Root Causes 519: AI Is the Room

Hosted by

Tim Callan

Chief Compliance Officer

Jason Soroko

Fellow

Original broadcast date

August 18, 2025

AI is not the elephant in the room. It is the room itself. Jason explains what he means by that.

Podcast Transcript

Tim CallanSo let's talk about the elephant in the room.

Jason SorokoThe elephant in the room, Tim, and remember, this is the Root Causes podcast, a PKI and cyber security identity podcast. Do you see AI in that title? No. But we talk about AI a lot.

Tim CallanWe do.

Jason SorokoIt's because it's the elephant in the room. I'm gonna make this episode a declaration, Tim. Sometimes we have these episodes where we make a declaration. Our declaration here today, between Tim and I is AI is not the elephant in the room.

Tim CallanSo it's not the elephant in the room.

Jason SorokoIt is the room.

Tim CallanAI is not the elephant in the room. AI is the room. Got it. Explain what you mean.

Jason SorokoSo I have fully embraced artificial intelligence into my digital computing life and a lot of people, when you and I were first talking about it, it was about prompts. What can you do from the prompt.
We're very early in creating synthetic voices of ourselves in order to talk about deep fakes and all these kinds of things. So you and I were really, really, really early, but you and I were doing generative text, generative audio, generative video. We were doing all of that stuff. And I think that that is most people's experience with AI right now. I wrote a Substack article not long ago that actually talked about Tim are many of us becoming technical geriatrics? And here's what I mean by that.
I have spoken to extremely smart, extremely experienced and extremely well-meaning people my age and around my age, who, for their entire professional lives you would never call them a Luddite, and you would never, ever suspect they were anything other than capable of picking up a technology and using it like any tool in a toolbox. Like if you were a welder and you were amongst a bunch of other welders, and you had a welder who just was confused by welding equipment, you'd be surprised by that.
I'm seeing that now though in our world, which is, it is an incredibly sophisticated tool set for us and I see a lot of people who have traditionally been right on top of their tools, and can't use these tools and don't understand them. And let me tell you how that sometimes sounds.
I'll be describing my current work within creating a second brain for myself, creating a rag system, or a graph rag system, and being able to not just create, going to any of the models that are available to me, and being able to do generative AI based off of whatever it's been trained on, or whatever else. I'm talking about, really, really sophisticated rag systems that are just proving themselves to be incredible right now. And I'm describing that to a colleague, and a colleague might say, hang on. Do you mean, like, just ChatGPT. And I was like, within that moment, it made me realize, oh, you stopped being curious after ChatGPT 3.5 or 4.0. You’re not aware that Open AI, and there's so many other models out there right now. Anthropic’s Claude system and Google Gemini and they're all incredible. And I use them all for different purposes. And my response back to that person is just a bit of shock. It's like, wow, you're a welder who now doesn't understand new welding equipment.
And I have a theory that Apple's botched implementation of AI, and I think we can call it that. They say they're going to improve it. I haven't seen it yet. I have no doubt they have smart enough people to do it. But Apple's botched AI implementation, people's initial curiosity around early ChatGPT and the early models, and then Microsoft Copilot equally being its initial, here's the cool things you can do. And then people actually using it, it didn't live up to its hype.
And, so I think a lot of people were initially like, okay, the uptake of it was massive. That was shocking, in terms of just, and it was great to see, but people didn't follow through in the way that I would have expected. They kind of lost some of their curiosity around it.

Tim CallanWell, I think a couple things. I think this statement you made about not living up to the hype is, first of all, accurate and, second of all, incredibly important. I remember when I first started playing with these tools, it was amazing. Like you could tell it to write a sonnet about a refrigerator, and it would. And you’re like, I just wrote a sonnet, but then if I tried to, like, get it to make the sonnet better, I couldn't. Or I could tell it to make a picture of something of a person, and it would do that, but that person would have seven fingers. And no matter how hard I tried to make it have only five fingers, I couldn't. Just couldn't get it to make only five fingers. And so then you start to say, okay, it sort of has this real gee- whiz factor. But when you start getting into pragmatic, real world usage, it turns out to be just so much more limited than the original gee-whiz factor.

Jason SorokoThat’s exactly what I'm talking about. Exactly what I'm talking about.

Tim CallanAnd that's the hype thing.

Jason SorokoExactly. But here's the gigantic, however. So people, a lot of people - not all, but too many people have not kept up with the latest models and the improvements.
I'll give you the one that clinched it for me, and the reason why I said I'm now an AI first welder, if you will. That's now going to be at the front of my bag, not the back of my bag, in terms of my tool set. And part of it was Open AI - I call them out – Their 03 currently, the full 03 is inventive.
In other words, I stated once on this podcast with you, and I don't think anybody liked me saying this and tough, because it was true. It's not true anymore. AI at the time, generative AI, was math plus plagiarism equals what I'm looking at.
Therefore, really what I was saying was, AI is not capable of original thought. Well, thought is a big word, so let's not even use it. Capable of original ideas.
03 was the first time I saw inventiveness. Like this didn't come from anybody else. And I think that some people in certain fields are actually starting to realize there are specific models used in specific ways, where it's not just generative text, generative imagery, generative audio, but generative of original ideas, and that has changed the game. Now, is it always right? No, you still need a human being to be in the middle of it. But this is part one of a part two story, Tim.
Part one is once it showed me inventiveness, even if 5% of the time it's inventive, that's incredibly useful to me.

Tim CallanAbsolutely.

