AI Risk Warnings Mount as Debian Votes on AI Accountability

It's Tuesday, and it's time for Techstrong gang. Welcome, everyone. Wow, we had a great show yesterday.

I almost didn't want yesterday's show to end. If you didn't catch it, check it out on YouTube or on Techstrong TV. It was actually a great conversation.

But I got a feeling today's conversation's going to be as good or better because, as usual, there's no shortage of just good topics, meaty stuff to sink into here. Let me introduce you to our blue ribbon panel of gang members for today. Joining us in the-- Well, we would used to call that a Hawaiian shirt, but I guess now they call it a tropical shirt.

My friend, Alastair Cooke. Alistair, it's good to see you. Always a pleasure to be here with you all.

Thank you. And then, in the study up in Maine, but very close to New Hampshire, our Kate Scarcella. And I guess still up in Canada, Chris Blask.

Yo. And then, of course, joining the three of us is the man in the high tower, Mike Hazard. Mr.

Hazard, how are you? I'm well. Good to be here.

Good. Well, let's dive right into today's Tuesday's work. I don't know, clutching your pearls, the sky is falling.

Oh my god, AI apocalypse is coming again. It gets better, though. You have to kind of stay tuned.

So let me walk through all the different piece parts here. So the G20 summit is on, and it's down in the Carolinas. And in advance of that, the governor of the Bank of England put out a two-page memo saying that from both a cybersecurity perspective and also from a market perspective, because the stocks might be overvalued, that we're on our way to hell in a handbasket with AI.

And his point being that this stuff is too dangerous. And this is an important conversation because, well, at the same time, the Trump administration is circulating a suggestion among the G20, a thing called the Carolina Principles, which calls for limited to no regulations of AI, and just basically has invited some of the leaders of the AI industry to the summit to kind of make this case to the other members. Meanwhile, in China, they were floating out some comments that basically says that we need some clarity around regulation and pointing a finger specifically at Anthropic and saying that we need models that are truly safe and proven.

Now, this is a little on the rich side because if you haven't followed any of this, there's been some analysis that suggests, and I think it was Booz Allen or somebody like that that said that they seem to have detected that when a Chinese model, at least some of them, detects that it is a US company or a US agency that is using them, that the code that it generates and the suggestions it generates are altered from what you might see from somewhere else. So there seems to be some clever behavior going on in this. So Chris, this all feels like geopolitics AI style at scale in a way that is just mind-boggling.

But what's your take on what's going on here? Well, yeah, the world is dead. Long live the world, right?

And I love the heading for this stuff, you know, AI apocalypse, ah. Look, it's a complex world. These are extremely complex systems and issues we're talking about, and they're playing out.

And if you look at where we've been coming from to this point right here, each of the issues you just mentioned, Mike, they're each artifacts happening in a space over time. And since Kate and I get to join you every Tuesday, and we talk about this week to week to week to week. What you said at the end, we need models.

No, stop right there. We need large language models, the individual little chunks of code trained by, they're at the center or involved in these large systems we call AI to be trustworthy and so forth. I would challenge that as a premise to begin with, right?

A really good functional, particularly large language model, frontier sort of thing, has to be capable of doing all sorts of things you wouldn't want it to do. Otherwise, you're not generating into the space to get the answers and solutions you want, right? Now the very concept of you can have secure models, I don't think is a thing.

It's systemic. And then, thinking of the geopolitical realm, yes, large nation-states, notably those two, the US and China, have their interests, and those are playing out in public as well at the same time. It's not overwhelming at all, really.

This is normal strategy stuff, right? It's more than a 15-minute segment we can really break down here. We can talk about parts of it, but each of those issues inside the geopolitical interest, right?

The current US administration has the position that is sort of understood. So the unlimited open model or AI system threat is being pushed. That shouldn't surprise anybody.

What are the implications thereof? I'll just end my little rant with there are more than two countries on Earth, right? And I am an American with a Canadian wife and family and live my life back and forth between the two.

And the answer for the 40 million people in Canada or the other 7 billion people on Earth who don't live in the US and China can't be answered by either country. We need systems where the evidentiary trail is there, so whomever you are can make repeatable and resilient and reliant and reliable choices. Systems that tend to play out against that will probably lose in the end.

