AI Reality Check: Returns, Buyouts and Meeting Trust

Right. Hey everybody, how you doing? Welcome to the latest edition of the Techstrong Gang.

As you can see, we have a full house here. It's just unusual sometimes, but we got just about everybody represented from all parts of the IT universe, including security, development, a bunch of AI folks, and of course, John, who is our resident business tech expert in my mind. But anyway, everybody, welcome.

Thanks for being on the show. I'm going to skip the intros because, well, they're just too long, and when I get to the C block, we'll explain who Fred Cohen is. But, let's just jump in because I have a simple mind, and right now I'm at an event in New York for Cloudera, and they have a new report out that I'm going to get into in a minute.

But let's start with a couple other things. We talked about this briefly yesterday, but Apollo has a new report out that's suggesting that outside of big tech, large companies are not seeing a whole lot of productivity benefits from AI. Cloudera has a report out talking about how, well, 89% of folks are retrenching their AI strategy, largely because the costs are bigger than anybody anticipated.

And also there's all kinds of nasty issues around governance and things that need to be addressed. And then underneath all that is the Anthropic IPO, and there's a number in there that kind of leapt out at me, but it was something like, and John, correct me about my numbers here, but so they're saying that they generated 16 billion in the first six months of this year in revenue, and then they're saying they have 12,000 enterprise clients who are generating about $100,000 annually. 2 billion for the year.

2 billion to align it to six months is about $600,000 for enterprises. Yeah, just really quickly, if you do the math according to Anthropic, they had an overall net loss of $80 billion for that same six-month period, and roughly one fifth of their revenue comes from two customers, but we can talk about that later. Yeah.

And so now you come to it. So Stephen, I'm going to start with you, but it doesn't seem like enterprises are spending a whole lot on AI, and it seems like we are kind of treading water here. And is this going to get any better anytime soon, or it just seems like the math is all wrong.

Yeah, it sure does. I said yesterday on the Gang, I brought up the Apollo study. I'm so glad that you decided to lead off with this and invite me back.

I said yesterday, I shouted into the camera, "Show me the money," and I'm going to do that again. I just did that again. It's so weird because what you just said, I actually disagree with that.

The enterprises that I talk to feel that they are spending a lot on AI. They feel that this is a large expense for them. " The Fortune 493, which is basically the 500 minus The Magnificent Seven, they really have not shown any kind of productivity or revenue per employee gains at all.

They are spending money, but the amount that you need to spend, again, to your point just now in the IPO, the amount that they're spending is actually significant. If you look at that, that's how much they spend on enterprise software. That's how much they spend on Slack or on Office 365 or whatever.

So they are spending on AI. They are not getting a measurable result on revenue. And I think they are starting to wonder if I can quote and phrase, "Show me the money.

" So Robert, I want to ask you this because in my opinion, at least from what I see, most of the spending is driven by developers who are generating a boatload of code, a large percentage of which isn't making it into production environments. So at the end of the day, are we just kind of playing around or what's your sense of what's real and what's not? Well, developers are seeing a huge increase in productivity with AI.

It is definitely improving their lives from a security perspective, and automation. But if we're looking at this productivity gains in a classic sense that we're going to take total revenue divided by number of employees, of course, The Magnificent Seven are going to be higher than anybody else. That's where this big growth is coming from.

So these folks are selling the shovels during the gold rush. And of course, if we're looking at classic productivity gains, they are going to outstrip everybody else. Now, a couple things could explain this.

One is the companies have adopted AI and are seeing improvements in efficiencies and productivity, but they haven't changed the workforce to align with that yet. The other thing is, is that they aren't seeing those gains. So we just don't know yet.

I can speak for our friends in open source and development land, we're quite happy. Senior engineers are ecstatic to not have to deal with chores and drudgery and those sorts of things. Developers were the first to get a handle on hallucinations from your AI agent and to call it out.

And I'm happy to see that we are going back to, or not going back, enforcing things that we learned 10, 20 years ago about CI/CD, unit testing, integration, not giving production keys to engineers that the AI can pick up. So things are great in development land, but I just don't think that that has translated to the financial reports from the 493. Well, let me ask you the obvious question then.