Jason SorokoEven if 95% is garbage, that 5% is something else, because that propels me forward, in ways that I couldn't have in any other way.
03 though has another bag up its tricks, and so do the other models now as well. Some of them are catching up. But 03 is in itself an agentic AI. And what I mean by that is this. It uses tools itself. You can actually watch as 03 is reasoning and trying to generate output. It will actually write its own Python code on the fly to actually form something for you. And it has other tools in its tool bag as well. And it's very inspiring to watch.
And so Tim, here is my statement for this podcast, and then I'll get your reaction and we’ll leave it there, which is, it's not about the generative aspect anymore. It is about automating everything.

Tim CallanAnd really, I mean, again, generative AI, it’s very gee-whiz. And it’s a great way to introduce somebody to the idea, and it's a great way for all of us to get our hands on it. If you have a browser, you can use it. You can get results. You can get text. You can get pictures. What's harder, though, is what I think is ultimately going to be the bigger and much more transformational use of AI, which is that it's agentic, and it's incorporated, just incorporated into our systems and our processes.
It’s the difference between me asking the AI to give me a set of instructions for how to do something, whatever. Build a machine. And asking the AI to just operate the robot and build the machine.

Jason SorokoAnd that's the point. It’s not just we're getting there. It's for as an information worker, it should now be the front of your tool bag.
And I would say that anybody who's not incorporating MCP servers, A to A servers and various other toolsets, and if you're thinking to yourself, that's not really for me, or, what's worse yet, I'm not really a coder. That's the one I hear all the time. I'm not really a coder, so I'm not going to use things that I might have to get my hands dirty on a command line and type in something in order to make the computer know what to do, what to point at, how to hook up the server, things like this. I think that Tim, you're going to hear us talk a lot more about AI and from a fundamental level, because I think we have to start inspiring people to use AI tools at the front of their bag. I think, Tim, it becomes incredibly important now for information workers who, if you consider yourself not a Luddite, you've got to learn some basics.

Tim CallanBut don't the AI tools just become tools.

Jason SorokoThey will.

Tim CallanAnd I'll give you examples. Number one is like, okay, so everybody, we all use AI in a non-generative AI fashion if we do any kind of search queries, because the top result you get on both Google and Bing is an AI- generated summary. And it's just there. And nobody decided to use it. They decided to use the same tool they're using before, which is Chrome, or which is Google search or Bing search, and the first result I get is AI telling me a thing. So there's a good example. There's all kinds of things, like software that it's just using AI now, and you don't choose, you don't consciously choose to use AI. You don't hit an AI button. You just get results that are AI enhanced.

Jason SorokoThe audio software that I use to process these podcasts is entirely AI, and you wouldn't know it.

Tim CallanAnd it's invisible to you. It's behind the scenes. And maybe there's styles and things you can change. But the AI is just part of the tool. It's almost like the way programming languages change over time. If I'm an end user of a piece of software, I don't perceive that the language it was written in was different, except in so far as the end result is different. The product that came out is different, and I think that's how the vast majority of the humans are going to use their AI most of the time. It's going to be a bespoke tool for a different that for a different purpose, and that tool will just be better.

Jason SorokoHere's my argument though, Tim, I hear everything you're saying. I completely agree with all of it. In fact, the conclusion to it will be, we probably will not just browse with a browser less, because of the fact that we'll be searching in different ways and consuming data in different ways. I think what we're going to find is, take a look at what some of the big AI companies are doing in creating physical form factors, glasses, etc. Eventually, human machine interactions completely change. Even mobile devices themselves the way they are. You might use them differently or less.
My own patterns of usage of search have completely changed with AI because of what you just said. But let me go back to my original argument and what I said at the very top. If that's all it was, then AI as of right now is not the room. It's the elephant in the room.
What makes it the room is when you push it a little further than just - -being fed AI on a silver platter by the big tech companies, if that's all you're doing, it is probably only 10% of what's capable to you right now. And that's the truth.
And so let me make my argument, if you're willing to just learn some more and even start to investigate no code orchestration systems. Learn how to hook up an MCP server to it. Learn what those acronyms are. I’m not even gonna get into that right now. Hook up a Docker container that's basically running some of these servers for you in the background and so easy to install. Plug and play software at the server level. I'm not talking about for my mom. I'm talking about for knowledge workers, technology professionals who are somehow allergic to some of these technologies that accelerate you massively. Now you're right in saying, eventually, all this stuff will just be incorporated into all of our tooling, but we're not there. What I am saying is this, 90% of why I'm excited has nothing to do with what I'm handed on a silver platter right now.
And for those of you who want to go on a journey with me, I want to do that with you on this podcast. And this isn't just to go into general AI and general automation. We're going to keep it to PKI. We're going to keep it to automating credentials, digital identities, those sorts of things. So we're going to keep on point in terms of the subject matter of this podcast. But if, if it is the room, and we're going to declare that, then let's make it the room. In other words, I want to talk about PKI, digital identities, certificate lifecycle management, all that good stuff in context with AI. And not just simplistic silver platter productized AI, but a little bit of exploration of what is out there for you that doesn't require three PhDs in computer science. It just requires you to not be a Luddite. So that's the intro opening to what hopefully is a lot more podcasts on this subject.

Tim CallanAll right, well, we've already had a lot of positives generally on the subject of AI, but I like this.

Jason SorokoWe've just begun.

Tim CallanWe've just begun. All right. Thanks, Jason.

Stay informed with expert insights

Subscribe to Root Causes for engaging discussions on PKI, digital security, and best practices for protecting your organization's critical assets. Don’t miss an episode!