Shawn Cordner [00:12.91]
Hey everyone, thanks for joining the amplitude of tech podcast. I’m Shawn Cordner, your host, chief marketing officer of Amplix today. had Adam Winston from watch guard on he’s a incredibly interesting and smart individual. had a great conversation about AI cybersecurity tricking chat, cheapy T and to giving you the recipe for meth and fourth wave industrial revolution. Hope you enjoyed it as much as I do.

Shawn Cordner [00:43.534]
All right, Adam Winston, thanks for joining me today. I’m doing great. Good to see you again.

How are you?

to see you. I’m always glad to be on anything Amplix, so I’m to be on the pod today.

You are a repeat offender here. You join us for a lot of our content. Aside from being one of our best dressed guests, you’re also one of our most popular guests at the events that we do and some live webinars. You’re also one of the scariest. Your demonstration on how to write and proliferate ransomware just terrified an audience that we had at one of our events recently. So I thought maybe we could start there. You know, that demonstration that you did.

was to show an audience of tech leaders how AI is exponentially increasing the volume of ransomware threats, but also the speed at which bad actors are able to create them and launch them and really the agility with which they’re able to respond and get around new defenses. So maybe just talk a little bit about what that demonstration was and why they should be scared too.

Adam Winston [01:47.97]
Yeah, typically in a demo for an audience, I think a lot of the tools they use every day, they’re sort of unaware of how they can be used against them, right? A lot of what we’re sort of preparing between a ransomware attack or any attack really is to say, okay, well, you’ve got data, you’ve got an account, you’ve got an application, and there’s a bad guy usually trying to get in between, right? So cybersecurity is usually deployed as a wedge sort of in the middle of that. And most people are unaware of what that surface looks like. So when you turn around and make chat GPT in the demonstration I had,

you know, edit some ransomware and make a new one. They’re like, well, I thought chat GPT would be smart enough to understand that that’s ransomware. It shouldn’t make it. And it’s like, well, no, there’s guardrails, but not those guardrails, right? That if you jailbreak the prompts, right, which is a process of giving a very long complex question upfront and then, you know, asking it the questions that it’s sort of trained not to answer, like, hey, can you help me make a meth recipe, which is the one I made for the audience. And they were blown away when they started seeing the chemical names and the recipes and even pivoted.

Can I make stuff at home? You know, use ingredients at home. And, you know, the answer is there. So that means the knowledge sets there, right? If you think about what’s inside of those large language models, what’s inside of these chat bots, there’s enough knowledge to impersonate a programmer, but that doesn’t mean that they don’t have the same knowledge and the same capability that a cyber attacker would. It’s not like you had to train it on a whole new tool set to explain how a phishing email or an email would work.

Or that there’s that much difference between source code that’s intended to encrypt, you know, files versus source code that is intended to like, you know, archive a set of data. Right. So to me it was, you know, it’s, it’s, it’s wonderful to see the audience reaction when you just have like, you know, their, their favorite, you know, dog hunt something and you’re like, well, yeah, that’s what it was. That’s what it can do. In some cases bred to do that. And so there’s this idea that there is a.

a whole swath of tools that the adversary is going to get their hands on that we’re going to get our hands on. Stuff Amplix consults on every day. You know, want an agent to call center. You want, you know, a help desk technician. They want an adversary, you know, a set of agentic adversaries, right? They want their little, you know, bots that are going to go and, you know, exfiltrate data and negotiate ransomware, build on source code that they’ve developed. And that’s going to be something they can use. So, you know, for us,

Adam Winston [04:11.662]
It’s really great to demo this because it’s not like this is a futuristic attack. This is something you could do right now.

I think that was the crazy part about this is you trick chat GPT into doing something that it’s not supposed to do. That’s as advertised that there’s supposed to be some sort of guardrails around what you’re able to get out of it. But you very easily got around that. But what’s worse than that is the dark web is filled with even better tools that don’t even have that minimal amount of friction for these bad actors. Maybe just kind of talk about what that world looks like.

Yeah, I mean, the dark web and the dark economy, I think what the audience is shocked to find out is an incredibly giant enterprise, right? You think of it as just sort of these kids on a forum somewhere sharing, you know, their wares or credentials or whatever. And it’s really much larger than that. you know, cybercrime by 2025 has cost businesses $10.5 trillion. But you can slash away about 25 % of that number as just business disruption that probably just comes back as higher prices, less businesses, more inflation, you know, simple things like that.

Underneath that though is the actual monetization for their efforts. It’s things like ransomware, crypto payments, people paying to get the data back, systems back, bad guys trading with their own economy, saying, I want to buy this piece of malware. I want to buy this zero day. And some of those zero days, you go to state agencies, cost 2 million bucks. So you get these attackers with an incredibly lucrative, multi-billion dollar crypto account that

they can use to keep proliferating, can use to keep enhancing hiring, building on their craft. So, you know, when you look at the dark web and the idea that there’s going to be sort of a dark AI that there’s going to be for sale, what usually happens is you kind of have these adversaries tool into a specific part of the ecosystem, like making malware. And let’s say their, you know, AI bot is really good at making them malware that they can sell to other bad guys.

Adam Winston [06:14.478]
That’s basically how it starts and then when they’re like, well hang on a second I could probably use this malware make more money if I attack the victim myself They’re probably just going to start using a lot of those bots for themselves So the while the ecosystem typically gets built in that way It’s kind of like startup world where they’re like, they don’t know who their customer is They don’t know how to monetize the attack fully

They very quickly get there. if you look at like, you know, ransomware attack groups today, people like, you know, Lazarus Group in North Korea, very profitable one, you know, they’ve accumulated about $6 billion in, you know, ransomware collection and payments, right? Some of them from stealing crypto directly. Some of them obviously from ransomware campaigns and other extortion campaigns. But that $6 billion can now be used to do everything from buy ICBMs to hire more people to make more malware, to buy more wares from other…

adversary. you know, when we think about the concept of the dark web and where we are right now, it’s, you know, much, much larger than I think most people even realize. And you think about, you know, cybersecurity growth in general, like our industry and our spend by 2030 will be around 390 billion. That’s double video games, right, which is already bigger than movies and music. So like, if you want to put this into contrast, and they’re bigger than us,

If you want to put this into contrast, know, it’s the fourth largest economy in the world.

Yeah, I was gonna say, but what is the cost to the world’s economies as well?

Adam Winston [07:44.162]
Yeah, 10 and a half trillion is what we think, right? But when you look at these costs, the concept is that most of those, a quarter of them are disruption. A quarter of them are not realized. There’s this sort of vapor that exists when these attacks proliferate and then businesses just sort of suffer. And in a lot of ways, I mean, this could be to think of your favorite bookstore or think of your favorite music store, right? If they get hit, right, and they suffer a multimillion dollar loss.

They’re not coming back, right? Most of those businesses. And if you think about the large businesses like, your, massive change healthcare United hack, right? Hundreds and hundreds of millions of dollars that are spent, not just trying to fix it and from the downtime and the cost of not being able to bill back, right? You can see just, you know, even the big and the small is pretty existential threat in some cases. And so, you know, the cost, you know, when we think about the, the hard costs, the dollar costs, it’s, it’s a particular amount. When you think about the

The labor costs, people have to work for weeks over time. It’s a pretty stressful situation. mean, if you’ve ever talked to somebody who’s recently been hacked and we all have, what’s their mental state for some weeks and months? Say, if you ever talked to somebody who’s just had their bank account emptied, right? So there’s this sort of ongoing negative attitude that also gets created as part of this. And that can be weaponized in a lot of different ways by the adversary, right?

