AI open source clones, jobs and deepfake fraud

Hello everybody. Welcome to the Techstrong gang for Tuesday, where chaos reigns once again supreme, and we're going to dive into all these fun topics of the day. But let me introduce our guest, Kate Scarselle.

Good to see you, Kate. Nice to be here. All right.

Chris Blask is back. How you doing, my friend? I'm loving life.

Excellent. And Sid Nag, how are you? Good morning, Mike.

I'm great, thanks. You appear to be back in New York again. Is that true?

Yes, I am. Thankfully. I don't like to be traveling anymore.

It's kind of old news. Yeah. All right.

Well, one of these days I'm going to have to get on Metro North and come see you, Sid. You should. We can meet in Grand Central.

There you go. Right outside Grand Central, lots of good cafes over there. Hey, I'll come from Boston and come see you guys, too.

Excellent. My favorite, Persian Cafe, unfortunately, is closing down. I don't know if you heard.

Oh. Okay. That's too bad.

Okay. Which one is that? Persian Cafe?

Yeah, it's the one across the street from Grand Central. Historically, it's been my remote office where I meet a lot of people, because then I don't have to go far from the train. Smart.

I guess I haven't been you there enough, so it's closing down. That's part of the problem. But there's fun things happening in the world, and let's get started.

There's this software developer, I think his name is Brandon Thomas or something like that, who has created open source clones using AI of Adobe software, now Microsoft software, and even World Graph, the game. " Kate, is this going to be the new thing? Am I going to wake up every other day now and there's going to be a clone of some application out there, and will the entire financial model on which this industry rests start to fall apart?

Just saying. I think it's quite possible, actually. And I don't think it surprises any of us, does it?

To some degree, I think we were expecting this. So yes, developer Brandon Thomas illustrates that. So he used Claude to create open source alternatives to Adobe, just like you said, Microsoft Office, AutoCAD.

AutoCAD was really super pricey, and now getting these free and these commercial products. I always think that they're, and Chris, we talked about this too, this democratization of software and of hardware, too, and I think that this creates a totally different economy, and I don't think that that's bad. So yeah, this is what's happening and hello.

Let's see it. Yes. I don't know.

Sid, let me ask you about this, because you live here in New York, as we were just talking about. So on Friday evening, I'm walking down the street in Bronxville on my way to dinner, and I overhear a conversation taking place on a street corner where people are talking about this Adobe open source package. So it's like a mainstream conversation already.

I was floored. I had to take a double take as I'm walking down the street. I'm like, wow, people are actually talking about this stuff on street corners.

So is this going to gain a boatload of mainstream traction? Or maybe it already has, and this is a new reality. Yeah, I think the price is right.

That's what they say. So let's see how that plays out, but I have my own doubts about how this is going to... It certainly affords more customization, portability, customer control and all of that.

No doubt about that. And yeah, the price is right. It's going to create heavy duty competition for license-based and subscription-based software, depending on the software of your choice that you've been using.

Okay. But I think AI-generated applications in general require much more rigorous testing that they're currently going through in the world of AI apps. There's vulnerability remediation issues, there are cost issues.

By cost, I don't mean just purchasing the software, but it's more about maintenance costs or running costs, deployment costs. There's not enough knowledge out there to take these things and just throw it in the enterprise world of robust application environments where people are running mission critical. And I'm not saying Adobe is a mission-critical application.

Maybe it is for some, maybe it isn't for others. But things like migration, training, infrastructure support, all of that come to mind. So I think there are a lot more issues than meets the eye.

And so I think those are the things we ought to tackle before we could all jump headlong with the lure of AI-generated applications that we've seen where the price is right. Yeah, I'm sorry. And I know Chris wants to jump in here, too.

So yes, from a cybersecurity point of view, where is the trust and everything? But the only thing I'll bring up is that vulnerabilities are all over even what we're paying for. So does it really matter if we're paying a lot of money for these applications that are full of vulnerabilities, as well as not paying money for...

It's sort of the same thing. There's still vulnerabilities. We're not getting this right, so we might as well not pay if we're not getting it right.

I think Kate just said we all suck equally. All right. There's one other thing that I want to throw in, which is a lot of these applications today are becoming, I hear the word workflow productivity a lot these days.