If we're developing more code and developers are happier and things are actually getting shipped, is that software that matters? Because it doesn't seem like it's making any impact on productivity. So we have all this new code and all this new software, but to what end?

It doesn't seem to actually move the numbers. Well, security fixes are not sexy. They're necessary.

And that's what I believe a lot of developers are working on right now. I believe a couple of weeks ago we talked about a big backlog of vulnerabilities. There was a HackerOne study I think we were discussing.

And I think that's where all the work is going right now. There was this huge increase in number of vulnerabilities, and developers are actively working on that. I think the CEO of a bank, Bank of America, right?

Came out and said that they are bringing in a lot of contractors to take care of this. Correct me if I'm wrong on who that was. I think it was BO...

Was it BO... BofA, I think? I can't remember, but- Yeah.

Okay. I got it right. Bank of America.

So, I think that that's what's happening. We're seeing a lot of just not fun updates. We're seeing a lot of security updates.

And for folks like me, I don't know, Fred, if you agree with this, but I love those things, because that means less work for us down the road. And as we all know, a good developer is a lazy developer. Do work early so they have less work later.

All right. Well, since you mentioned Fred, let's just introduce Fred. Fred, how you doing?

Welcome to the show. Thank you. All right.

So in answer to your question, actually I think you're talking about the Fortune 500 here, but the real benefits of AI right now, especially in the code space, come from small companies. Especially early stage companies that used to have to spend a quarter of a million dollars over eight to nine months to develop an application. The CEO is now using vibe coding to develop an equivalent application in a couple of weeks or months, and that has a dramatic effect on the smaller businesses.

Large enterprises, like any big ship, turn slowly. The smaller companies are taking advantage of it, and that's where the real productivity gains are. In terms of security, security tends to be reactive Instead of proactive.

If they would architect things right in the first place, they wouldn't have the vast majority of these problems. So what's happening with these fixes is AI has found lots of vulnerabilities that were already there. And so now they're doing the fail and fix approach, which is a failed approach from the beginning, and they're just doing more of it.

That's not going to solve the problem. Architecture is what solves the problem. The other thing is the AI-generated code introduces a lot of vulnerabilities as well.

It's really bad at making what you would call more secure code. So they're generating lots more vulnerabilities and then having to fix them, which is what the security business has always made money on. All right.

So just to put some context around that, Chris, who is Fred? So if you know the term computer virus, you know Fred Cohen. That came from his PhD thesis where he did the original work on that and many, many, many things since.

And just to expand on this, I find it fascinating that we're talking, and I agree with what Fred was saying, and also we're talking about AI use for creating software, which is a great use case. I do it all the time. But that to me, software is a small part of actual business and economies, and AI tools for doing everything else that is not just engineering software is amazing.

And my productivity in that space is off the charts. And I don't think, on this show every week we talk about this, we keep talking about coding and coding and coding. Business and the economy is way more than software, and these tools used well and in the future, I think that's where the economic and productivity gain comes.

So Stephen, coming back to you, the paradox here seems to be is we are more productive, but it doesn't necessarily translate into revenue that's measured in the way that businesses would track things. So, maybe every employee is having a better experience, but the ROI to the business is a little less hard or harder to actually put a finger on. What do you think?

Yeah, I would like to say that. First off, I'd like to say, holy cow, it's Fred Cohen, ladies and gentlemen. But let me get back to the point.

Yes, I would love to say that there is productivity gains happening here. It's paradox because I am not an AI hater. I use it all the time.

Claude is my co-- I was going to say Claude is my copilot. Microsoft's going to come down on me for that one. But Claude is open right now.

And I'm having it help me with stuff. Claude is my bicycle, and it makes me go faster. So I feel like I'm more productive, but, I'm also a president of a business unit and I don't see increased revenue, not from that.

I see increased output and you look at the software coding, I would question, far be it from me to question Fred, but I would question the contention that small companies are shipping new software based on AI development. I'm certainly seeing vibe coded applications going out there, but I'm not seeing this as a major driver of increased productivity. " AI is just solving the first problem.