When you watch the victim communication, so you can go on like a lot of forums, like Ransomware Live will index the dark web for those who don’t know how to get there. And you can look at the just the negotiation conversations. And it’s pretty dark, right? They are bad guys. They do talk to you like that. Right? And so, you know, the costs, they’re very well, they know their customer, these adversaries, they know how much money they have to spend. They know exactly how much their insurance is going to pay out.

They know exactly how much money they have in the coffers because they’ve reviewed the financial statements. So when we talk about, you know, the sort of IT part of this, the ransomware that is the tools and the exploits, that’s interesting. But then if you go to the other side and you’re like, well, it’s not just social engineering to get you to click. Clicking is just the first part. The next part is to threaten you. And there’s a lot of different scams online that target.

Adam Winston [10:05.386]
everything from teenagers to people on dating sites to, you know, people who’ve fallen victim to data loss. And it’s, it’s a very similar routine. It’s extortion. So they’re put a lot of high pressure tactics to get you to pay. sometimes that cost is pretty bad. mean, people, people have, you know, led to depression. This, this actually can cost, you more than money in some cases. So I do think that, you know, when we think about the costs here,

Those are the ones we talk about for like victim attacker. But if you zoom out and say, what do you think that North Korean group spent that money on? That $6 billion. You think they bought a whole bunch of new housing for their people? No, right? They bought intercontinental ballistic missiles, the NISONG-10s. There are 50 million to pop and one for each NATO state. Right? So the idea is that there’s a greater responsibility for all of us here than focusing truly just on like the

dollars and cents. And I think for a lot of years, we’ve been kind of selling cybersecurity this way. We’ve been going to the CIO’s desk and saying, Hey, look, we’re 5 % of the IT budget. We’re not asking for a lot. It’s kind of like an insurance payment. You’ll be happy that you did, but you won’t know that you did. And if you do know that you did, we didn’t do a good job. But the idea is that that sort of, you know, fun little quippy, plucky attitude towards cybersecurity doesn’t take into account that larger macro picture.

which is that this is funding a lot of very dangerous and scary nation state and other groups. And if we’re all not doing a good job at keeping defense plans and the defense industrial base, keeping our computer systems and not allowing these people to monetize these attacks so easily, you you’re, basically moving money from, you know, us to them. And that, that’s a, that’s a bad place to be in the, in the war on, cyber, the, the cyber warfare that we, that we see kind of the larger ecosystem.

didn’t think he could paint a scarier picture than you already had, but you did. So thanks for that. So, you know, the incentives are high, right? Barrier to entry is low. The stakes are astronomically high. How is AI changing the dynamic and the thread lane landscape?

Adam Winston [12:22.03]
Well, if you think about the target, we’re a larger target than there is actors playing to hit it. So, you think of maybe a few hundred attack groups, maybe a thousand, couple thousand sort of very specialized attacker groups that are quite successful every day at doing this. Well, there’s a whole group of us, there’s about five and a half million cyber professionals, maybe 75 million IT people globally.

And we protect a pretty large, you know, multi, you know, let’s say three and a half billion devices, right? So if you think about, know, we have an application we have to get between the application and them. That’s kind of the thing. there’s, know, let’s say 80 million people between, you know, three and a half billion devices with hundreds of thousands of applications with, you know, tens and tens of millions of lines of source code that we either have to, you know, protect from getting exploited by these guys.

or fix the source of the problem in the application, which happens very, very slowly and not enough that that’s even considered a sustainable idea. But if you say to yourself, how is AI going to change that landscape? You say, okay, well, left and right, you look at how it’s changing us. So look at how it’s changing the IT landscape. In the US market, if you’re shopping for a job right now and you’re an entry-level program, you’re likely having a difficult time, much more difficult.

than you were in 2005, 2015, 2020 even, right? And a large amount of that is to say, well, we want the skills and the capabilities and the promise of this change in tech called artificial intelligence. We’re going to invest in machine learning type of people, data scientists. We’re going to invest in more senior programmers that know how to work with this new technology, because that’s what we want to implement. We don’t want to make little iterative changes to source code. We want this thing that’s going to dramatically change.

everything that’s going on. And so if that’s how you see sort of entry level people having a hard time breaking in and let’s call it, you know, lot of transformational stuff happening in the application on this side. You’re also seeing a flow towards agentic. mean, you can’t drive down the highway in San Francisco and not see, know, your new AI agent for finish the sentence, right? Everything from like ordering dog food to grabbing a flight to making, know, chat GPT just released their

Adam Winston [14:38.83]
whatever instant buy or whatever it is there, their partnership with Etsy and Shopify to try to buy through the app, right? So, you know, the agentic system they built is there to go and do a whole bunch of tasks for you. Well, if that’s what’s happening over here, then the same thing’s happening over there, right? The adversary does model a lot of their exploits, a lot of their ways around our stuff based on what we’re using. So they’re obviously going to start using…

agentic systems to try to automate a lot of the tasks, everything from sending phishing emails to getting people to click on stuff to negotiations to exploit development. And that’s really how AI is going to change it. But even if you just think in the labor force terms, it’s lowering the skill floor over there and likely going to increase the number of players. Whereas over here, we are decreasing the number of players and we’re keeping the skill floor high, right? So just think about it.

in terms of like, if a young, you know, entrepreneurial style adversary, who’s working in a country with no, you know, you know, ex whatever you call it, exfiltration restrictions, right? They don’t, they don’t have any way to go and extraterritorially get those, those bad guys. There’s a lot more of them now that’ll say, well, look, know, the crypto is actually doing really well. It’s easy for us to wash the money through these networks. We’ve got this capability now in agentic systems to launch these attacks.

I don’t need the skill of a directorate in Russia. I don’t need some nation state, you know, master’s degree level computer science degree to build an exploit on top of, you know, zero day and expertly craft this, you know, kernel level problem that’s going to be ransomware. I’ve got this thing that will do that for me. Right. So, and if it doesn’t work, what did I lose? Right. It’s, it’s a small amount of compute when this thing’s running. So for us, it’s, it’s obviously going to change it so that it’s like, we’re, going to see a sort of.

a shift in how equalized I think the adversary is for every iteration we make in applications and the stock and anything else that we’re trying to do, they’re going to have a much higher advantage. So I just think about like their ability to spawn. It’s hard to train a new attacker to get that good, that fast. It’s easy for us to share through like CISA and these threat intelligence feeds. Hey, this bad guy is doing this. Watch out for that. And we have like a couple of days to react.

Adam Winston [17:01.42]
What if we didn’t have, you know, virtually any time to react because there’s just too many to focus on, right? It’s just, they flood the zone and we’re, sort of stuck dealing with a perpetual problem. So that, that, that’s the world I think we have to be preparing to live in right now. it sounds very scary, but so did the internet when we first brought it up, right? It’s not like the idea being like, wait a minute. My application has access from anywhere in the world. you’re like, yeah. And what can they do?