It seems like many of these AI-generated applications are not just applications by themselves, but they're very agentic-powered, if I may use the word. And in that sort of vein, these agentic-powered applications are not monolithic. They're basically connecting to other applications, be it calendars, emails, productivity applications, video applications like the one we are on right now.

And they're building knowledge graphs and making the end user very productive. So when you start to think about these AI-generated applications, you also think about how far can they go deep into the workflow experience of the end user, and I think that matters. So I think building monolithic applications with using AI agents is not the answer.

All right. Chris, theoretically, there could be more than just one developer working on this. There could be a community, much like every other open source community that's out there, and will they address or help address all these concerns that Sid brought up, especially because, well, they can work on vulnerabilities, they can fix vulnerabilities, they can update the applications, and we've seen this model kind of work reasonably well.

And who knows? Maybe there's an enterprise support package from somebody who provides support for open source software today around these applications. I don't know, is that feasible?

Well, yeah. So the regular Tuesday watchers of the show can go back, particularly this calendar year, and stitch the comments together. So I know that I said multiple times that all I need to know to generate your application is the functions, the brochure.

What does it do? And I'm assuming that, not having looked at the topic of today's blog, that's what they did. " And anyone can do that.

We had to plan for anyone being able to do that. And I can. And from the early part of this year to now, I can take almost any feature set and express that in software quite easily, but with a non-zero point.

And to get to Sid's point of this, I don't have infinite amount of time and energy. I got enough thing to do already. I don't want to.

I'm happy to pay someone else to actually take that, and I think that's persistent. I think Kate, you're right. It changes the economy, but there's still economies.

And it comes down to what, Mike, you and Sid were just saying. Just because I can generate some functionality doesn't mean that I want to rely on it, and it certainly doesn't mean I'm going to give it to anybody that depends on me, a client, a customer, whatever. That means, Mike, maybe where you're going, maybe there are open source communities for some sets of this, but I think there will be companies, organizations that are monetizing the maintenance of these functions, but it's not going to be the 1980s class lock-in of vendor licensing forever.

If that's your business model as a vendor, think again. And just to Sid's point, the last point, the monolithic thing, people are getting this idea that my AI, the AI, is going to do everything. I can tell you, the laws of thermodynamics and efficiency and focus don't permit that.

You cannot put infinite amount of information into a single AI system and get anything out of it. What you do, you were saying, you have specialist systems that know a lot about that streetcar or whatever, and they do their jobs, and it's the coordination of the systems of AI systems. But I said enough, so I'll stop.

Well, theoretically, if I understood you correctly, an open source community could build a bunch of AI agents that are targeted to those specific functions, and then there's individual little teams, AKA working groups, that are updating those agents on a semi-regular basis, and that may all work at the end of the day. But the part that I'm scratching my head about is I get that there are enterprise customers out there that may not feel comfortable with this, but how much of the software that we're talking about here is actually consumed by average Joe end user who may not care about any of these issues because they're just people and they're running Adobe for some job or task, or they're a small one-person shop, or they're using Microsoft Office Suite because, well, that's what basically came when they bought the computer, but now they're looking at the licensing fees, and there is this thing in the world that everybody's talking about. It's called affordability, and if everybody can save a couple of bucks, how much does that ripple through the software economy?

What do you say, Kate? I absolutely agree. It's a for real thing, and we've seen this throughout the decades that we've been in this business where free is real.

When you compete, from an IBM perspective, we could never beat free, even regardless of how many features we would show that cybersecurity could do. At the end of the day, the tools matter, and the way that they are configured, and yeah, it matters. Because I noticed that when we would go in and present this package, at the end of the day, you still need people to implement it, and that's what I see.

What I see is that the people matter, the human in the loop actually matters, and that if you can produce something for free, you're going to use it because you're still going to need the human hours to get things working. So affordability, it matters. Free, it matters, and I think free, well, is the way we're going.

Only because of necessity. And everybody needs to play right now, whether they want to or not. And so in order for them to play, they're going to try to get what they can in order to present themselves this digital presence that we all need.

Yes. I think there's one other element, which is I think free is great. If it works, it's great.

But in an enterprise environment, free is great until the IT support guys get the call, right? What happens then? I just remember in my previous job, when I first joined the company, everybody got a standard issue of a IBM or Lenovo laptop, right?