" Because that's what AI's doing. Well, we'll come back to that in a couple of minutes. Yeah.

Chicken. Yeah. I'm glad Steven mentioned that and Chris mentioned something about business and the business of software, and I think one thing I find fascinating about this topic and this particular study is the absolute ridiculousness of trying to measure the impact of a tool that is in its infancy.

I mean, these tools in the enterprise are somewhere around six months to a year old, and I would love to have seen a similar or parallel study that was run 5, 10 years ago, 15 years ago. What was the impact on the Fortune 500 revenue and productivity gains when WebEx first came out or Zoom first came out, right? These tools changed the way we work that may not be measurable, have an instantaneous measure on top line numbers for Fortune 500 companies.

But for every 500, now there's 10,000 other industry companies out there, 100,000 or a million, compared to the Fortune 500, all of which see tremendous gains. And I've got massive numbers of stories and parables talking to engineering companies that do stuff with less people because of AI, whether it's on the marketing side where I am now, where we are able to do the equivalent of, we're three times smaller than the staff we used to have to do the exact same stuff. Yeah.

For our teams. So we talked- Other teams. So at the beginning, we were talking about how it's so much easier to track AI spending than to see the incoming revenue.

And I think that was underscored by the Anthropic IPO, which is going to get a lot of attention for a lot of numbers that is checkpoints that are going to change over the next couple of years. But the thing that bears for now, the storyline, is that it's a $2 trillion valuation. They brought in $16 billion of revenue the first six months, but they also spent $9 billion on their infrastructure outlet.

They also issued, in addition to the autonomous AI behavior warnings that they issued in this prospectus, they talked about customer concentration. They have too few customers, who make up most of their revenue, and they're talking about slowing top-line growth. So, I mean, that's going to be the story for now and the fixation for now, and they're going to try to push this IPO.

We've talked about this last week a little bit, or Mike and I talked about it. They're going to try to rush this IPO before the election. They're going to try to move it as fast as possible, but there are a lot of numbers that kind of jump out that underscore what we've been talking about.

All right. So there's two things I want to add before we close it out. One is, to John's point, competition is heating up and you're starting to see OpenAI put out models that are less expensive, and then there's the open-end models that some of which don't come from China, and you can use them happily if you're so inclined or concerned about the security implications.

But the point is that there's other ways to skin this cat now, and so I don't think Anthropic's going to be the only game in town. The second thing, though, to Jack's point, try to be competitive without using AI and see how far you get, because basically you won't be competitive. And that brings me back to my point, and I think I've said this before, but I'm not entirely sure that there is a hardcore ROI for AI.

Is it not just the new table stakes that I need to remain competitive, so therefore we're all in? It's just trying to put a number on it that equates that investment to revenue. I'm with Stephen.

I don't think we're ever going to see that number. Fair? Fair.

So it's kind of like when dotcoms were providing three meals a day for their employees, and that went away. Something you got to do because everybody else is doing it, but then we're just going to make it fade. Yeah.

And not to mention the cost side, because revenue going up is fine, but cost going down is better for profit. Right. And I think- At least it's a topic ...

I'm not sure we've seen the cost going down part of this equation yet. Yeah. How about revenue going down and costs going up?

That sounds great. Yeah. Absolutely.

But listen to what everybody's saying. Fewer people to do the same job. That ultimately results in cost savings.

Well, and I also have one more thing, which is, you look at the SaaS spend per employee for a company. You got $20 for your Microsoft license. You got $10 for your Zoom license.

You have $10 for your Google license. You're probably spending $150, $200 a month on SaaS per employee. If you're a really big spender on Anthropic Claude, you're at $1,000 per employee.

So maybe that $12,000 a year that you're spending on their Claude cost or your AI cost is coming out of their next raise instead, and their salary's going to stay flat, and your spend per employee will You can follow the same curve, but it gets distributed differently across the balance sheet. No, man, I'm going to go work for a startup and make more money working for the startup. To Fred's point, there's going to be more startups.