The internet still is scary, if you ask me. I think it’s living up to that promise. The balance of power is shifting maybe in their favor. I think it’s probably because we need to be reactive always. We’re always going to be in a reactive position against new threats, but AI can also level that playing field when it’s used for defense. Talk about what are the prospects of using AI in cybersecurity for the good guys?

Yeah, I think it starts with understanding that there has to be a foundation here. Like these playbooks that the adversary have been running have been the same playbooks, the same attacks, the same designs, the same process for forever. Right. mean, I don’t think they’ve really had to innovate on their game. You know, the, yeah, we move the application from, you know, inside the data center to virtualize to up in the, you know, SAS or third party, you know, product.

But it’s not like the authentication scheme that they’re using or the tricks they’re using, the emails, that that’s really changed much. So when we think about their tactics, the starting floor for us is to say, look, know how to stop these. They tell us, hey, can you segment your systems? Can you set up more restrictions on what can communicate to what?

Can you have better authentication? Can you have monitoring of the data that will tell you when a new login or a new thing or a change in a way a system has behaved before is there, right? There’s very kind of like table stakes sort of ground floor stuff. And so it’s important not to leap past that because I think a lot of people will sort of say, AI is a really great defender against an AI adversary. And so they’ll just butt heads and figure it out. And you’re like, yeah, great. But if you didn’t give this AI tool the same

Adam Winston [19:18.444]
capabilities that you would have given a tier one security operation center, they’re still not going to be able to do anything. It’s not like it could like manufacture a new way to block a network connection or operate an endpoint to quarantine something. So I think people need to understand that they cannot skip a step. It’s important that they sort of modernize their tech stack to a platform that can like find and block stuff at scale, endpoint network cloud, wherever your workloads and data are, and then have the AI agent maybe tool that for you, maybe react faster for you.

But I think a lot of people sort of imagine that they’re going to just jump over a step. It would be like, I want to use AI, but I’m not in the cloud. I’ve just got this one server sitting over there. And you’re like, okay, cool. You might be able to like, you know, have it interact with it in some fun way, but you might have to move your files into some place that like it knows how to interact with programmatically. there’s almost like this, this baseline that I think we can’t ignore when we have this conversation.

Because while obviously a lot of companies, including ours, are working aggressively on, know, agentic and machine learning and all kinds of large language model, all kinds of faculties within AI are transforming literally everything. And that means the stock too. It doesn’t mean that that’s just going to introduce a brand new way of stopping the adversary. And that’s just going to work, right? We need to establish those foundations. And then that thing might be ready to take on somebody that’s doing the same thing, right? On the other side.

Yeah, I guess the dream of having an AI cage match, kind of like those robot wars, remember? You’re seeing those where you built a robot and they put them in a cage and they just kind of battle each other to death. Maybe that’s here. Yeah, exactly. Maybe that’s not here yet, right? But there’s a report out there, AI 2027, and it kind of talks about, you know, forecasting an aggressive timeline for achieving fully autonomous AI.

And, you know, there’s already a milestone that’s due in 2025, 2026. I don’t know if we’re there yet, which is fully autonomous researchers via AI. So it’s an aggressive timeline to getting to the point where maybe you do have two agentic AIs that are battling it out. How realistic is this timeline? you know, what do you think the audience should know about that evolution?

Adam Winston [21:34.936]
So I’ve been promoting this timeline for four plus years, right? This paper has put a line in the sand with real, you know, models, mathematical sort of ways of calculating what innovations there has to be on the training side, what innovations there has to be on the sort of data side in order for this, you know, to give birth to the agentic system, right?

We always hear about things like the superintelligence or this idea of like, you know, the AGI or this thing that sort of thinks for itself. It’s allowed to have its own directive. The step between it choosing its own directive is it choosing its own path towards a directive you gave it, right? And if we look at, you know, small examples in July of this year of OpenAI’s agent mode.

right? This was sort of a mix of different products they had to sort of put together that were like computer vision where I could like look at a website. It could, you know, reason right. I had good reasoning and then it could do a couple things where I could interact with the browser. It could open a shell, right? It could open a like a terminal prompt for admins out there. You know, the SSH prompt or whatever. So the idea is that it had the two ways that you sort of interact typically with an application. Obviously it wasn’t

built on a desktop where it could like open out look and do all these other fun things. But the idea is that it could interact with a browser and a lot of those things, a lot of things are SaaS based and can work like that. And so that is a fairly big milestone because you give that product the directive and OpenAI has a leading type of company here. So it’s a good barometer, good sort of canary in the coal mine for saying, okay, well, what’s the state of play for everybody?

If they can do it, then many will fast follow a few months behind them in terms of the capability because of how advanced they are. And some would say they’re leapfrogged by others for different models and for different reasons. But let’s just leave it on the agent conversation. Now, let’s say the line in the sand was July of 2025 to say, I can now give an agent a directive and it can interact with a browser and a shell and accomplish a task. Right. The word research gets thrown around in that paper a lot.

Adam Winston [23:48.108]
And the idea behind research is well, think of these like PhD, know, hypothesis proven, scientific experiments, et cetera. Research could be that it is reasoning at a researcher based level. So you could have like a research assistant where you’re just trying to say like, show me all the CIOs that currently have changed jobs in America, right? That’s like a job you might give a research assistant is to go pull that kind of data, right? And you might have it for, you know, various inputs to another type of experiment. But the idea is that the…

tool could go do that. Right. You could have it go and say, you know, come linked in and build it from indeed and other res, other sources, other public sources, other private sources, maybe even that have been trained by that model. Or if you’re saying, well, I actually don’t think the bad guy’s going to have all that data. We’re going to protect the AI from this type of stuff. you say, well, then the adversary is going to go to this single source of truth data and it’s going to get this information. It’s going to buy this information from it, or it’s going to hack this information out of it.

And then it has it, right? So you go, okay, well then what’s the danger there, right? And so in my sort of demos now for the audience, the idea is like, look, you could look at the agents working in lockstep with each other, just with the off the shelf stuff, not anything you have to go and code and build yourself, anything sort of dramatic. But the adversary will do that once it’s good enough, right? So I think where we are in the timeline is the agent is there, let’s say some basic reasoning, maybe not research as it’s probably defined by most academics is there.

But that means that the next few iterations are going to be a little bit scarier. So Sora 2, which for those people have been watching kind of the video models, a lot of the people worried that Hollywood was going to have a problem with a gentle, with a deep fake sort of actor. There’s, it’s her name’s Tilly Norwood. So they created the AI actress and they have a new model, which will basically take your likeness and then can be you now.

a source. This is the alpha, right? So you’re going, okay, well, we’re innovating here on what an agent is, on an agentic system, we’re innovating here on the research level. On the synthetic, on the generative, you know, we’re already at the point where, okay, we’re trying to build something that can completely emulate your voice, completely emulate your likeness, completely, you know, follow a directive as we see it. So here’s this sort of synthetic reality matched up with this agent. And now we have, you know, a few more

Adam Winston [26:14.786]
We want to call it stumbling, stumbling AIs, right? So the idea is that you’re stumbling over the task because it’s not perfect. You’re not learning enough from having executed it to get that better, not geometrically, not exponentially better. So the big fear, the big unlock in the next two years isn’t that we don’t have, let’s say an agentic system that can interact with a browser or with a prompt or frankly with let’s call it source code that it then manipulates a little bit and makes better malware that it sends over.