But there was a few of us who wanted Macs, so we went and got Macs, right? Our own Macs. And then we realized, my God, how do we support this environment, right?

And so we formed a support group within the users like us that supported each other without IT getting involved, right? So that worked for a while, but that was kind of, when I get to it, I'll respond to your question kind of thing, because I still have to do my day job, right? And then when more and more people got Macs, then I could start to support it, right?

So I think you're going to see that kind of evolution is- And think about phones. Phones went the same way, right? " Then everything started to go onto phones, and that became a whole security nightmare for us.

My point being is that people want to play with the tools that they want to play with, and if it's free, by golly, they're going to be playing with it because that's what they want to do. That's who we are. We're curious humans.

And Sid, to your point, I'm just going to route that ticket to my AI agent and go back to playing Warcraft. What are you talking about? There you go.

There you are. There's no such thing as a free lunch, right? And to be clear, the open source world is not perfect, not at all.

And we all do open source things for reasons. Because sometimes, because we're hoping it leads to a job, sometimes we want to make the world a better place, sometimes we want to contribute or be seen to contribute, right? And this conversation just brings it up again.

Open source is wound all through commercial products, but it's not really ethical and epical. The economies don't necessarily give back to the contributors. And same thing, I look at standards and working groups, there's a lot of what we would call free volunteer efforts, but we're really just not counting the costs.

And this sort of thing, I know we're at the end of the block, but Mike, your last comment about we deal with this every day. Small businesses, it turns out the average car dealer has, I think, 17 subscriptions, right? A little business has seven or eight or nine, Google this, that, and the other.

I think that number will go down. A lot of that function may be, quote-unquote, "free," but it's still wrapped into some support structure, some vendor. You're just not buying 12 different things because you need to, you're buying two.

Yeah, I think the one other point is that open source is great, but when you start to build commercial products out of that, it gets extremely hairy, right? I actually remember I was involved in building a cloud offering using an OpenStack, if you guys remember OpenStack, right? It was very attractive because you've got the whole stack, right?

It was open source. But the difficulty is not the software itself, it's like all the different terrains that sort of spawn from these open source software structures, and then maintenance and all the support tools around tools that you need to run a full cloud stack. All of those headaches start to creep in.

So initially, it seems very attractive from a pricing perspective, but actually building a service that one can use on a day-to-day basis have a lot of other challenges that are not really apparent right away, right? Well, I'm going to leave this here, but I'm going to bet one thing. A lot of CEOs of ISVs are calling their legal teams right about now to figure out if there's something they can do about all this, because it's just going to replicate through.

And yes, maybe folks, it's not their job and it's not their full-time thing that they want to do, but if they're getting enough joy out of doing it, they'll continue to do it. So we'll be back in a minute. Now, shifting a gear here.

The IMF, I guess, had some conference in Europe. It's the International Monetary Fund, and the leader was basically warning folks, and we've talked about this on this show before, but saying that it's not just a question of AI and replacing jobs, it's a question of what's the economic impact. And so they're pointing out the fact that if jobs go away or people make less money, the tax base is going to dramatically fall, and that next will create a ripple effect throughout the economy.

And I think it was described as a forthcoming tsunami that's going to have this massive impact. And tech folks have been kind of warning IT leaders and business leaders and folks about this for a little while now. But Chris, it feels like now it's part of a mainstream conversation, and the people who actually manage the banks are getting worried about it.

So how do you see this all playing out? Well, I'm glad we're having the conversation. I'd be more worried if we weren't about now, right?

One of my articles was a reference for this, and in that piece I'm talking about The world is always full of this large number of possible futures. And one of the things that can make you as a person, and systems, whether they're literally computers or institutions, organizations, markets, more efficient and effective is to pick some. Pick some future you'd like to live in and see if that's actually narratively possible.

Can you see yourself getting from here to there? " We make movies about it all the time. We have a pretty good idea of those.

We don't want to go there. And we have a pretty good idea that it is possible to get there from here, by any given definition of here, like today or last year or next year. There's also positive outcomes.

Right? And I think that's really the much more efficient and effective use. And given that all of us want to live in a world where there's a Persian restaurant across from the rail station we go through regularly, that we meet people and so forth.

We can engineer it so New York City is all robots and no humans walking around. That's possible. Why the hell would we want to?