No, don't imagine you're going to make more money. Yeah. I run Accelerate, the largest accelerator in the San Francisco Bay area of California, and our experience goes along with the industry that 70% of companies don't return the investment after 10 years.

Uh-huh. And by the time you're at break even, you're about 92% of the company. So about 8% of the companies are sufficiently successful to be called winners.

This is like opening up a restaurant. It is. Hey, I've done two of them.

Horrible way to make money. Yeah. It's awful.

All right. I'm going to leave this here because I'm sure we're going to be returning to this topic again in the future. But everybody seems to be watching the same numbers for a change, so that's a good thing.

Speaking of numbers, let's switch to our next block. John, I'm going to ask you to lead this off and explain what's going on here. But Microsoft seems to be paying a large amount of money to tell people or ask people to leave.

And in my experience, that's the thing you do just before you tell them to leave. Right. You give them an option.

You give them the option. Do you want the severance package or do you want to get laid off? Or do you want to take the gamble on you're going to still have a job in a couple of months?

So we talked about addressing this idea that your revenue can't keep up with the spending, and I'm starting to see this pattern. We have Microsoft, which is offering voluntary separation packages to about 60,000 employees. That's 27% of their global workforce.

That is incredibly significant. And this isn't the first time they've done this, this last year or so. It's been the third time.

And they're trying to curb, they say, their labor expenses while they just plow money into AI infrastructure. And I heard the same thing from Barclays and Kroger's, to a certain extent, American Express. And we talked about this off-camera yesterday in a meeting where Mike and I were talking about is this the beginning of a number of companies doing this, and what's the motivation?

Is the motivation because the business, they foresee maybe a slowdown in business, maybe it's a greater economic issue beyond their control. We acknowledge the AI spending is significant, and they do have to cut back. But I'm wondering if they're aggressively doing this, and AI is being used as the convenient scapegoats to accelerate these layoffs above and beyond what they probably initially talked about.

So I think that we're going to see more of these and Microsoft gets more aggressive. Yeah. Let me throw that question to Fred, because we were talking about that in the previous block.

But in my experience, Fred, people, companies go out and they lay off 20,000 people, and then they figure out they need to rehire 5,000 of them, and then they bring them back, and it becomes this weird little cycle, and it smells of financial engineering to move a stock price than a business strategy. Is that going to be any different now? What do you say?

Well, first of all, I don't know how many of those experiences you've been part of, but I'm pretty sure that we're at a step function, right? So AI is not going to go eternally up. It's going to come to the top of the step, and people are trying to figure out how close they are to the top of the step.

So they lay off too many people, they discover it, and they bring some back, but the net effect is it's beneficial. In terms of their financials, that really has to do with timing, right? So they're going to do it sooner or later based on the timing for some report that they have to file so that they can have a better stock price at that edge piece.

So Fred, I'm glad you pointed that out because what's unusual about what Microsoft is doing, normally they announce these types of moves the day they announce their results. They did it before. The other company, Amex, I believe, announced theirs the day of their results.

So they're trying to get ahead. And in a sense, it does affect, the stock does go up. Mm-hmm.

Wall Street loves layoffs. And then they eventually bring people back. So John, to your point, are we on the cusp, or maybe it's already happening, of a big tech massacre?

Are all the big tech companies going to cut these headcounts? It's slow rolling. It's happening.

There's another one. Here's an interesting theory going around is that if California passes Proposition 40, which is the billionaire tax initiative- ... that Mark Zuckerberg, first he's trying to move to Florida.

Now, the talk is that he might take this out on the workforce to compensate for the $5 billion it's going to cost him in personal wealth. There are some crazy theories going around. But I actually do think that a lot of these companies, whatever their ultimate motivation is, are trying to get ahead of the economy.

Because I think that they see business- What impact on personal wealth? Meta said they were going to cover his tax fees if he stayed in California. Yeah.

Well, I'll believe it when I see it. They're spending what? Almost $100 billion on this AI infrastructure push they're doing.