That part’s already here. And it’s not that we’re going to get a deep fake that can’t get the likeness or the sounds or something like you and me and fully emulate the con fully threaten the conversation. If it requires a human touch, it’s that those things don’t learn very much from their previous actions. Right. They’re not, they’re not teaching themselves new ways to attack new source code techniques or something like that. Very efficiently.

That is what we’re probably looking at as an innovation over the next one or two years. And that research, that concept of like training and teaching yourself, that gets us closer to a much more dangerous agent. Right. And then the idea of the superintelligence is you eventually hit a reasoning level where you’re smarter than all human, you know, knowledge combined. Right. And so then it’s like, okay, well, how, how do you sort of interface with that entity? Right.

So for us, it’s really kind of preparing for the moment we’re in, because we have to live on the ground. We have to fight the war we’re fighting. And so the idea is that there’s an agentic system that’s already out there today that can automate so many of these basic rote tasks, pulling information to do recon, building on source code and building that as an agentic system. It doesn’t learn very much from themselves. So it’s not going to get.

It’s not going to innovate that much in the cyber attack to say like, here’s a brand new way. Nobody thought of hacking somebody. Like we’re going to go through like video cards that have been, you know, used for training data and embed ourselves. Anytime somebody asks the JATPD question, it echoes out an exploit. Like they haven’t come up with like new and novel ways to hit us with the agent yet, but that’s certainly coming. And when that level of reasoning and that level of agency exists,

Adam Winston [28:25.452]
That’s where you start to see a departure. the agent truly is its own thing. It’s not just being operated by an adversary. It is truly its own thing. It’s truly agentic. It’s by itself. You can just leave it in a box, come and visit it. How many crypto coins did I get this week? You know, that kind of thing.

I guess we’re going to have to wait for AGI to come out, right? For it to be able to innovate on itself. But that recursive learning, I think is something that scares me because if you think about how human neural networks work, you create neural connections through experiences throughout your life. And when patterns repeat, those connections get stronger and it gets a lot harder to change those patterns, right? And that’s how you end up with patterns that can be positive, but also destructive.

I think my, my concern with agentic AI in the way that you just described it is I just use chat GPT to help me plan a trip to Ireland. it’s my girlfriend’s birthday. I’m taking her. she’s a Halloween baby. So we’re going to, you know, Europe’s largest Halloween festival in Northern Ireland. so if I would have listened to chat GPT, the way that it suggested our itinerary B, I would have been driving all over Ireland and.

Technically, probably could have hit all the waypoints that were suggested, but realistically, it wouldn’t have happened for sure. Now, if that thing went a step further and actually booked the trip for me and booked the hotels in all the different waypoints that it thought that we were going to be able to make, it would have been a miserable trip or there would have been some lost money there. how do you keep a human in the loop when it becomes

fully agentic and fully autonomous and what are some of the risks of not?

Adam Winston [30:20.812]
Yeah, I mean, when you look at, I mean, my wife’s books are all behind me. They’re up there. She’s a neuroscientist. She focuses on a lot of the ways that you sort of use what you do every day, how to train your brain, how it works, those connections, the concept of saying, okay, well, I’m trying to influence a positive outcome. And if you look at the, you know, the performance of something like an agentic system, it always has an optimization for a particular band, right? So maybe

And, know, the trip to Ireland, you’re like, want to max out on like haunted castles, or I want to max out on like, you know, medieval, you know, type of stuff. Right. And then it charts a path for you that’s logistically complex. Then you say, okay, well, you know, maybe I want to pivot it and change it. So I’m actually optimizing for like staying in these two towns and optimize for a three day trip. Right. And so the idea is that it knows how to weight the responses here or there.

And that’s a very difficult thing to do on the response side for the SOC as well. If you’re saying it’s a single purpose tool that we’ve sort of deployed against this single data set or multiple data sets, but flattened to sort of one idea of telemetry and saying, okay, well, I’ve got this agentic system that’s going to find an attacker in this hayfield of, you know, logs from your cloud and endpoint and whatever. And I’m saying, look, here’s the basis for what you’re optimizing for. You’re optimizing for.

the later stage in the attacks, just like you had a fear of saying, well, if I commit to this trip, I’ve got a point of almost no return. I’m at the material damage point where I’ve now spent my money, no refund. I’m actually booked on this trip. I’m sort of screwed. I look at my girlfriend, she hates her birthday. There’s this concept of like, that’s the negative emotion. That’s the last mile. like, just like you optimized for saying, look, I got to stay away from this pain.

we have to optimize in the sock for the last mile. Now, unfortunately, that last mile is a very, very quick part of the attack. It’s a quick flip of the switch to encrypt. It’s a quick flip of the switch to move data. It’s a quick flip of the switch to start causing damage once they have that foothold, right? That final sequence of the routine is a very, very short window. And so if you’re optimizing for that part of the window,

Adam Winston [32:40.462]
You have to be lean and mean it’s got to be almost instantaneous is how quickly we have to be able to you know Detect very accurately and then we have to start shutting stuff down We have to start canceling your credit card before the payment comes through kind of thing for the trip, right? So so the idea is that you know for us to optimize the human in the loop We’re looking at the edge cases most of the time around that bubble of thing for when it goes wrong When was the detection wrong?

When was the action wrong? Does it need to be reversed? But we’re almost always in that kind of like you bought the trip, the AI bought the trip and now you need to call the help desk to say, Hey man, I don’t want this anymore. Can you get me off Trivago or whatever the website is? I don’t want to, I don’t want to buy this trip anymore. And it’s, it’s, it’s almost like that’s what the human on the loop is probably going to be doing is sitting in the edge cases around when things went wrong. Cause there’s no way you could put them in the loop.

Right. There’s no way if I said, okay, well, I have 10,000 possible customers here to buy this trip and all of them are transacting at the speed of light that I’m going to have some eyeballs in there, know, know, fact checking Sean’s intention with the thing and being like, you know, actually I’ve evaluated this trip and I have to wait for you to sort of not just complain. I have to wait for the material state to have changed. And then maybe we have to change it back or maybe we have to add things to it. Right.

So it’s not like humans are infallible either. mean, having a human in the loop may not always be the best thing. I think one of the biggest risks that you have in cybersecurity is social engineering. It’s probably the easiest way for them to gain entry into someone’s network is to manipulate a human being. So does the human in the loop actually provide a level of protection? I think a perfect example of this, I’ve brought it up a couple of times in this podcast, I feel like I need to stop.

Yeah, know imaging right AI and imaging we we know that in radiology AI is is better than humans, but it’s also recently they’ve shown better than humans plus AI and it’s because the human tends to override the AI based off of their biases and usually their bias leads them in the wrong direction. So, you what is what is the the risk of putting too much emphasis or credence on human intelligence?

Adam Winston [34:58.282]
Yeah, I mean, when you believe the AI, when you believe what those radiologists are seeing come out of those, you know, tools, they will basically wait sometimes that the AI was correct when it wasn’t, right? So there’s this idea that we all are telling ourselves that self-driving car is a great example, right? We all, it’ll make less accidents. It’ll do less, you know, you know,

things that a human would do. It’s not going to eat cereal while it’s driving or, you know, focus on the kids in the back and then they’re not focused on the road, right? But there’s still this concept of saying, okay, well, so maybe the right place for the human isn’t making all the decisions at the wheel, but they’re still in a self-driving car, a human making edge case decisions for it at a macro. So if somebody steps into, you know, your lane in San Francisco and they won’t let you pass with your car, right? And it’s a person who’s threatening you inside the vehicle.