Right? So it's good to point out these possible negative futures and design ways to avoid them. But I think the flip side is just as important, figure out where we want to go, see if it's possible to get there from here, and then every day and every month and every quarter and every year, try to keep going down that path.

And we spend a lot of time on this show, in this world now talking about AI and LLMs. And I can tell you at a technical level, your AI's, LLMs work better that way. Bound the possible outcomes you want, generate in that space, you'll tend to come up with solutions down that path.

Humans work much the same way. We think we're going in one direction, every moment we take steps in that direction. And personally and institutionally, we need to recognize that.

I think we need to tax AI agents is what I think. Yeah. Again, once you assume we live in a future that's the "Star Trek" future, not the "Star Wars" future, right?

"Star Wars" is everything sucks forever and always will because life is that way. "Star Trek" is we keep working through things so that yes, we get the automation, we get the AI, and there's still things paid for and a reasonable world to live in. So how does this work?

We tax the AI agents and then that revenue is used to create some sort of basic income that we're giving to people so they can have enough money to go to the restaurant and buy something to keep the whole thing going? I'm a little bit confused and as much as I like the world you might be painting there, Chris, my experience is the opposite. Capitalism is hardly centralized and hardly rational.

So it's a bunch of companies individually driving down for their own efficiencies and devil be damned for the next. But this is where previous experience is useful for future planning, right? We all know these things.

These are the outcomes. These are the mechanisms that lead to the outcomes we want to avoid. And we can and we should ponder that because that's the path we're on.

But we should think about the path we want to go down as well. And maybe it is taxing AI agents. Frankly, I think human attention gets more and more economically valuable as we get more and more of these systems.

I told Will, the excavator operator that I mentioned him on the show, right? Because we were just talking before this. He's got a Caterpillar, or actually a John Deere excavator, big diesel thing with treads and tracks, and he's working on the fiber trench and so forth.

And I think he's not going to get a job in the future. And it's not because he knows how to move the levers to make the excavator work. It's because, as we're talking about, the landscape is real.

The real things are real and human beings paying attention to those is high value. And I don't think that's replaced by AI like we're talking about it today. Yeah.

My concern is that automating junior tasks could remove the experience through which people develop professional judgment, right? That's a key concern, right? If I'm an employer, I should redesign entry-level roles in a completely different way in light of all of this, right?

Around supervised AI use, right? Around code verification and similar problem-solving. Why we can track hiring and progression and wages and that sort of thing, right?

And I think the transition needs practical steps. You just can't do a flip of a switch and do these kinds of things, right? So I think, Chris, in your article, your central argument is really the desirable future needs a plausible path towards a...

That keeps in mind present conditions. It's not like- Hoping things will happen ... a magic shift, right?

Organizations and enterprises should really connect productivity ambitions with credible workforce transitions, right? To me, the next step should be measurable value and preserve people's ability to participate in the economy, not alienate them. We don't want to create a dystopian society where AI becomes this destructive force, right?

That's the way I look at this problem, right? You know the human... Go ahead.

Sorry, Mike. No, go ahead. 0.

The only thing that I'd like to bring up is that when you look at the oil company in Canada, and I think it's Suncor, and just overall, if I were to look at gas oil companies and the automation that's happening there from these big rigs being able to drill and move without humans intervening, I wonder not only does it become safer to drill because they don't ignore the alarms going off, but if they do the calculation of do we ignore this alarm? And if we do, the catastrophic events that could go from there. And I don't know.

There's a lot to be said for having machines that can do our jobs. I think I have a Shark that vacuums. I really don't want to take out vacuuming again.

There are some good things that are coming from this. So I'm unclear, though, how it's still going to be executed, right? I can't force a company to hire X number of people to keep them employed, so theoretically, the government's going to have to give them some sort of allowance that they're going to collect through this tax.

Because the companies are going to be like, you hear people talking about it, some people are out there saying it's only a matter of time before there's a company that generates a billion dollars of revenue with one employee, AKA the CEO. I don't know if that's ever going to be possible, but the point is, it all sounds great from a public policy perspective, but I just don't see the will of the folks involved who are the investors and the people running these companies are going to be like, "This is not my problem. " So they move the kick to the government, and the government is not prepared to deal with any of this as far as I can tell.