Alan, what is your assessment here of the Red Scare that we constantly pull with the Chinese models? And is that for real in your mind, or is that a bunch of US companies with a particular axe to grind? Well, the Red Scare, what's funny is the Red Scare wasn't originally Chinese.

It was Russian, Soviet. Oh. They just transferred it over.

" Where the Trump administration decides to make a G20 meeting down in the Carolinas without telling the rest of the team, the rest of the G20. They invite the Russian, I almost said Soviet, the Russian finance minister to take place in the G20 meeting when they haven't been invited to take place in it since the Ukrainian war. " Then on top of that, you got the AI billionaires who paid their ticket, and they're entitled to come in, too, and say they're-- Because they do own a piece of this, it seems, right?

They've bought their influence. And they're allowed to come in here and say whatever the hell they want to say. " Right?

Because they were quick to badmouth Anthropic and regulate Anthropic and tell OpenAI they got to delay their thing. So we don't want any regulation unless we do it. That's a pretty arrogant world, a pretty arrogant view from where I sit, right?

Then you got the Chinese, and you got to give these people credit, too, because they're no less arrogant. They're saying, "We don't know if we want to do an AI deal with the US. " Especially those Anthropic people who always blame us for distilling our models off of theirs, right?

While Anthropic is being sued for ripping off videos and everything else and rare books and the rest of it. When I said a couple of weeks ago that this is a soap opera, this is truly not even a daytime soap opera. This rises to Dallas level, nighttime soap opera, right?

Well, the only thing missing here is sex. No. This is crazy.

Well- That's all I got to say. How do you know it's missing? Just curious.

Well, it may not be, right? That'll be the next thing. Who in the administration has an AI sexual partner or something?

Who knows? There you go. Yeah.

Well, and again, I know this is a USA show, and at least four to five of us are US citizens and so on and so forth. But this is the internet. We're communicating to the world.

And even in the American businesses I'm involved with my entire career, there's lots of non-Americans involved, right? So we all understand- And you know, Alan, we have on this show, my position on the current administration is well-known. Don't approve.

Zero of five stars. Wouldn't recommend. But it is what it is.

We've got to stop being surprised, particularly security people. Look, adversaries act in ways. Right.

" People do things. Our job is to figure out, given that people are acting certain ways. China, to be clear, I don't approve of the Chinese government either.

I don't like the single party rule thing, everything else. I'm not a Chinese citizen, so I'm not involved in that conversation, but it's not a terrible country. It's got its own pros and cons.

The rest of the world, including the US, does all sorts of business with it and trades people and humans and ideas and products. The actual world is not caught up inside the narrative of any country. We don't sit here and talk about what Putin is doing with his inner man where can-- I mean, he does stupid, insane things.

We know that. We factor that in. I will posit that my country is sort of operating that mode, too.

It's got this sort of chaotic bag of cat stuff. You don't have to actually know which cat is which. Right?

It'll do the things that it does. How do we navigate around that instead of just talking about the cat fight? All right.

Well, Alistair, you're down in New Zealand, so what is your end of the world saying about all this stuff? Well, small countries, particularly remote small countries, like to have some sort of point of differentiation, and there's definitely a push in this country for careful use of AI, careful building of AI data centers here. And the idea that there should be unfettered freedom to create anything that a corporation wants to create is, I think, a uniquely American idea.

The idea that the society is at its best if the corporations do whatever corporations do. If you start looking at other places, you start looking in Europe, you see far stronger protections in place. This is why a lot of US companies find it very expensive to operate in Europe, because regulatory regimes are much stronger.

And this is across AI as well as kind of almost every aspect of life. Rule of law is stronger through Europe. Rule of corporation is a little stronger in the US.

So these things reflect the natural culture of the places, the players that are involved. AI, of course, there was a strong axis that is US, China, and that leads to, as Chris Blask saying, inevitable conflicts. Everybody's doing what's in their best interest.

And there's conflicts between that because what's believed to be best is different in different cultures. This is fundamentally what we're seeing. But also to Chris's point, politicians, crazy cats.

And often not as well informed as you'd like. So maybe a little more freedom is good, a little less regulation of what these large AI companies are doing. I wanted to touch on one of the other aspects in here that you opened with, Mike, which was the concerns around financial chaos and the end of the financial world.

