0055: Sean Lane

Hey, this is Ted and I'm here with Sean Lane, CEO of Cross Checks out of Columbus, Ohio. Welcome to the show, Sean. Thanks, Ted. It's good to be here.

I appreciate it. Yeah, can you start by telling us just a little bit about yourself? And I'll ask you more about Cross Checks, but why don't you give us the elevator pitch for Cross Checks and then who you are and how you got into this line of work? Absolutely.

So, Cross checks, we're based in Columbus, Ohio. We've been around for about five years. We're predominantly focused now in healthcare space. And, you know, a short clip of what we do is we create AI bots, essentially, to help hospitals and healthcare organizations automate a lot of their workflow.

Okay. And, yeah, we focus in healthcare today, but we're quickly moving into other verticals. All right, good. Well, I'm going to want to know more about that.

How did you get into the world of AI? That's a great question. So, my background, I'm an intelligence officer by training. I spent most of my career at NSA at the National Security Agency.

And I've always been a tech nerd growing up and have done a lot of work and technology in the intelligence community, mostly working with just enormous amounts of data. I mean, as you can imagine, at NSA, we were collecting a couple hundred thousands of rows of data every second. I joined NSA in about 2004 and spent most of my time in Iraq and Afghanistan. And we were collecting enormous amounts of data, all very, very heterogeneous, you know, different from different sources.

Could have been like technical data, could be pieces of paper that were scanned images, you know, that came from people's pockets. And we really had to make sense of all that and figure out how it can be applied to catching bad guys. So, as you can imagine, it's a huge data problem. So, that's where my interest in what we call machine learning now, which back then we just called linear regressions kind of started.

And it's really to figure out, you know, how do I make sense of all this data? And then how do I apply it in software to actually get people information to help them make decisions? All right. So, how does a person get ready for a career at the NSA?

What were you doing before that? So, I was an intelligence officer in the Air Force. And I, you know, ever since I was a kid, I wanted to be a spy. So, that's kind of the path that took me on.

So, I decided that, you know, military intelligence would be a great place to start. I did ROTC in college, became a lieutenant in the Air Force, I did intelligence, spent a year in Iraq as a kind of regular intelligence officer, and then got recruited into NSA about a year after I was in Iraq. So, really just having a background in technology, the good thing about being a spy that I realized was, it's more about being a nerd than it is about being like James Bond. So, the fact that I had a technical background that I could do some coding and understood, you know, how to deal with large amounts of data, that was really the thing that allowed me to excel at NSA.

Okay. So, when we use the word intelligence in military intelligence, or with the NSA, I'm imagining that that's not the same as when we're talking about artificial intelligence. Is the word intelligence meaningful in this conversation about what you do or what you have done? Yeah, and the intelligence community, intelligence is all about the kind of information or knowledge that you can gain out of data.

So, intelligence is being able to understand intentions, be able to predict the next steps. So, it is different in artificial intelligence, but it's the same in that you're trying to use it to understand intentions and then predict on what to do next. So, it is very applicable from that perspective. Artificial intelligence, of course, is being able to do that, not humans, but with machines.

And I feel like we try to do that a lot with humans, that can be an intelligence community. But quickly, the intelligence community itself is moving to start to adopt more artificial intelligence and trying to help predict and shape intentions. Okay. So, we fast forward to cross-checks.

How did you make that transition from NSA into the private sector? Where did this company come from? So, I was on my fifth tour. I was in Afghanistan and just talking to my fiancé and my wife.

And I'd been gone a lot and she said, are you going to be gone like this forever? I said, well, probably, this is my career. And we decided like we wanted to try something different. So, I decided to start a company.

And I got out of NSA and started my first software company in Baltimore, Maryland. And we built technology systems for the intelligence community. And we did a lot of particular analytics and visualization tools. We built a cell network in a box.

It's a pretty incredible little company. I grew that to be about 60 people. We sold that company in 2012. And it was about that time.

So, I just sold the company. And I'm from this really small town in Southeastern Ohio. And they were having a huge problem with prescription drug abuse. And they asked if I could do anything to help.

And I started to look at healthcare at that time. So, it's 2012. And when I uncovered, kind of pulled the covers back on healthcare, what I saw very quickly was there was no identity layer. And what they had was a bunch of silos and information.

And they needed the internet. And there was really no way to connect all this data together if you didn't know who a person was, who a patient was. And it really was an exercise of entity extraction. So, understanding this is a patient, this is a doctor, this is a drug, and all their data to bring it all together.

And as we started to explore that in healthcare at cross checks, so we started cross checks as a company in 2012 to address this problem, we went around to a couple hundred health systems over the past four years after we started the company, giving them an identity resolution capability and building this massive graph database of every single patient, all the provider locations and lots of other very interesting entity information. And once that was in place, that was like the routing table of the data. But we determined very quickly that we needed a way to route that data. So we built the routing table, but we couldn't get the information from one crappy piece of enterprise software to another.