So the math for now is crazy. The spending is up significantly. They have to find a place to cut.

Who else to cut but the rank and file, the bottom line? And then they're going to rehire people eventually. But in the interim, Wall Street will probably like this, but it's going to be chaos and craziness among these tech companies and their employee bases.

But let's be clear about one thing, though, which is that Microsoft and Meta, Facebook, whatever they're called these days, and Anthropic, there's a certain class of companies that are transitioning their business into something that is extremely capital intensive, right? These companies are not talking about millions of dollars in investment, not even billion dollars in investment, but double digit, triple digit billions, right? Tens to hundreds of a billion dollars of investment.

That level of capital and percentage of the company's capital that it is, is hard to come by. The big enterprises, 50,000, 100,000 employee companies, are not dumping $10 billion of capital into this type of play. They're in the millions to tens or hundreds of millions of dollars.

So their need to rebalance the balance sheet to go from operating expense and employee expense to a pile of cash that they can invest in AI infrastructure is completely different. And I think we have to be careful to not take what the big seven or the top 20 or 30 AI infrastructure players are doing and apply those to the rest of the business world. These are extreme examples.

They're extreme. They get all the attention. But the interesting thing is Meta will not hesitate going through the carnage because they did it before with the Metaverse.

Yeah, but they've all said they want to invest anywhere between $50 and $100 billion. Yeah. That's a pile of cash that they just don't have right now, and they have to get it somehow.

I think there's another effect that's underway right now, and that is the upcoming election in the United States and the fight against data centers. To some extent, this election is about data centers, and a lot of people in Republican districts are objecting to data centers, and this is creating a whole set of other issues. I think if you wait a week, you're going to find a lot of different opinions from these companies all the same.

Yes, you will. Exactly. They're going to pivot to whatever the consensus is among the voters, and it's going to be interesting to see how they back away from some of these projects or scale them back or conveniently forget about them once the Democrats- Or move them overseas where they can cut down tens of thousands or hundreds of thousands of trees.

Yeah. Well, I think Finland has an issue with that, but we'll come back to that. Robert, I want to ask you something.

Is there an upside here in that all these people who have been working for these big tech companies are now available to work for startups, and all kinds of interesting things could happen as a result? Absolutely. I would love to talk to every single one of them.

Yes. And especially if you're located in Austin. I'll buy you breakfast, hot cup of coffee.

But I did have a question for John, and speaking of upside here, didn't they carve out, they had a few exceptions there, didn't they, John? Oh, where? For Microsoft?

Yes. Oh, yeah. Yes, they were talking about level 70 employees, and they were talking about just certain types of employees.

But you know what, Robert? I've gone through this before. Yeah.

Anybody who worked in journalism has gone through this before. Mm-hmm. And it's the worst feeling.

You're basically, if you've been at a company for a long period of time and you make a fairly decent salary, you're older, and they're probably concerned about your benefits and your health Eventually find a way to sweeten things to get rid of you. I know that from my own experiences a couple of times, unfortunately, and they will get the cheaper labor, the younger folks who are healthier, and I think as Microsoft- Oh, yeah ... is getting younger.

And that's my point, as far as upside goes, it says, what? A data center operations, commercial sales, marketing, security. So if I was less than a level 70 and in those job functions, I would be talking to my boss's boss about a raise.

Obviously, the work that is important. And so I would encourage the Microsofties, I don't know if they still call themselves that. They do.

But, dating my Gen X self here- Yeah. But, as far as upside, Mike, where there is chaos, there is opportunity. Mm-hmm.

So for the folks that are staying, ask for more. All right. Folks, I'm going to leave that there because there is an upside all the time, and I can remember some of the best jobs I ever got was in a downturn in the economy because, well, there's always opportunity somewhere.

All right, I'm going to shift the gear to something else. Well, it's true, I- Oh, I'm going to say goodbye, you guys. I'm sorry, I've got to go into a webinar.

Oh, sorry. So have a great show, the rest of the show. Great seeing you all.

All right. Bye, Jacob. We're going to have a conversation about chickens.