We've talked repeatedly about the possibility that there's a bubble, and how would we know for sure that these tulips are actually worth what we're paying for them? That does continue to give me concerns that we'll suddenly find in a year's time there's a new way of doing AI that doesn't need all of these things we've committed trillions of dollars to building. That might be somewhat problematic for the people who have to repay those trillions of dollars.

This is true. Kate- They might repay them ... I'd love to get your opinion here, but is there any chance that maybe we could just create some non-aligned think tank for AI, that the world can just share the research and maybe take all this political crap and throw the politicians out of the conversation altogether?

What do you say? As the Pollyanna of the group, I'm going to say yes. Seriously, I do believe that because first, I'm tired of the Chicken Little, "The sky is falling, the sky is falling," and the sky doesn't fall.

And so we sort of get used to these warning labels over and over and over again that haven't, Chris being in the green room, we talked about it, right? I mean, we're all still here. " I think we see that in the next article even.

And so I'm hopeful. I'm hopeful about AI, where it can actually go, and I think people are crazy. I think everybody's going to the Carolinas and all this.

And look, I understand that there are people who are responsible for the systems going to the Carolinas. However, there are so many people who don't know what they're talking about, making up decisions and coming up with these crazy decisions. " As a typical person in Midwest Ohio here, the yeah/no.

Yeah, no. That's not happening. So let's just, Chris, talk to me.

The one now who's on Lake America. Talk to me. There goes that thing.

To answer your question, Mike, I'll take Pollyanna up a level. If we think there's going to be a third party, a neutral thing on Earth where we all share kumbaya, no. But there will be many, right?

The whole idea that we're going to build our national critical personal family business systems on single points of failure, like everything coming from a certain country or some nonprofit third party being the panacea, no. There will be lots and lots of sharing among the lots and lots of... And US and China represents a lot of the businesses on Earth.

Those businesses, they aren't America, they aren't China. They're people and organizations that exist there and interact with the outside world, too. We're all connected, and there's lots and lots of layers of sharing that- So, Chris, I've got to disagree though ...

are already happening around that. Go ahead. I've got to disagree a little bit on that.

The concept of a company or corporation in China is a very different concept than a US corp or a Western corp. They are controlled by the state, and I don't think... I mean, go ask some of the billionaires in China who were brought to heel for getting a little too outside of this Communist Party's vision.

Well, as a third party, though I'll debate the US and China are entirely different on that spectrum. It's a matter of scale and time. But I agree.

I don't have a big- And I'm not saying the US model is panacea or nirvana either. I think we have the opposite here. In the US, the corporations, well, first of all, we have more government ownership of corporations right now than we've ever had.

Thank you for that, Alan. Good grief. Yeah.

I mean, that's the new MO, is you want to do a deal here? Give me a piece of the action. Okay, you want to bring up Lake America?

Google has to put Lake America on there because they need royal decree to exist because they're an American company. So everyone in Canada deleted Google Maps. But Apple Maps hasn't done it yet.

They will soon. Thank you, Apple. Thank you.

Again, two to adult and zero for Apple. Well, but who knows? MapQuest has not done it.

Yeah. But the US government-controlled companies like Apple and Google and Meta and so forth, these are part of the US government, I will just say it this way, exactly the same as major Chinese corporations. As an external party, I trust them.

I would recommend- Never more so than with... No, Chris ... anyone in America.

Absolutely. I agree. But the only thing- Go ahead ...

the one thing that I'll say, and we saw this with cybersecurity, are cyber mercenaries, right? I mean, the idea that we have these boundaries like we do with countries. The internet all connecting us together has really changed the model.

We may not visually see this as of yet, but it's out there. And I say that, to go back to Mike's point of my little Pollyanna, there's many of us who are really trying to change the way systems are delivered, especially to homes. And I don't think it's forever.

I don't believe in a centralized model. I believe in a distributed model, and I think we see that with cyber mercenaries, and we see this with people actually standing up and trying to change. So that's my take.

Pollyanna. Okay, Miss Pollyanna. On that note, though, we're overtime.

Mike, we got to jump to our next one. We do. Let's lay it out.