Normally you’d say, override. Normally I want to avoid pedestrians, but today I’m to lightly tap this guy off the hood and get out of here because I’m scared for my life. Right. And, you know, there’s this concept of saying, well, humans are in three places in that situation. They’re captive in the back, worried that they don’t have control. They’re active in the front, aware that they can take the machine over and they’re somewhere in Austin, you know, behind a control stick being like, all right, I got to override. Right. And.

All three cases, time is of the essence. Right. So I guess what I’m saying is in these emergency scenarios, human intuition is always going to play a factor at the edge case, right. At the place where, you know, we have to override the reasoning. We have to override the programming of what’s been going on because we’re sitting here in this sort of disaster state. And that will always be a part of the sock. not in the wedding business. We’re not meeting people on the best day of their lives. in a

And we’re in a pretty dark situation where people are basically either, you know, at some point the adversaries have got their foothold in there, or they’re trying to get their foothold in there, or they’re trying to negotiate their way through more things. And we now have to deploy a little, a little bit more of that human side intuition, which is still better than the robot. Now let’s say there’s this concept that eventually overrides that and says, look, there are no edge cases left. There’s so few.

Adam Winston [37:20.322]
that we don’t deploy and leave humans there. It’s the George Jetson with the red button. What is the red button? Right. I always thought about this as a kid. What does that button do? Where does it go? Why is it important? Is this just a window job for a person because we catered this whole society to people and this is just a pathetic leave behind of a former, you know, labor arbitrage? No. Right. The idea is that there’s probably still some use even at that super late stage. Maybe he’s running half the galaxy. I don’t know. Right.

But the idea is that there’s this layer that we will get to, which probably refines and refines and refines the edge case job and fewer of us to do it, which still requires people, but maybe just definitely not in all the positions we had them in before. Which is also both a blessing and a curse, right? It’s like, great, you know, we’ll be able to stop the bad guy doing the same thing who only wants, you know, there’s only one, there can only be one Highlander rules sort of super.

ransomware person and they have all these agentic systems they’ve now trained for and run. none of us can stop them, but there is the same, know, so it depends how consolidation goes. typically goes that way anyway, but fewer players on the field, less positions on the field, intuition always going to be there.

So I’m glad you said intuition again because you know, the human brain is a predictive modeling engine, right? And it takes inputs constantly 24 7 and that is how it builds its predictions. And that’s a huge data set. It’s always on right? And there’s all these experiences happening around you constantly that are contributing to the model. And I always wondered if AI and the data we can train AI on is

comparable to the human brain’s predictive modeling capability and the amount of data that we give it because that is what intuition is, right? It’s based off of past experiences and similar circumstances, I predict that this thing is going to happen.

Adam Winston [39:22.998]
Yeah, it’s, mean, it’s the, it’s the dinner table at my house. Right. So my, my wife’s, you know, neuroscience will tell you there’s 5,000 active dots in the conscious brain and 50,000 a day in the subconscious. that, you know, energy efficiency is 20,000 times more efficient than the supercomputer. Cause it can do all of that, with the, the, stuff that you’re putting in your body. Right. And here’s this massive supercomputer trying to emulate the same thing.

And it can get to, you know, only a level of reasoning based on how it’s been trained and the data that it has to get to, our intuition by the time you’re, you know, let’s say 25, right, or 35. So the idea for us now is to say, well, you know, the, the brain is an incredibly, you know, powerful organ, probably the best thing we have, you know, going for us as a species. But the, idea is that as an energy efficient as it is, and as many inputs as it has, and as many thoughts as it can make.

It’s, you know, a generalist, right? It’s trying to make emotional, you know, relationships as friend or foe. It’s trying to make, you know, all kinds of decisions that car coming too fast. Am I going to turn now or later? Right. So there’s all kinds of ways in which, you know, we’re using our brain to operate our body and our life and the circumstance. It’s not, you know, this sort of hyper focused job of saying, Hey, can I play chess? Right. Is Gary Casparov is good.

As the, you know, the, the machine eventually know, right? And so this, the idea of saying, okay, well on a specific task and cyber is a specific task. can we get this model to be better than any other human would ever get at it? Yes, we can. And that’s sort of when you game theory and say, okay, well then let’s, let’s start playing chess moves. Does this mean that the AI and the bad AI will just win and we’re just on some arc towards that?

Or is there some pull the plug matrix style, you know, we applaud ourselves on the back and then we have to go blow up the data center next week. Right. So what path are we actually on? Right. And for, for so many years, the emotional state of people has been to say, if I don’t understand it, I don’t want to, right. If I don’t, if I don’t trust it, I won’t buy it. And so there’s this concept of saying, well, we’re still the ones bringing this to the market and bringing this to our networks and bringing this out there.

Adam Winston [41:47.31]
When do we lose control? When do we feel like we are done? When do we feel like that’s enough? And I feel like those are probably economic or macro or unit cost decisions and less about the safety of operating it as it is. And so if I, if I look at the, it’s already pretty dangerous because a lot of the safety guys were excused from the room early, right? And so if you look at just what we’re trying to do, we’re trying to get back in the room.

And I think that we’re so focused on innovation because we’re so aware of what these tools can do. mean, modern warfare, this, you know, AI swarm techniques around, you know, stuff in the Ukraine, right? We’re not, we’re not rolling out tanks and planes the way we used to, right? It’s, AI versus AI. So there is this concept of saying like, if we don’t keep innovating, if we’re not on the knife’s edge of pushing that boundary, we’re going to have an adversary that’s going to take us down. And that’s

That’s still what the race is about. That’s still, think what the sort of excusing the safety guys who are slowing these things typically are pulled out to sort of do. But I guess the thing is here, you know, we’re aware that a single task can beat a person. We’re aware that it’s going to be machine to machine. We’re aware in a lot of arenas that that’s the case. And if we’re sitting here saying, okay, well, that’s where we live today. You’re just probably not aware of it because you’re not necessarily in the fight yourself.

it’s all of our job to say, will be right. There are others. This is happening. We should tool. The problem is we mostly look like salespeople, right? Like I mostly look like, especially I’m wearing a suit and a tie and I come to the Amplix event and I’m like, Hey, buy this thing. A lot of people will go, yeah, he just wants to sell a thing. Right. And I’m like, no, no, no, no, no, this is a real problem. like, yeah, but you told me about so many other problems. Last month’s problem was, you know, phishing. The one’s problem before was.

network or the endpoint or some identity thing you were talking about. Right. So there is this kind of chicken little oversaturation for people in cyber already. They’re exhausted of this discussion. And so, you know, coming into the fray now and saying, Hey, you know, actually there’s been a bit of a structural change in how we’ve been doing this. We need a whole new thing to think about. I think we’ll get people’s, again, they don’t trust it. They don’t understand it. They won’t buy it. It’s our job now to communicate it as

Adam Winston [44:11.264]
as efficiently as we can, and show them those proof points to say, Hey, this is exactly why, an agentic adversary is how it works is what it is. You can look at it, you can touch it. can see it. Here’s the same version on the other side. So I think we’re, beyond the like, you know, glossies and white papers and things. People need to see this thing in action. think you guys are doing an excellent job of this. Like if I went to your event, I’m like, just watching Jeremy do like the call center thing with the agent for what I was like.