And that's really where we started to look at AI. And I had kind of avoided going down the trap of the buzzword search and trying to figure out how do we do AI. But instead, we thought this could be an actual solution. And the way we define AI is a little different than some.

We were very Turing AI type people here at cross checks. We believe that it should, you know, mimic a human as much as possible. So our definition of AI is AI equals RPA. So robotic process automation plus global awareness, so an understanding of entities across many enterprises and then machine intelligence.

So those three things together to us creates the most powerful AI. And that's really what we do today. Robotic process, again, robotic process, what automation automation and is, could that be considered a narrow AI by itself in another organization? Yeah, I think in a lot of cases people do consider that narrow AI.

So to me, RPA or robotic process automation has traditionally been just automating software, usually web software. So some people call it scraping, using different tools to automate native applications. That's a very kind of scratch the surface level of RPA. What really gets exciting with RPA is some of the advances we've made in computer vision.

For example, our platform uses test-racked for, which is some of the newer computer vision capabilities, as well as creating our own convolutional neural nets to be able to read the screen and understand the software. So the bot can really log into a software just like a human, understand the software, and start working within it. So that's a really cool capability. Wait a minute.

So you're using computer vision, you're using a camera to read a screen rather than have the bot be inside the computer that's displaying the web page? Or are you confused? Well, the bot is inside the software. It's not using a camera, it's for say, but it is using computer vision.

So it's really what the bot sees. It's like the eyes of the bot. So it's looking at everything on the screen, it's understanding what it's seeing. It knows that something's a search button or an inter-key or whatever it may be on the software and getting a real understanding of that.

And that's what's really hard. So your typical screen scraping, it's just looking for a word or clicking on a certain thing. This is really understanding the user interface through computer vision. So it's a whole other level of robotic process automation.

All right. All right. And what are the other two pieces you said? RPA.

So RPA plus global awareness. And that is, global awareness is a really important thing for us. So let's imagine you work at a huge company and it has 50,000 employees and 30 different enterprise software systems doing something or rather. So you have your bot logged into the software operating on it just like a human, human user would be, you know, seeing the software, knowing how to manipulate it, how to click through it.

And then as it's working, it starts to understand and remember the entities that it's dealing with. Now the entities could be customers, products, in our case of healthcare patients, providers, drugs, diagnoses, and it's building a graph. And that graph is sharing one brain. So as your bot is working on 30 different pieces of enterprise software, it has this awareness of an entity as it kind of crosses the different pieces of software.

And it's a really interesting way for a bot to be able to use many different software platforms, and understand they're talking about the same thing as they jump from one to the next. Does that make sense? Yeah. Can we break that down into an actual example?

Are we talking about like scheduling software at a medical office knowing that there's a patient that is scheduling an appointment with this particular provider, this particular doctor? That we're talking about? Yeah. That's a great example.

So, okay, so let's say in a healthcare organization, you may have a different scheduling software, you may have a different electronic medical record, you may have a different billing system, you may have a different tool that you're using to reach out to the patient and ask them questions or send them information. So as the bot is logged in and using all these different pieces of software, they need to know they're talking about the same entity, the same patient. So that's what our bots can do. So let's take it even a step further and talk about global awareness beyond enterprises.

Like let's imagine you are traveling across the country and you're in Wisconsin and you chipped your tooth and you need to get it fixed. So imagine being able to text a bot and say, you know, our bots call it all up. So you text all up and say, hey, I need to go to the dentist. All of those who you are, all of it has login credentials and is logged into every system across healthcare, can look you up across all the different systems, find your insurance information, find the dentist nearby you that can accept your insurance.

Your logged in so she can schedule your appointment for you. She can take all of your medical history and dental history from all the different places and bring it to that one place where you decide you want to go to the dentist. That's basically RPA plus global awareness. So understanding who the patient is across many, many, many systems.

And that provider, that dentist in Wisconsin doesn't need to opt in or anything. All of them can provide it to them in a way that they can use it no matter what. Yeah, so the patient has to opt in for the information to be shared. It's really the patient's call whether or not you can share the information.

In my scenario though, every provider would have to be on this internet. So every provider would have to have one of these AI bots that act essentially as an intelligent router connecting all the pieces together. Wow, okay. And what was the third piece then?

What was the second one? The third piece is machine intelligence. And this is the spectrum of machine learning from supervised to unsupervised from your basic linear regressions to your most advanced dynamic neural nets, for example, in the unsupervised learning realm. And this is really about your bots.

They're working in the systems like a human. They are understanding the knowledge in the systems across platforms. And then they're learning how to do the job better. Right now we're using historical data from humans to train our bots.

So let's say for example, at a hospital, you know, a human has looked at an insurance card and has translated an insurance card into the software by typing in what they see. And they've done that a million times. We can take that, all those insurance cards and all the examples of what the human typed in and train a neural net to be able to look at an insurance card and know what it says by basically the training of what has happened in the past. So it's kind of a convolutional neural net model.