Which is another form of, well, democracy. And the Debian Linux community got together and decided to have a little debate amongst themselves about, well, what is the proper requirements for using AI to create code for the project? And they evaluate everything from outright banning to forcing people to at least disclose their entire usage.

And they kind of settled on something that feels like the middle. " And well, okay, this kind of seems maybe ridiculously rational, and is this the way other projects should go, and is this how it's all going to turn out, and it's all good? I think open source leads the way.

So yeah, and our good friend, Mitch, whom I haven't seen for a while, so if you do see him, say hello. But he points out, I think the issue that we've had and that we solve with cybersecurity, and that is the people problem and capacity. And we've had that, and we're still having that, this whole human in the loop.

So he points out that AI can produce code at machine speed, but human review still happens at human speed. And so we can write accountability into a policy, but policy does not create review capacity. And again, humans, humans, humans.

We have seen this as an issue since the beginning. And so I love their thought process. I love the way they have gone about this.

Brilliant. But at the end of the day, we still have the issue with humans. It is an issue.

But what do we do? So... I don't know.

I read somewhere, I think, that maybe 40% of the patches being created with AI coding tools were breaking systems because, well, the AI coding tool didn't have any context for the production environment. And we wound up with this infinite loop of, well, we created a patch that broke something faster, and then hopefully maybe we went back and fixed it with a human. But Allistair, is this cycle kind of broken?

What's going on? The cycle of make changes as fast as possible with as little review as possible, and surprise, things break when you do this, is kind of predictable, particularly given the challenges with getting an AI system to have enough context awareness to be able to work out the consequence of changes. So often you'll scope the context awareness to just the very specific problem you're trying to solve because it costs a huge amount of money to have the sort of context awareness that you have from a senior developer.

So here it is. The AI is good at doing things that a junior developer might be okay at doing, but tends to struggle with the things that a more senior, more experienced developer with years of history, therefore huge context, can cope with. Is this just a transitory thing, though?

Is the huge expansion in the size of models, the size of context that's being stored as we're learning more about how to operate inference, which is what this is, the inference part of AI at scale, is this a problem that'll go away? Probably. Yeah.

It'll probably- I think this is the kind of thing where digital twins will give you that context that you need, right? Recreating you. You can recreate the failure, but you can't recreate the knowledge of how to resolve the failure with a digital twin.

Yeah. I hate to tell you- Yeah, you can. I think it goes to what you said.

They're rushing this out, maybe. And quite frankly, as someone who's been in the vulnerability management business for 25 years, this is not new. This is why it used to take us 30, 60, 90 days to roll out patches.

Because before we would roll out a patch, we would test it in the lab in all different configurations and make sure it didn't break something worse than the thing you were fixing. We've gotten away from that. I think a couple of things.

First, I personally hate digital twins. The reason why I hate digital twins is because I feel like you're reproducing all these vulnerabilities, and you're just giving more access to more vulnerabilities to be exposed. That's more coding and everything else.

So I really believe in software de-bloating and reachability. And what I do think, to Allistair's point is, I believe that AI will actually help us to become very lean. And what you said in the first, we're building something prior to what we actually know of how well it's actually going to work and how well it's actually going to be orchestrated.

What AI has done really, really well, it has shown us our weaknesses and our vulnerabilities in code and in our products and everywhere. And that's really, really good. It's good because it should wake us up and say, "You know what?

We do have to go on diets. " And that's not bad. Well, and this takes us into where we left the last segment, right?

So, there's more than two answers and single sources, and there's nobody at the policy level, particularly in politics and so forth, who even, Kate, you touched on this before, even understand the question. We're debating it now among the developing world folks building this stuff, and we don't have the right term for it. But in the terms we're using, right?

And I think, Allistair, you threw this example. We have the vulnerabilities and bugs. We're using the same words to mean a whole bunch of different things, right?

AI coding tools can be just something as simple as this little LLM and a little harness around it with this little context window. That leads to exactly the kind of things that we're talking about now. And to Allistair and Kate, digital twin, yeah, also the wrong metaphor.

I agree. It is a way of doing it, but that's not what we're talking about here. And with all my biases known, but Ashraf El-Hag, my co-founder, you guys all know him, he's obsessed right now with the term situational artifacts.