I think that's why AI NF is screaming about this thing, but... Ultimately, I hope that businesses with that mindset are out-competed. And we've seen this over and over through commerce, and capitalism is a separate thing.

Commerce has been around literally since the dawn of humans. And untenable structures get replaced by better structures. I think it is better, faster, cheaper to do things differently.

And again, Mike, some of the things we debate, I'm not debating that these powers don't exist. If you do things a certain way, you get the trillion dollars, you have all the control. That's the way things work up to this point.

And maybe it stays that way forever and then we're in the "Star Wars" future for 10,000 years. I don't think so. Now, is it in the next three months, the next three years, there's enough change in that, that commerce and control changes?

I don't know. But I see, to your point, Sid, step by step. I think there's a future out there that's attainable, that doesn't have all these Orwellian downsides.

And every day I find things to do that seem to fit in that future. So we'll see. But I'll throw this out as well.

For all those captains of industry that think that they're going to have companies with one to five employees driving a billion dollars worth of revenue, realize that Next door will be another set of five people who are using AI to compete with you, and the cost of it being in that business is going to drop to near zero for those five people. And so instead of there being, oh, I don't know, maybe 10 competitors that I'm complaining about, there will be thousands of competitors in that space, all of which will be driving pricing down to a level of zero in their zeal to compete with each other because that's what happens when you have an oversupply versus demand. Because as I was chatting with somebody the other day, it's like, wow, it's going to be great.

Hospitals will be more efficient than ever. But as I pointed out, there will now be more patients. So the number of patients and the number of customers remains finite, and if those customers don't have any money, it gets even smaller.

So- Yeah. So the idea, the question is, if you're going to use AI and have five employees, and your competition's going to use AI and have five employees, if both parties are using AI, how do you differentiate from each other to win business? " Because now you're using AI to do everything, and assuming AI is fair and not partial to one party over the other, if it's universally applied.

So I don't know how it plays out. I think that's something worth contemplating, right? I think we can look at Henry Ford.

While he did a lot of things wrong, and I've talked about this before, what he did right was that he knew that if he was going to be able to sell his cars, that the people who were actually working on the assembly line, he had to pay them enough so that they could buy them. I think that as we move forward, that's how CEOs should be thinking in that specific idea. Like, if I'm creating something, then how are people going to be able to buy it?

I don't know. The buzzword of the day that I keep hearing is moat. Everybody's trying to figure out what their moat is going to be to maintain their ability to compete in the age of AI.

The problem is most of the moats that people are describing seem to be around either some sort of secret ingredient, AKA Colonel Sanders' chicken recipe, or it's the workflow. And the workflow is not much of a moat. I'm just saying, because to what Chris described, if I can see your workflow, I can reverse engineer it.

Yeah. " And I said, "We built a team that would die for each other. " So this isn't about AI, and this is not about trust, proving to your actual customers that you don't suck, that you're still going to be there, that you're paying attention.

I pay for that, I'll keep paying for that forever. A license because I have to have it to do a thing, fine. But that vendor doesn't know I exist, doesn't care about me, and is going to change the features next month to screw up my own workflow.

That's the current model. So build trust with your customers and adapt. And I'll add one more little thing before I change the subject.

But the cost of switching from one vendor to another vendor in the age of AI is also dropping to zero. So that moat is very shallow. All right.

I'm going to pull a conversation a little bit into the realm of cybersecurity, but we're talking about deepfakes, and somebody has a report out saying that there's already $40 billion worth of global fraud floating around because of these deepfakes. And there's even one incident where somebody wound up on a meeting with a dozen other folks and wound up sending like $23 million, and it turns out that everybody else in that meeting was a deepfake, and they couldn't realize it or didn't realize it. Now, I admit that some people may not be as bright as others, but Sid, as this continues to progress, it seems like the cost of grading deepfake attacks is dropping to zero, and there's still plenty of gullible people out here.

So is this $40 billion going to be $80 billion, and how big does it get? Yeah. Deepfakes are essentially turning familiar faces and voices into unreliable evidence of identity, right?

We're on this podcast, right? People could take our pictures and faces and impersonate us. So I think that's where we're headed to as a society, and as a new world of AI we live in, right?