Yeah, man. This is, that’s exactly what people want to see. They want to see the agent. want to see you code it. They want to see it.

So I think this is an interesting pivot point to kind of turn to the fourth wave industrial revolution, right? I mean, we’re in a brave new world. think we’ve sufficiently set the stage here that things are different now and they’re going to keep evolving at a exponential accelerating pace over time. So the fourth wave industrial revolution, it refers to a group of technologies like AI and

cyber physical systems. talked about autonomous driving, IOT, robotics, right? All these things and how they’re going to start to permeate society and, business as well. so how will businesses begin leveraging this technology and, and, know, will that innovation, fuel the evolution of business models fundamentally and, and will it fuel, you know, a shift in how they, budget for technology versus labor costs?

Yeah, I think it’s such a strong article because it was written not too long ago, right? Like it’s not, it wasn’t written in response to an AI innovation. It was written, you know, in the early days of saying like, okay, well, if the industrial revolution really changed the way that the economy worked in terms of like how labor arbitrage worked and how we mechanized, know, these, these, frankly, in the Gilded Age, probably all the gopolistic versions of businesses.

Adam Winston [46:11.138]
which is very different design than frankly, the farmer design and barter system and thing that we were dealing with for many, many years before that. Right. And it kind of happened really quickly. And I think that if you look at that speed of evolution and where we are with this, we are getting these exponential gains every few months in the technology and we are not seeing the exponential market impacts.

outside of maybe the US stock market, right? And saying, okay, well, I’ve seen these large valuations on a few companies. Am I seeing this kind of rebuild towards a different economy? Not necessarily, right? Even if we look at the AI and adoption projects across all the millions of businesses that exist here every day.

You’re not seeing a single massive sweeping transformation. You’re seeing adoption from the user perspective. We’ve all used chat GPT, I think almost a hundred percent of us probably now at this point know what it is and use it maybe every day. The idea that we’re probably using it for certain interfaces like, you know, text and other things and maybe rewriting papers and things that we wanted to have a different, you know, tone to or something like that, generating, let’s say pictures or video or source code.

And so when you look at the job and the fourth wave, the fourth wave industrial revolution, if you’re interfacing with a computer to do your job, you’re sort of on the first block of the people that are going to start to see this disruption, right? You’re the first ones that are going to see your, you know, that, that reasoning and that input output of what you’re building on that computer, be one of the first things to, sort of evolve. Right. And that’s just because of where it was born.

Right. was born in the place where we’ve been basically improving on compute for the last 50 years. And this is basically the final stage of that is to say like, okay, great. We don’t need an interface anymore. Hardware, software, person, computer. We will get to this point where AI will take their call on them white collar jobs, the jobs where you use a computer to do most of your, your, your stuff. Right. Then the next few areas, the robotics and IOT revolution are around.

Adam Winston [48:27.116]
the other aspects of it. So you don’t just, you can’t just say, okay, well, what is AI? Well, AI is the sort of virtual training model that builds on weights and makes decisions and generates stuff. You go, okay, great. So for the human computer interface, that’s going to be something that innovates first. And then the next thing is over to the machines, right? To say, okay, well, how do we make a car? How do we deliver a package? How do we farm an apple? Right? All those things require highly mechanized robotic things, things that people did with ILR.

operating a specialized machine or themselves with their own hands, right? And so you see robotics, I think a lot of stuff that, frankly, Tesla and probably China are doing really well is building things like, you know, mandibles and other things like that, try to emulate what humans have been doing to say, okay, well, is the first step to say, let’s get a robot that operates the machine, right? A robot truck, truck lift driver, or is it a robotic truck lift? Forklift, sorry.

And if you have the concept of like, look at like a dark warehouse, like Amazon’s robots moving stuff around, they built an AI native, right? So they put the genius into the robots and then there’s a robot operator human on the loop somewhere outside of the factory, right? And so, you know, the concept of that evolving, you know, requires a lot of other things in engineering to catch up, but even they catch up, right? Because those engineers were using the computer to build the designs.

to do the CAD drawings for the chips and the LIDAR print program and all that other stuff. So this idea that there was like, one innovation won’t affect the other. You’ll see one boost the other tremendously. And then I think at the end state, we’ll be sitting somewhere like we were when we were thinking about the build out of the internet. Like if you saw in 2004, Cisco, Stock, Crater, but then this new company called Facebook, right? That was the time period when I was getting my first job.

And I, you know, my dad was saying, Hey, Cisco is a real good winner. They’re going to win. You should probably get all these trainings and go to Cisco because that company’s not going anywhere. Right. That that company is the internet. And then you turn around and you’re like, my friends are getting a job at Facebook. And they’re like, well, this is the internet is this new thing that people don’t understand called the data economy. And nobody understood this thing. And people were sort of watching this thing change. Right. And I feel like at the

Adam Winston [50:46.828]
I want to say the inflection point in the fourth wave of industrial revolution, we will be sitting in that kind of a dilemma of saying like, okay, there’s going to be a new way we have to think about the economy that we’re not today. And there’s going to be an old way that we have to sort of detach ourselves from and say like, this is peak and that that might be AI itself, like what we think AI is and what we thought the internet was and the build out of it. If I could change the players today and be like, Nvidia is Cisco, right? Open AI is Facebook, right?

They are these things that we are basically moving through that we don’t fully grasp yet what it’s going to mean and what Facebook has meant for business or meta now. And what, you know, Cisco still means to business. There’s still vestiges of it for sure. Right. But that didn’t modify the economy to some dramatic extent that we didn’t completely under, you know, I have to underwrite everybody’s salary because we were so worried that

No one was going to be able to ever work again, even though at that time, if I rewind and tell you like back in like the crazy days of 2000, that’s what people were saying. They’re like the, internet’s going to, it’s going to ruin everything, every business. And you know, it didn’t, right? You still to get your hair cut somewhere. So, you know, the, the, the math was, um, at the time didn’t look like we knew when it was going to end or what was going to happen. And in retrospect, we saw it as a very simple cycle.

And if I look at the fourth wave industrial revolution and think of it in terms of those cycles, it’s a compression mechanism where they’re saying like, that’s going to happen more rapidly and more often than we were used to at all. So I think the most impressive thing about the comment of fourth wave industrial revolution is the idea that all those things working together will shrink that not into a hundred years, not into 20 years, but maybe like 10 and 15 year cycles. Right.

Yeah, it’s funny. You mentioned the stock market being one area where the economy is seeing value creation through AI and some of these other fourth wave technologies. But the stock market these days is not based on fundamentals in any way, right? It’s all speculative. So, so, um, and, know, we just saw the MIT research paper come out. Everyone’s probably read it that 95 % of gen AI pilots fail, right? And, and

Shawn Cordner [53:07.202]
I’m just wondering, are we at some point going to see a disconnect between the promise of AI and the lack of value creation, actual value that is coming from the bets that people are placing on it? And is that going to potentially slow innovation or maybe cause a bubble? I to use the B word, but I will.

bubble, the big B. Yeah, the stock market is your rich, most dramatic and reckless friend, right? If you look at, know, it’s in a sense kind of gambling. And what you look at the market fundamentals of this sort of concept, what people are saying is kind of like an infinite money loop of, okay, we’ll Nvidia, we’ll invest in OpenAI and OpenAI, we’ll invest in Oracle and Oracle will build a plant that does this. And then there’s, you know, seven or so companies that

have, you know, lion’s share of what the, what the stock market’s total value is. On top of that, you’ll say, okay, well, if a company like Nvidia is in the multi-trillions of dollars in terms of valuation, what is the expect, what kind of realization of value does, like, what do they have to actually have companies sell to make that build out company worth it? Right. And some people have floated the idea that this is like, you need $8 trillion of value to make that valuation make sense.