And it's another way to express this. When you say situational, it means situational. That exact specific situation.

So I'm a coding agent or a coding agent system inside some vendor, exactly with all of the past, present, future, the context around that. Not the context window. Context is the kind of context a human needs.

And the context window inside a coding tool with an LLM is this little string. It's not enough. We really have to throw everything that we actually think that we know almost out the window, and because we have to really rethink this.

We see a brokenness, and we continue to want to go on our brokenness. Stop. Just stop.

And let's rethink this. Let's almost run in a deprecated mode, in a depression mode, in a safe mode as we rebuild. It's going to really help us in the long run.

I think, though, there is something to be said, like there's a group of people who want to shame folks for using AI, and there's a bunch of folks who are using AI that don't want to admit that they're using AI. And maybe we should just come to some clarity around the point that just say, "Hey, whether you used AI or not is besides the point. " And you should at the very least say, "If you want to tell people you used AI, you don't need to put an asterisk around it.

" And that's a great idea, right? It's part of your canonical, like Chris, these canons, right? It's part of our canons going forward, and there's nothing wrong with that.

I was speaking to somebody at a party this weekend, and as I was listening to him, I was thinking to myself as he was talking, and I actually... He was a C++ programmer, blah, blah, blah. And as I was listening, he was talking about how he hates AI and blah, blah, blah.

But I was listening to what he was wanting to teach students. " Like the train leaves the station, and I can write the Python and... " So I keep thinking, we're not doing this right.

So yeah, let's talk about AI. Let's talk about how it's better at coding than us. It's more efficient, I think, or can be.

And there's nothing wrong with that. I think the same is true for writing, the same is true for marketing, the same is true for a lot. Yeah.

Here's the deal. This is a transitionary phase we're in right now, where they're looking to put the scarlet letter on something that's AI. And Mike, you say, well, because this way people will know there might be hallucinations.

Why? Because when people write stuff, we assume everything is correct there, and they're not hallucinating? Come on.

We're rapidly getting to the point where the quality you're getting from AI is going to exceed the quality you get from humans- Yeah ... for all but the most highest task. But I subscribe to something Reid Hoffman, one of the co-founders of LinkedIn said about it.

Because we wrote about this a couple of weeks ago. I wrote a whole series on it. The Scarlet Letter was one of the articles.

My friend Brad Feld chimed in, and Reid Hoffman wrote, not because of what I wrote, let me be upfront. But Reid Hoffman said, "Hey, if you're willing to put your name on it, if you're willing to stand behind what it says- Yeah ... if you're willing to be public about it, it doesn't make a difference whether you wrote it word for word by yourself, with AI, AI wrote it, makes no difference.

" And I think that's what we're going to get to sooner than later. Because everything's going to have some AI in it. Like this whole thing with the Claude watermarking within the words.

All work's going to have some AI in it. What are we going to start making second-class work here because that one- Yeah. That's a passing phase- Doesn't make any sense ...

for people like Kate, like this person you were at the party with. Yeah. For the troglodytes who are going to wear...

I wrote this on a LinkedIn post, a comment. " Yeah. " Because they wear their we don't believe in AI badge- Yeah ...

proudly. Other people may think it's a dunce cap. Well, I have always, and now that I'm deploying for a company again, we take attribution absolutely fundamentally serious.

Not just take it seriously, that is a primitive. When I publish things, and Alistair said, I publish it, that's me. And maybe somebody else co-wrote it for me, maybe I used AI, maybe I used pigeons.

I wrote it, I'm taking responsibility for it. And publishing an article is the same as signing off on the pressure valve reading or anything. There's some person is taking responsibility for that.

" Look, who is the person responsible, or the person, the organization, the entity? Nail that down from the bottom up. And a lot of these issues, at the human level, people are going to go through the phase of trying to...

And a lot of AI generations, AI slop is well-named. But that doesn't mean there's not wonderfully coherent AI content out there. Pictures, code, art, writing.

Yeah. Is a person behind it is the real question. Absolutely.

But Mike, there's a person behind the clock here, it's me. We need to hop to our third segment today. All right.