And I'm not talking about- Okay ... Chicken Little or the sky is falling chickens, but it has to do with, well, how our meetings are changing in the age of AI, and I'm going to let Chris explain what's going on here. But, the whole protocol in the way that we interact with each other is changing for better and for worse.

So Chris, walk us through this column you have over on Security Boulevard. Well, I'm going to try to weave together the comments of the rest of this show, because Robert, you just say, the startups and so forth, we're in hog heaven right now. We love all this stuff.

And, as Quietwire, we build things. But in the meantime, when we started this company a year and a half ago, me and Ashraf and Fred and a couple other people talking about this, and we had to use the tools that were out there. And Gmail, Google, we're still using Google products and so forth, and we're transitioning away from various chunks of them, but the transcription part.

Turns out transcriptions and AI systems, if you use them right, are amazingly helpful. I'll be talking, in the middle of a call, I'll say, "Lumina," and when you read the transcript, it does its thing. And my AI, Lumina, reads the transcript and does all sorts of things.

I don't touch it. But there was a meeting two weeks ago, and it was with a couple of our overseas folks, and I didn't make that call. " And so I went back, and I pulled up the transcript, and it's a two-and-a-half-hour conversation, and since I'm the only English-speaking person, I wasn't on that call.

The call was in Arabic. Right? And if you haven't seen a Gemini transcript, it has the transcript itself on one tab, and it has the AI-generated Gemini Google...

And I've just got to say, to be clear, one of the biggest companies on Earth with all of the AI skills and money to apply to this, they will give you a very confident summary that doesn't make any sense. Right? And it only understands English in this setting.

So you have hundreds of pages with one word, one word, one word. You see this Arabic sound word that's sound in English or sometimes intellectual property might be this sort of Spanglish, Arabic thing. " And it's like, so ship chicken has become a metaphor inside the company for this ridiculous extractions.

Because the summary said, Ashraf and Hashim met together, and they decided to move all the intellectual property of the entire company to New Zealand, and all these other very confident things. And to be clear, I'll get off my soapbox, but we focus on continuity in systems, right? When I say Lumina, I'm talking about an AI system that every time I touch it, knows everything we've done together to this point, knows what to do next.

Gemini, Google's massive AI, has been there for hundreds of Quietwire meetings. Hundreds and hundreds and hundreds. But it has no idea that every single one isn't brand new with a bunch of strangers.

To the point that the summary is that we transferred all our intellectual property to New Zealand. " I just love the whole thing. So ship chicken is all inside the company.

If you say ship chicken, we all know what happened, right? Some big huge system just went completely insane. So my question is, is this unique to you, or are we all seeing variations on a theme here with these meetings, and- Absolutely.

All right. Absolutely. Yeah, look, sometimes you have to help the transcriber.

So one of the things that I do in meetings now is make sure that there's an agenda, make sure that it can read it, and so that gives signposts, breadcrumbs for the AI transcriber. But also at the end, to say, "These are the things we agreed on. " And any human knows how to do that.

You pick up non-verbal cues of agreement, those sorts of things, but AI can't do that. So you have to help it along. But this is again, a symptom of, and I'm not saying that Chris did this, but business leaders are relying on AI and treating it like it's the gospel truth.

And that is the problem. The problem is not transcription not being good enough. It will continue to improve.

The problem is, business leaders are abdicating authority, abdicating responsibility, and being lazy, but not in the good developer lazy way, the bad executive lazy way. And there is a responsibility there when you use a tool. When you're using a knife, cut away from yourself.

And we don't give scouts knives until they go through training. I don't think we should give it to executives until they go through AI transcription until they go through training. All right.

Well, you know- Stephen, I'm going to hand this to you for a minute. I've got to go fix something at the front door, but doctors and lawyers, trust them? Exactly.

Yeah, first off, oh my gosh, Robert, do you really think that you could tell an executive that they need training on a tool before they use it? Give me a break. Go back to- Well, I can, but- ...

my third pencil. They have- I'm a rarity. But at the same time, it is hilarious because, what you say is absolutely right, that ultimately, this is a tool.