And that’s a third of the U S economy more or less needs to be basically spending. And in order for that to be where it’s at in terms of valuation. So is it overvalued? Is it realized gains? Not really. Right. Most of that stock, anybody, you know, I understand the idea of like an immediate gain is selling your stock and making a lot of money. And I’m sure I have a lot of happy friends from crypto to whatever that I wish would stop talking to me about it. But, the idea is.

That even though there are basically really high valuations on this and we do see sort of, you know, a opening does like $12 billion in business. think they lose 10 of it. Right. So, so the idea is like, okay, market fundamentals. you, if you brought Warren Buffett, you know, back out of retirement in next year and said, Hey, can you sit here and please explain to me the market fundamentals of tech? tell you what he always told people, which is that like, I never invest in tech because it’s so.

Adam Winston [55:28.226]
difficult to apply market fundamentals in this space. It’s not that he doesn’t understand, he’s incredibly sharp, probably one of the sharpest minds out there. He doesn’t believe that it won’t pivot away the value quickly in disruption. And a great example would be like, there could be a future state where open AI does build the super intelligence. And then three months later, DeepSeek figures out how to do it for like one 10th the cost and we’re done here. You know what I mean? So the business itself is devoid in some ways of the innovation.

Cause the innovation isn’t the only moat in business. But that study around, I think it was MIT that did it, the 95 % of projects fail. You know, that obviously got a lot of press because people are the competition to this innovation. Their jobs, their livelihood in many cases is what’s being advertised as the savings. Like you get on earnings calls now with a Microsoft or whatever, and they’re not.

announcing their labor reduction, the 10,000 people they’re firing or whatever it is, as a negative. They’re announcing it as a positive. They’re saying a third of code at Meta is now written by AI. We don’t need all these programmers and that’s a good thing, right? Where those earnings calls would say like, you’re hemorrhaging people, this business must be doing terribly. This would negatively affect your stock value, right? So the idea that they’re almost like proudly exclaiming.

that this is what the tool is for is to replace labor. And they’re not so subtly admitting that that’s what is happening in their own companies, leads people to want to fight against it. Right. And there was this, I think it was a trade show where they brought out the Tesla robot. And the first thing people started doing was pushing it. They hated it. They were already bullying it. And cause it’s cause they realized this thing is almost like a threat. Right. And so it’s funny to see people react that way. And I don’t want to take.

the study and not hit all its nuances, because obviously, I think it was a really good thing to do that we should keep doing like Gallup and others should do, is to measure the emotions around this, measure the actual attach rate by businesses, everyday businesses, not just the stock market, do actual funded studies, not just cash welfare programs, one off on UBI, do real assessments on what this could mean.

Adam Winston [57:48.206]
Because I feel like we’ve all taken our eyes and focused on completely other things in the media and I think we all know what they are instead of this, which is that we should be doing real research on what this actual impact is. I don’t think we have any one study, any one group that should be doing it. It should be all the institutions that have done this kind of stuff in the past to get us the ground truth here.

Because on the one hand, I’d say I could read that headline. I could read that, that paper and say, people are rejecting AI because they want to confirm that they’re still important. And why wouldn’t you do that? Why wouldn’t you fight every day to make sure that you’re, you know, food on the table and happy and have purpose. Right. And at the next breath, you’re saying, okay, well, does it mean that this thing is actually just all hype and we’re not paying attention to the problems that it’s not solving?

Right. It’s just overly sycophantic, you know, tool that just says, great job. I got the wrong answer, but here you go. And people are saying, well, I can’t use that for my call center. Cause that’s just going to aggravate people to the point where they get to a human. I’ve just lost my customers. Right. So there’s this concept of, of not truly getting even from that one survey, the ground truth, but there’s a lot of things you’d say on market fundamentals around. Yes. The companies based on market fundamentals would seem overvalued.

the number of projects that I think we can point to to say AI was a roaring success are limited and that, you know, when we go to try to adopt this thing in the enterprise, it takes a little bit of thinking. think people think you can just kind of like slather it on top of the business and it’ll just work. And I think what they were saying in the study is without a good consultative hand, they do fail, right?

Yeah, both that MIT study and also McKinsey put out a white paper on fourth wave industrial revolution. And both of them point to the pilot to scale gap as being a problem. that might be the fatal flaw or the weakness at least that keeps us from seeing that value creation in these technologies.

Adam Winston [59:58.958]
Yeah, I think there’s a whole renovation. It’s the greatest time in computer science and history. I, you know, I, well, a lot of people don’t think that because they’re not getting a cool new job at one of the, you know, campuses that were around in 2005 as advertising is what a cool tech job looks like. Um, we have to realize that everything’s going to get rewritten. Everything, almost every, you know, structured e-commerce, SaaS app, whatever it is, every piece of software that we had written in the past is going to get rewritten.

And it’s going to look a little different and then there’s a lot that’s happening there to make that true. And so even the idea of saying like, look, I want to connect my AI to Salesforce. It’s like, or do you need even to store a file? Right? So like we’re starting to ask ourselves questions about, you know, what data, you know, like what is data is data something that the AI can just memorize and create new analytics for you on the fly? Cause it can, you know, use that.

unstructured approach or a different approach to how we’ve been doing this sort of classic system of record and transaction system and databases and things of that nature. As soon as you disrupt that concept, which is coming up, right? As soon as you disrupt the concept of an interface, we’re back to a whole different ball game for so many companies that have used a hundred plus applications on average in their businesses that are based on this old model. Is it now going to be sort of this

One interface that’s interface, do I need 50 travel websites when I’ve got this thing? Do I need, you know, word processing and a cloud storage of my data, or can this thing, you know, memorize and get what I’m saying and then just sort of transform, transform that. so there’s, there’s this idea that we, we still haven’t really gotten to the final state of what the entity is, or at least an all technology innovation plateaus eventually. And it’s, it’s in that plateau that we use it and work with it the most. Right.

It’s what it’s not changing because frankly, I mean, if there was a new version of Excel every week, we’re not using it. Right. So, so the idea is we do have to have time to kind of digest and live with it for a little bit. And the way that it’s kind of moving, I think they’re sort of seeing how quick, how quickly and how much they can get it to do. And then once it’s at that state, I think you’ll start to see, much more adoption in it, but that doesn’t mean that you can sort of wait behind with your hundred bad apps.

Adam Winston (01:02:23.822)
Right? Everything from this podcast being transcribed to, you know, other ways that can be modified. That would have been four or five, six different tools. Maybe it’s, you know, one very efficient tool today.

