don't know, mark, you think it people will become invisible in the age of ai? I think we'll just move up in complexity. I mean, if you think about it, um, early on we made around close and then we mass marketed in and people can go to Nordstrom where they can go to Target to get their close.

But there's still, uh, plenty of folks that want tailored solutions. And I think that's where, where the opportunity is, is that we can get past that 100, 200 level of time consuming, uh, things like making a website. It's not everybody, everybody who wants a website should be able to have one, but only Bonnie can have her unique going climate change and how, um, how we should be doing that.

And I just think it just we're, it's a, we're shifting ourself to doing more deep, meaningful work. And, um, I'm super optimistic, which is not typically my personality, but for this, I think that there's so many things I've wanted to do for so long and I couldn't because I'm doing websites and I'm figuring out how to make the best graphic and I'm just learning how much content goes on there and optimizing it for search and all of those things are not, not all that valuable. What's valuable is, or what I hope is valuable is my opinions and my thoughts and the way I organize my ideas.

All right. Well I think the most important thing is to make sure we all have plenty of hobbies. 'cause we're not sure exactly how all these job functions are gonna evolve.

It is only one of them, but we're along for the ride. Hey everybody, thanks for watching and hanging out with us. And I also want to give a plug to Bob Ruman who's coming on next to talk about the latest and greatest issues of the day that are probably rubbing him the wrong way.

And once again, thank you all for being here. And Mark, thanks for being on the show. Thanks.

Good to see you guys. Great To see you. And Bonnie, as always a pleasure.

Oh, as pleasure as well. All right, thank you all. We'll see you next time.

Hi. Okay. Uh, this is a story about epistemology and epistemology is the branch of philosophy that deals with the nature and limits of human knowledge.

Tenant wade, the story goes like this. Uh, last week I got a, uh, message in a LinkedIn chat window asking me for my opinion on my leadership, uh, skills. And I responded, I was flattered that somebody would ask me about my leadership skills, whatever they are.

And uh, I responded. And then another message came in through the LinkedIn chat window asking me, uh, another question and another question, another question. And I responded.

But after a while, I got the feeling that it was not a human being asking me the questions. Was this really AI driven? I really had no idea who the person was.

So I asked the, uh, sender are you are human? And of course, the uh, message was, of course I am. But you know, the reality is I really had no way of knowing.

I didn't know the person. I never met them in real life. And the only way I could tell if they were real is if they showed up on my front door and introduced themselves to me, or if they showed up at the front door of a friend of mine who I knew in real life and made an introduction.

Otherwise, all it is is just data in the chat window. Which leads me to the point about epistemology. Uh, and epistemology really is about how do you know what you know?

So imagine the scenario. Let's pretend I have a cup of coffee and I wanna know if the coffee's hot. And one way I can test the coffee's temperatures to put my finger in the coffee and feel it.

And if it's warm, I know the temperature's hot. Or the other thing I can do is get a thermometer and put the thermometer in the cup of coffee and read the temperature of the thermometer. Now, this assumes that the thermometer is reporting the temperature accurately.

I am depending indirectly on the thermometer to determine knowledge. And this is really significant because these days the world is a terribly complex place, but a lot of stuff rolling around and we don't interact with most of what we come to understand is the world directly, rather we understand indirectly. So what does this mean?

So imagine I'm watching a YouTube video and I see a car crash, right? Well, how do I know it's really a car crash and that it's not, um, some models or some AI generated scenario? And the fact is, unless there's a thing called provenance, I don't.

And provenance is how one determines the art origination or authenticity of something. And provenance technology is a growing, uh, initiative. There's a lot of provenance technology coming down the bike, but there's more needed.

Um, but an example of Providence technology is that, um, Reuters and Canon, the camera manufacturer, are getting together to create a technology that gets embedded directly into a Canon camera. And once a photo or video is taken, the origination information of that data is sent to the blockchain. And once it's on the blockchain, it's immutable and publicly verifiable.

And that's an example of provenance data. Using provenance technology and provenance is very important, particularly as we go into these realms of anything could be anything at any time. It's getting really pretty amorphous out there.

So, but this is leading me to think about something else. And it's this, I mean, does it matter that we're living in a world in which, you know, computer driven special effects and AI powered interaction might have more credibility than reported facts? Who I, I don't know if it matters.

I think it does, but you know what? No matter what, it's definitely worth thinking about.