It doesn't matter what tool you're using, your name is the one that's signed to it. And it reminds me of the story of the lawyer who generated a bunch of nonsense in his legal briefs, and the judge said, "Look, I don't care what tools you're using. This is not about AI.

" Mm-hmm. And it's the same with doctors. You look at that.

My doctor, I'm sure your doctor probably has a, instead of having a medical assistant in there taking notes, or instead of spending time staring at his computer while he's supposed to be talking to you, he's got his AI voice listening the whole time. Well, first off, you know that you can request access to those transcripts before they- Mm-hmm ... from your doctor.

You know that? You can. Mm-hmm.

And, you might want to check them- Sure ... so that you too don't have a ship chicken moment in your- ... next prostate exam.

I'm not sure- But I understand you can remove ship chickens by an operation. Yeah, the surgical thing. Well, but as funny though, I'm sorry, but when you start talking about ship chickens, I'm thinking about ship cats, which was a real thing, like that cat in the Navy.

And of course, then it made me think of The Penguins of Madagascar, who were basically ship chickens who ended up with the lemurs. So, it's a whole mess. That's what's going on in my brain.

Yeah, the lemurs had to make a show. But I might really want to go back to Robert, what you said, because to be, as I say on this show every week, you have to understand that those of us in this space, I am on many, five, I think, different AI governance working groups with international standards organizations, and we're all trying to figure out inside the groups what we're even talking about. Mm-hmm.

But Robert, what you said. Transcripts are wonderful. " And just keep moving on.

The productivity is amazing And at the end of the call, emails are sent and documents are created. I don't have to do anything. And they're all right because we line our systems up really well, so they know what we're doing.

But the ship chicken is wonderful just because even somebody in our own organization, who's very good, who's an Oxford professor, had to come back to me because Gemini made a summary from a transcript, and it sounded way confident. It was like, apparently- Oh, yeah ... and you have to peel that apart, even at our level.

So the average executive does it- I hate the confidence they have. So I get it. Part of this is, I think, is you talk about the confidence.

We have an unrealistic expectation of the confidence in this AI particular tool, and AI and its capabilities right now should be considered equivalent to a new college grad intern that you just hired. Yes. Absolutely no idea what's going on.

I'm sorry. New college grad interns are much better. But look- It's like being tedious.

It's like I've heard it on the internet ... if you relate it to humans. " So they interpreted the Spanish differently than the court transcriptionist did.

They were both multilingual, but they had two different interpretations of translating the words from Spanish to English, or in your case, Arabic or whatever the language was. And right now, it's the exact same problem we have with Tesla and other cars with full self-driving is there's an expectation that if you have a computer controlling the car, it will never, ever, ever get into an accident. So every time a Tesla gets into an accident, it makes international news, which is absolutely insane given the variables involved in something as complex as either driving a car or transcribing and translating a foreign language is a very hard thing because all languages are context-sensitive, and the computers don't have the same context.

As Robert mentioned, body language is a big part of it. And I've mentioned this on the show before, but I think it's important enough I'll mention this again. There is a famous scene from the television series "The Wire," which is a five-minute-long conversation with exactly one word that is uttered, and it is the F-bomb.

And it is an entire conversation that you can follow and understand based on body language, intonation, facial expressions, but it is one word for five minutes, repeated over and over and over again. And that's something that an AI will take millennia to figure out because we can't give it all the context right now. Well, it's hyper-contextual, and language does this.

In translation and transcription is a perfect example because what I mean in American English, in this context, in this north-south whatnot, even there, down to the individual specific group, we all know this, and we're trying to solve it the top down. So we're going to have a universal translator that speaks English and Arabic. Which English?

Which Arabic? In which context? My soapbox, we can build it from the bottom up because groups actually already know their own idioms, but there is no way to universally map it all ever.

I think you need to lay the blame at the feet of the AI companies and the AI hyperbole, which is not new to AI, right? Yes. So for the last 50 years, people have been saying, "Oh, AI," blah, blah, blah, blah, blah, and it's never been true.