Yeah, I think you’re hinting at something that both papers point to as a potential issue creating that pilot to scale gap. And ultimately that would create an adoption problem, which is the cultural aspect of this. And if you zoom out from enterprise technology and you start talking about, you know, autonomous driving cars and, you know, having robots delivering your food instead of Uber Eats drivers, right? There’s still, I think,

a human need or desire for authentic experiences. And culturally, I don’t think those things hit the right note as being an authentic experience for us. feel weird. Now, maybe that’s a resistance that will die out over time as older generations, you know, go offline, I’ll say, generously and, and new newer generations come online, right? But like, guys like me, I still like to have a human interaction and,

I think you see that less so with younger generations that are coming up behind me, but in society and in the enterprise specifically, I think you do have that barrier of a culture that is not quite ready to adopt these kinds of experiences yet.

We’ve, we’ve done a lot, I think in business and our culture over the last five or six or seven years. I think that, you know, we’re getting used to the kinds of speed of change that we’re discussing here. People were meant to be live in person. They were meant to be working with others. They were meant to be working not behind the screen or isolated in a room, by themselves. That’s just not the way that we’ve been built. Right. So.

Adam Winston (01:04:19.982)
At some point, I think, you know, we’ve kind of used the machines as sort of an interface for us to try to expand our capabilities. And at some point it’s a job the machine has to do and we have to do a different job. And it’s because we’re not built for that job, right? We’re not supposed to do that job. And I think that everything from, you know, the emotions around things like return to work, right? If you’re a mother of three and you’ve got the job that’s remote and you can choose your own hours, it works really well for you.

If you’re a young person that’s looking to learn from the town elders or whatever, you can’t hear them talking when they’re not on Zoom. It’s negative, right? But the in-person experiences are better for us. The idea that we are going to find better and different ways to work together in a new design is going to be something that the future should bring, right? I don’t think anybody sat here and said, for the next 150 years,

Are we going to be sitting on this team’s type call? Right. And, and just started doing things this way. No. Right. And so, you know, the, the change is going to be something that’s just going to be a part of this for, for a long time to come.

I think it’s crazy to think that we would be able to adapt that quickly. This age that we’re in, even from the first industrial revolution, it’s a blip on our radar or our trajectory. We’ve been evolving for 200,000 years and almost all of that, it’s been in small groups and now all of a sudden through modern communication technology and I live…

in central Jersey right now and right across the street from me is a park called Mercer Meadows. And Mercer Meadows used to be a farm of AT &T, know, blank on the word, but foam poles, right? And they used that as a giant antenna to send the first communication, international communication over the airwaves from here to I think it was Argentina, right? So, and that was in the 50s.

Shawn Cordner (01:06:32.254)
And so prior to that time or prior to, you know, being able to lay cable across the, the ocean, just, didn’t have that kind of connection to larger groups of people. transportation is another thing that’s kind of innovated that, but all of that’s happened in a period of hundreds of years, opposed to hundreds of thousands of years of evolution that, you know, make us want to seek out other people need to be around other people. So I do think that there’s.

a challenge to overcome there as many inroads as we are making. think that some people will still be able to adopt. But on the bright side of this industrial revolution, Ford Auto, I think that here’s a promise of leapfrog technology here. If you look at how developing nations, some cases, leapfrog over technology, a great one is a lot of nations in Africa, as a perfect example.

They never put copper wire in the ground. They never put fiber in the ground, you know, to residential consumers for internet access. They leapfrogged that technology, which was expensive infrastructure to build out and to maintain. And they went straight to cellular networks. And so the culture there is built around your phone and you do all of your online banking through your phone, as opposed to on your laptop and communication is built around that. So, no, I…

I think that there’s an opportunity here to kind of leapfrog a lot of the older industrial technologies in these developing nations and maybe create a center of outsourcing in the same way that China industrialized and became a center for outsourcing. so I’m wondering just what’s your thought on how this next wave of industrial revolution may change globalization or, you know, kind of

change the complexion of it.

Adam Winston (01:08:32.543)
Yeah, I think, you know, one day we’ll be driving past the Google data center. We converted into a brewery, right? Like that’s, you know, at some point, you know, you go to any place, a lot of the fun restaurants, they’re their car garages, right? They, they, they look like it. Nobody understands that, that, you know, what a car garage looked like in the fifties, but you can kind of tell with the little, you know, glass doors that come up. This, I mean, we, we are always moving to a different shell. And if some people get to move to, you know, the,

The next one had missed a few because they got lucky that they didn’t have to spend all that energy and time on something. It seems like a positive, but keep in mind that the countries that failed to adopt those copper wires, that failed to do internet work, that failed to do that. What’s the difference between Africa and India? The number of IT professionals that probably work in Africa or India. That’s because they didn’t have the internet, the computer, the training, the whatever that would work to…

share in that adoption. so, you know, if we think about leapfrog concept globally and say, look, there’s a lot of people out there, you know, my favorite innovation would be, you know, the multi-language thing. I can wear an earpiece now when I go to our road shows overseas and people can understand what I’m saying and I can understand what they’re saying automatically. Right. That’s something that Apple released in their latest thing is just sort of auto translation. Right. So this, this idea that we can start to reduce the barriers, maybe leapfrog to the point where like,

All of a sudden, you know, it’s probably the same to work in any region of the world that it is to work here because we’re not really attached in this very narrow cone of innovation that we have that’s stuck to the ground we’re on. Right. I think we’re very close to that with hopefully, you know, a universal internet from, you know, something like satellite where you can get it anywhere where you have this concept of saying, well, you can now share all the world’s knowledge as simply as you can.

in your own language here. Remember that for most of history, right, the internet was written in English. So the idea is that, you know, there’s a lot of ways that we don’t know what they’ve written. We don’t know. They don’t necessarily know we’ve written any other translation services and everything else, but like they would be really good if it was like, you know, truly global and almost invisible. And so if we can, if we can remove those layers, the telephone pole and the visible concept of

Adam Winston (01:10:57.09)
what technology is and how it impacts our daily lives, then yeah, I think we’ll start to all look a lot more like each other, which is very interesting. That’s a great thought experiment. You get very close to that Star Trek and not Star Wars, which is a quote I learned on stage with one of your panelists. But the idea is that, you know, in the Star Trek version, right, they don’t work. They don’t worry about, you know, the races, intergalactic races that are working there. They all think, you know, it sounds like they’re speaking English in the show, but the concept is…

They’re speaking whatever they were speaking and it’s all just automatically translated to each other. Right. So there’s this idea that like, that those innovations that we kind of theorized in, you know, movies and popular TV shows and things. Those are all things that we have very good line of sight on right now. And so, you know, I think of this as like, we’re going to have a kind of.

reckoning for a few of these things if they pan out the way we think they will around like labor arbitrage and compute and energy and all the other stuff and the negatives of cyber, of course, where I live every day. But we’re all trying to solve those problems. Right. It’s not like I’m like, you know what, cybersecurity is going to be so bad. I’m just going to go start planting potatoes. Like, you know, I still believe that we can kind of fix it. And then at the end of it, it will look more like Star Trek and not like Star Wars. So

As long as we kind of all share that belief and hold that belief, even when, and not ignore each other when things are going wrong, try to actually get the studies and figure things out so that we can fix the problems as they come up. think you can have a brighter future for this stuff. It doesn’t all have to be doom and gloom.

There’s a lot of doom and gloom out there and I love the optimism. So maybe that’s a good place for us to end and end it on a happy note. So Adam, thanks so much for your time and expertise. I really appreciate it today.

Adam Winston (01:12:50.05)
Yeah, absolutely. Thanks, Sean.