So I think it's important that recognizing an executive on a video call is not sufficient. There has to be other checks and balances that one has to institute, especially when it comes to payments and transactions. I think that the article talks about a Hong Kong employee transferring $25 million after a video conferencing involving a deepfake.

That's scary stuff, right? So I think you have to still worry about independent verification through established contacts or alternate channels, dual approval for high-value payments, additional checks for beneficiary exchanges, all of that, right? So, and in the world of autonomous agents, that could just further amplify fraud in terms of if you live in a world of unchecked authority.

So I think strong defense combination detection mechanism with independent verification, and especially when it comes to payment controls, those things that we have today in the world of non-AI still has to persist. And as we hand over the authority to agents to act, these controls have to be really made very tight, and have the right guardrails to be serving as an authorized business entity. And so all that needs to be built into the substrate for it to actually work.

So that's my thought. Chris, I know what you talked about... Go ahead.

I'm sorry. All right, Chris, let's go with you first. One of the things that people are calling for is some sort of trust layer to be added into this equation here, and that that somehow or other is how we're going to all know that it's you talking to me and that the transaction is safe.

But I hear the phrase kicked around, but I don't think anybody knows exactly what is required to go in to build something that would operate on a global scale to provide that level of trust. So, is any of that really achievable, or is this lovely digital society that we built for the last three decades about to fall apart? No, it is totally achievable.

All about the global part. I think, global, the short answer is yes. But, again, my approach has been bottom up.

And I have to start with the old security popcorn point, Kate, right? This is the thing. So let me get this straight, instant response, and your authority structure for spending $23 million and actually sending it is you think the boss probably said so.

Okay, great. Now understand your security control architecture. That's not a good idea.

It wasn't a good idea before deepfakes, right? I got an email from the boss, welcome to phishing. That is not a control.

If that is how you as an organization choose to spend money, then Kate and I will be happy to consult with you and architect better ways to do it. But to the purpose and to the topic, you can do this from the bottom up, easy. Easy.

Everything we talked about, it's easy to generate all sorts of code and structures. If you think about this, I am a small business. I have one owner.

Build up from there. Build into your systems the ability to have the controls and use the authentication that they should if they were autonomous. So how do you identify that a human's a human?

We can get into philosophy and neuroscience if you want. You can't. But pragmatically and practically, something better than I'm pretty sure the authorized person probably said so is a bad idea No, I get that.

You can introduce... I think as a society, we're losing the ability to think in a logical manner, right? I think common sense is paramount here, right?

You can put thresholds in, right? When I'm sending $1,000 to a friend or $500 to a friend on Zelle or Venmo, I don't send the $500. I send one cent first, make sure he gets it, or $1 first, and then I send him $499 as a follow-up, right?

6 million in a video conference, and you're sending the whole damn thing all at once? What sense does that make? This is not an AI issue, this is a common sense issue, right?

Well, maybe it was a bank, and 26 million is the rounding error, so who the hell knows? But Kate, I think you've wrestled with this in the past, so I'm going to toss this to you, but it seems like the cost of launching the attack drops to zero, but the cost of defending against attack keeps increasing. And at some point now we've got an imbalance in the system because the adversaries are spending less than a penny to launch the attack, and every transaction and every interaction has $500 worth of security wrapped around it somewhere when you average out the cost of this stuff.

So is the math going to break? Well, when you pose it like that, Mike, yeah. Talk about leading me down to having to say yes.

Dang. But anyway. But I think- Well, witness, please answer the question.

You've been hanging around Alan too much in this podcast. I think the problem that I have is that, again, we have tools. Let's say you don't have common sense.

We have tools. And why we don't use those security tools, we don't. And there's a lot.

Identity and access management has been around, gosh, I hate to point out Tivoli, guys, but Tivoli with IBM has been around for three decades. Of course, it's named something different now. But the whole idea around identity and access management is real, and you have things that actually combine identity and finances.

But we don't use them, and so if we create another tool, this is the problem at the end of the day. So I don't know, Mike, I guess I'm back to the way you pointed me and just say yes. All right.

Well, Chris, let's go the other way on this, though. So we've talked about this in the past, but are we getting to the point now where we should just assume that everything is false or fraudulent until proven otherwise? No.

And again, that takes you down the philosophy, and you'll never finish, right? Whoever you are, if you're watching a show like this, figure out your real pragmatic world. If you're the business owner, you've already dealt with all of the risks.