Well, it just keeps getting fun out there. So Meta is talking about at least testing the idea that they're going to put robots in their data centers, and the robots are going to install servers and pull cables and all the physical stuff that goes with managing a data center. Alistair, when you read this, what sprung to mind?

Is this the future? Is there not going to be IT people running around data centers? It's all going to be robots?

Yes. getting my hair blown away. Sorry, it's all gone now.

Yeah, I think more automation in data centers isn't a bad thing. And the interesting thing is, what kind of problems are you solving with what kind of tools in here? Because as soon as the article was talking about, we're going to have a robot that moves through the data center and presses the power button.

It's like, what the hell? Because the data centers I've worked in have remote management cards that can do this remotely. But then it comes to, well, this is because Apple discontinued all of their server models, and now we're putting desktops in data centers in order to run workloads on Apple hardware, and there's no remote management on those.

But I immediately thought, well, a little servo and the push of a button is a lot easier. But it's a single-point solution. A human body is a very versatile tool for doing a variety of different tasks.

And that's the sort of direction that I'm reading this from Meta, is having a robot that can move through the data center and undertake a task that it didn't know needed to be undertaken before it started. And that's a very human ability to go and problem solve on the fly and mentally work. And using a robot to do that is an interesting change.

I liked some of the commentary in there about robots are going to be slower to do this stuff than humans because their perception of the space they're in is much slower moving. Dual cameras and working out where to move a robotic hand is much slower than our movement. So there's an interesting kind of thought about to achieve the same amount of work outcome, you're going to need more robots, and at what point does it actually become more sensible to just retain that human?

There's a variety of tasks that they're saying still are a long way away, and specifically called out doing all of the complex connectivity that's required to deploy the latest generation of Nvidia servers. So they're still seeing there's a long way to go. But a lot of the time, the actual manufacturing, the filling of those racks full of equipment is going to be done not in a data center, but in a fulfillment center, in a big warehouse, which is probably going to be done by specialized robots for that task.

But yeah, process automation, robotic process automation of a physical variety is a reality. We've seen that in every other industry. Why would we expect it not to happen in data centers?

And is it- So to your point about that, therefore, is the future going to be there's some IT person that's walking around with their robot posse, and the robot posse's doing little tasks as assigned by the human? Or how do you think this will play out? I don't know.

Chris, what do you think? All the above. And Alistair, so this is like a lot of things.

We are going this direction. We will get there. Now maybe it turns out to be way harder to make metal knee joints in robots and it takes another 30 years, but we will get there.

And so for those of us who are advising or opining or doing the work, it's worth thinking about in my space, how long will it take to get there? And yeah, so auditors. I think I've offended a lot of auditors in the working group last week because I've always thought auditing goes away as a profession.

Now to be clear, there are millions of auditors out there making a good living, working 40 hours a week. But I would just say that if your idea, if you're 32 years old and you're going to work another 33 years and your idea is, as an auditor, my job is to show up and look through six months of old files, that's not what auditing is in the future. And working in a data center, physically crawling, and for anybody who hasn't done that, this is the fun stuff.

The freezing fingers and the trying to get the cable in, it's awesome. But this is industrial infrastructure. It's getting automated.

But this segment, of course, raises the Terminator, kill a robot thing. But as we discuss week to week here, we have those on actual battlefields now. So the questions are the same.

Do I have the reasonable boundary systems so that these robots can do what they do and not kill too many of us along the way? Which just takes us back to all the other stuff that Kate and I tend to agree on these days. Yeah.

One of the things I loved is talking about when I first thought about is running cable. You know how many floors I've had to lift and run the stupid cable to-- And I thought to myself, "Gosh, I don't know. " Yes, it can.

Yeah? I mean, lifting up the floors and- They're pretty dexterous. You know what?

There's a great report on the state of humanoid robots that we put up here on Techstrong. I think it's on Techstrong IT. It's a really good report I wrote just a week or two ago, and you should go check that out.

From a dexterity point of view, they are approaching, and even surpassing in some ways, human dexterity. If you've seen them do kung fu fighting and- Sure ... especially the Chinese robots.

But this is running cable. No, they have cameras on their fingers. So when they go under the floor into that cable caddy, they're actually seeing what's there, and they're very subtle.