It's a step function, and we're going up to the top of the step. At the top of the step we're going up to the top of is not general purpose, human-level intelligence that can solve all problems. It can't even retain its understanding across two or three turns of a conversation.

The way human- Am I the only one, though, where if I'm sending you an email and your AI responds to me and I can feel it, I can see it, I kind of feel like you didn't care enough about what I was saying to answer me. So- Nope, you're not the only one. I feel the same way.

Yeah. But there's so much sociology to this. And again, for those of us who've been around a couple of years, I came from the Sears catalog printing industry.

The idea that you would not send flyers out to your customers and you would do it electronically was massively radical in the 1980s, maybe '90s, but here we are. And Mike, I know exactly what you mean. We were talking about this a week or two ago.

I create an enormous amount of AI content. I can tell what it is. I think it's really, really good code and writing and so forth.

However, there's a certain human connection we want, and we haven't quite decided how we get it anymore. And we're all so cynical anyways because we live in the world of disinformation and blah, blah, blah, and now this. So I think that the well-learned cynicism of the population plays into this as well.

It's not just the tools. We're tired of it. Yeah, I get...

And that's why all my emails sound like a kindergarten teacher wrote them. Lots of exclamation points and stupid jokes. Yeah.

And nobody came for me. Well, that's what happens in shows like this. You've got people talking to each other, talking over each other, talking about the chickens and something for some reason.

Right. Well, and anybody who knows me really well already knows that I build AI systems that do that. Metaphors, jokes, referential, use words in a way that only I know because if you don't, I am starting to distrust the system.

If you can come back and, say, make jokes about coffee and dragons and mammals and make sure we're talking the same language, then I know the AI system hasn't gone off into the woods. Well, I think- So are you saying... Wait, are you saying that you need to put bad dad jokes and misspellings deliberately into the email to make it sound more like humans?

Well, no. I think you need to use bad spelling and dad jokes with your AIs until they reflect them back to you. And again, different types of AIs.

There's agents, what we call minions, who, "You're going to do this. " Then there's a relational system that I want to be consistent over time because it's my assistant or it's managing a system. With those, yeah, you want to have local, hyper-local metaphors and so forth that that system knows that's not trained into a model somewhere.

That's trustingly. I don't know, Chris, because that cuts me out from the relationship, and that's what I enjoy. That eliminates me from having an interaction with a person, and I don't like that.

No, no, no. We were discussing this this morning in a couple different contexts. I think human attention gets more and more and more valuable.

I think right now we're transitioning, and we expect... Look, I get so much crap email, and now we get crap AI phishing email, and it is also almost convincing, but it's just another curve up this crap scale. But it's- We all want the human connections anyways, and those are always real.

Look, the new spam is much, much better than the old spam. So I enjoy my spam. Yeah.

I keep my spam, I review it, and the transition from bad spelling to good spelling ruined a significant part of spam. And there's a certain joy in watching the new spam come out, but the problem is that it's so- You're scrolling up. Yes.

No, but it's just that the new spam is so ridiculously AI-ish that I can't tolerate it anymore. Well, I don't want to tie up Mike's time to beat us up, but I love this. I love the AI phishing stuff because I will get email from email addresses that really look real after the show that'll say, "Hey, Chris, I really like that show you did with Robert and Mike," and really like someone actually watched, and they did not watch, to be clear.

Right. I can totally tell you if it didn't criticize you. Yeah.

I'm going to end this here, but I will give Steven a quick nod saying he makes a good point here. If you didn't care enough to write it, why should I read it? So there's something to be said about that.

Agreed. I would also just point out that I'm here on Broadway, and yes, there's a new AI show opening up. I guess it's going to be called "Spam a Lot," by the way.

Can you hear it? Bing. Oh.

See, that was a dad joke for the whole thing. Right. And you can rest assured that no AI will come up with it.

There you go. All right. " Please stay tuned for the rest of the remainder of the Techstrong TV lineup that's coming up right behind us.

And with that, stay tuned for a word from our favorite sponsor, because it's us. See you tomorrow.