Understand those, map them forward, you'll be fine. But this reminds me of a call yesterday with, let's say John. John, if you're watching, I'll point this one to you.

And I was trying to explain, because this person, not an engineer, has great ideas, had AI tools, has generated some really neat AI applications. " Which is good, because that actually works. But literally everything is root, and just trying to explain that this is a demonstration.

You can't actually deploy this anywhere, because if anything breaks, it all breaks at once. So if you are running a company right now, and you've managed to avoid dealing with any of these issues, you're right. It's like Kate were to say, this will add costs.

Now, if you're starting a company, or if you're a little more agile, you're not in a crisis, and you build it in, it's not really adding costs. I think in the end of the day, it saves money. But again, as I said in the first segment, there is no such thing as free.

You actually have to take some time, take some attention, put something in place, and for some organizations, that may be catastrophic. I don't know. Maybe common sense.

I don't know. Kate, what are you saying? Everybody gets a secret safe word for everything they do, so that it's that person, and only you and they know what that secret word is.

In my case, it would be beer, but what do I know? I think we have tried that. And it's breakable, and I think we need to do some biometrics that we already have that goes along with anything dealing with not only finances, but our health, other things like that.

So I think biometrics will become more important at the end of the day for this AI scenario. How about systems as well? Sometimes you literally have to have the buttons so far apart that no one person can press them.

Well, how do you build biometrics into an agent? Well, you find out that it's an agent, right? And if it's an agent, as we talked about the control plane, then we understand that it is an agent.

I think there are some things where the human in the loop is actually very important, and I think when it comes to 23 million, I'd say anything in the million-dollar range, even less. The human should really be in the loop here for real, and have a secure transaction. This is going to become more important as agents, and especially as we're taxing agents.

So here's the part that I scratch my head about and I think will fall apart is that, so there are plenty of folks out there that think it's important that they should be able to be anonymous on the internet, and I understand all the arguments that go with that. But we're also having this conflicting issue with we want transparency so we know who the source is, and we can trust the source, and we need some ability to understand what's happening here, and whether or not this message I got is coming from some server in Africa launched by some Russian propaganda campaign, but I need to know what the provenance is. Chris, are those two things mutually exclusive?

So this takes us all the way to my beginnings of the internet, when everybody had a nickname and a handle and so forth, right? And I've been very clear on this from the beginning that I respect that, but I'm always Chris Betz. I am me.

My idea has always been who I am. Now, what you really want is validation of, right? So some other third party says you are, right?

So just a single token by itself is an identity. And you can be anonymous in the kinds of forums and systems I would create. You can be anonymous to other members of that space because one of them knows who you are, right?

So being anonymous in context should be supportable for real critical systems. I'm not talking about the general conversation in the public square and whatnot. If some other trusted party knows who you are.

That make any sense? Yeah. You know what, Mike?

You're actually going to leave me today actually scratching my head and thinking about this, because there's actually some real things that this article has brought up that we seriously do have to think about. Do we create, like we can create LLCs, that it's an organization, and that you are a part of this organization. So even though people don't know the identity of who you are, they know the organization, and then the organization...

It's a really interesting thing that I think we need to talk about and follow back up in, I don't know, a month. Because this is an important conversation, actually, and how we can solve this. Yeah, I think something Chris said caught my attention.

It almost seems like you need a third party, and what should be the third party, right? Well- It's obviously not a man in the middle. It's probably a role in the middle, right?

Are we talking about that, then? We've had problems- We've had problems with man-in-the-middle models where attacks have been generated by that entity, right? So it's clearly not a man in the middle.

It's probably a role in the middle with a trusted authority, right? And that role could be assigned to different individuals, perhaps. I don't know.

It reminds me of PGP. Chris, do you remember PGP, right? Oh, yeah.

This sort of reminds me of a PGP type of beginning problem way back, I don't know, two decades ago, probably. It's almost where we may need to revisit. Anyway, so thanks, Mike.

Yeah. We've got to end this conversation here because we're running out of time. But hey, like I started the show, first of all, let me thank Kate, Chris, and Sid for sharing their knowledge and their insights.

But as we started the show, hey, the forces of chaos have been unleashed. Have a nice day. We'll see you tomorrow.

See you at the Persian Cafe.