Mm. So there's no doubt in my mind they could do the job. But the issue right now we're running into it is doing it autonomously versus following sort of a script, right?

Where you may need-- My initial thing reading this was, oh, you mean those last five data center jobs are going to go away? Because that's all we've been hearing about these AI data centers- Yeah ... is there's no people there anyway.

Let's not kid ourselves. Yeah. Alastair, the days of you walking around with your tool belt kit with the Phillips screwdriver here and the flathead here, those are long gone.

Right? You're buying whole racks now. They're put in there.

There's no more data center dudes, right? And so there's five jobs left. Yeah, and the robots will do those, too.

When I buy a phone- Well, that's talking about the massive scale data centers. I completely agree with you on the scale that Meta and the other hyperscalers run it. They stack in rows of racks.

They don't stack in a single rack. They stack in rows at a time. And those come off a production site.

They're cabled into place. Kate, the cables go over here now. Particularly once you start doing water cooling, you can't have space under.

So there's no tiles to lift. There's just wires put on a rack, but they get put in once when the thing's commissioned. Yeah.

And that's it. And then three years later to replace them. Yeah.

I'll just tell you. What is making me sort of laugh right now is Fidium optic cables just came into our neighborhood. No robots, by the way, putting them in.

But as I'm trying to discuss with the person about where I want it throughout my house, I'm ending up having that cable, put fiber throughout my house. And so I'm just sort of laughing that I guess I need a robot, because I'm really getting into some small places. And that robot doesn't need to look like a human.

There probably have been robotic systems involved in getting that cable to you. It's just they don't look like a human. It may well be a bit of a creeper- Right.

I mean, obviously going to be like a- ... that goes through ducting ... spider.

Yeah. Yeah. All right.

So the data center jockeys, they're gone. This is the horse and buggy whip. Is that what we're saying?

Rest in peace. RIP. But that future's unevenly distributed.

There are still thousands of tiny data centers of the sorts where you're still skinning your knuckles as you're putting rack nuts in. Although hopefully there's now rack nut-less enclosures for your servers. But the idea that every data center is a mega-scale data center is not correct.

Meta's data center's are massive, but your local hardware store, your local big box retailer, there's data centers in all of these places. They're just- Yeah ... smaller in size.

And those ones, there isn't the economics of running a huge amount of automated systems in there because you're not making changes often enough to make that valuable. So there still is a place for data center jockeys. They're just not in those massive data centers that are tying up trillions of dollars of investment and for questionable future possible returns.

They're in the data centers where people are running production, where people are continuing to run the applications they've been running for the last 20 years on the same sorts of infrastructure they've had for the last 20 years. AI isn't replacing all of those, and robots aren't replacing all of the people who are doing this variety of different tasks. Do you guys have scars from these data center accidents that you're talking about?

I do. My blood is throughout the data centers. If they were doing like a NCIS or something, you'd find my blood throughout the data centers across the United States.

Well, I think-- Excuse me. And I think this is, I'm just thinking, if I was going to forecast, you got to understand this, the logistic scale of the planet. This is a 15-year arc minimum.

To what you're saying, Alistair, you seriously say we're going to replace, even just in the developer's world, build and deploy that many robotic systems in even just the major, and then the smaller, and then every data center on Earth. That's not in the next 10 years. That doesn't happen that fast.

So we're looking way over the horizon for most of our purposes. I will tell you, go look at BMW's assembly plants in Carolina. Oh, yeah.

They're all robots. But again, that took 30 years to get there. Yeah.

This is a whole new thing. It's all finger robots. But guys, we're about out of time.

So it may take that long, but the show won't. We're done for today. Alistair, thank you.

I know it's an ungodly hour out there, so we appreciate you getting up and coming in. Kate and Chris, always a pleasure to see my Tuesday cyber people on here. Mike, thank you again, as always.

Thank you for watching. We do this every Monday to Friday at noon Eastern Time. tv, Techstrong TV YouTube channel, our OTT channel, which is on Roku and Amazon and Apple, as well as iOS and Android.

Anywhere you've got a screen, we've got the gang coming at you. We'll be back tomorrow with a lot more interesting topics and great gang members. But until then, this is Alan Shimel.

Have a great day, everyone.