So that's an example of using machine intelligence to just learn and get better and learn, either learn from the humans or learn as they're going. Another example would be let's say a bot is in charge of reminding people about their appointments. And the bot is constantly looking at, okay, when I text this type of person and the person type could be created by the signature of information. When I text this type of person and tell them about their appointment, I'm getting a better result when I text someone at 637 PM versus 839 PM.

So the bot can start to learn the better time to text. So that's a basic regression example of machine learning. But we wanted the bot to use machine intelligence to learn and get better at what they're doing over the course of time. So if they see an insurance card that they have not been trained on, a different format or whatever, because they've been trained on so many, will they typically be able to read that anyway?

Yes, that's the idea. So the idea is to use the neural net to be able to recognize these. It goes beyond your typical OCR and it allows, it gives the information, it gives the bot a lot of ability to see and understand things that they've never encountered before. Yeah.

Okay, so Brandon and I have the show I've been talking lately about this idea that the machine can tell you, oh, this is a picture of people playing Frisbee in the park without knowing what Frisbee is or what people are or if Frisbee is something that you throw or that it flies or it doesn't really know anything. But it's been trained on enough images that it can spit back the label of this image. How is that related to what you're doing and how is it not related? Yeah, I think it's related in that, but in our example, I mean, we are trying to approach this from more of a supervised learning approach.

So meaning I would need to have a bunch of images and have had some ground truth around what they are. So, you know, and it's just a matter of if you've had a million images where someone has said what's in the image or a human has said that we think this is a person playing a Frisbee, then you're going to train your neural net to be able to find that much more easily. The more data you have and the more training data, the better your results typically. Now, being able to find a person throwing a Frisbee when you've never had that training data, that's a harder problem.

And one that I don't think we have yet the capabilities to do that, but I do think those are coming. When you talk about global awareness, what kinds of things does your bot know about? You said patients and providers, does it actually know what those are? What does that mean that it's aware?

So, global awareness, it just knows that it can distinguish one entity from another. So, you know, Sean Lane is not an incredibly common name, but there are definitely many Sean Lane's out there and it needs to be able to differentiate one Sean Lane from another. So, entity awareness is really about understanding that I'm a different Sean Lane than another and the same thing with doctors and medications and prescriptions and diagnoses. So, what results from that is this entity map to this big graph database where nodes are connected to other nodes and it knows like, oh, this is Sean Lane, the one that lives in Worthington, Ohio, it can then go to me as an entity and see everything that I've ever done and everything that's attached to me from the data that it's ever seen.

So, it's really a way of storing information that's a lot like the human brain. So, we're storing it in this kind of neuron model where every entity has new knowledge connected to it. So, marketers do this? Is this supposedly our anonymized data from our browsers or whatever and they follow us around the web trying to give us the same ad?

Is this similar to what they're doing? Yeah, I think it is. I don't know that they use a graph database to do that. They probably should or could.

They're probably using a different database, but essentially that's it. I mean, they create, you know, like you're going to have a bee cookie, like a cookie that follows you around and gets added to your visit history essentially. So, when you go to a certain website, they're going to be able to see the things that you had visited by the cookies that you picked up along the way and they're going to be able to target things to you as a result of that. So, like if you go to a site that provides online advertising as a banner or something like that, it's going to look at your cookies that you picked up along the way.

If we could go, you're looking at backpacks at REI. It's going to throw you an REI backpack banner from that. So, most of its data that's collecting in real time and not like a global knowledge base that they're storing forever. Most of it, they're just picking it up in real time.

I see. So, you're talking about interacting in physical space. If you're as a patient of interacting with Olive, then Olive knows which doctor you went to. It sounds like things are not, it's more of a direct connection in some cases.

How much do you need to infer and how much is things that's directly told, Olive? Yeah, most of it's directly told. I mean, we're, you know, it depends on the enterprise, you know, whether it's healthcare or another, how much risk you want to take on inferencing. Most of the information that, you know, Olive understands is stuff that we are pretty certain about.

So, we're trying to create a better experience for the patient ultimately and a more cost efficient healthcare system in general. But we want to make sure we do that one step at a time to where we're not taking too much, we're not taking too many leaps in trying to connect pieces of data that maybe don't need to be connected. So, we're doing, only things we're pretty sure about at this point. So, if we could then jump into the future and we can either jump into Olive's future into Cross-Tec's future or if it's better for you not to divulge anything proprietary, but where you're ahead and we could jump into a generic company in your space.

I don't know how you want to talk about that. But, you know, five years from now, how much change will there be? What kinds of things will an Olive like system be able to do then that it can't do now? So, we think of, you know, I think we're at this moment right now that's very similar to where we were in computing in like these kind of 70s, late 70s, early 80s where AI is a lot like the mainframe, right?

So, like people, there's a mainframe computer out there, not very many people had access to it, not a lot of people could develop on it. It was this magic box that did something incredible that not a lot of people understood. And it wasn't the mainframe that changed the world of computing. It was the personal computer.

It was that kit that those, you know, Steve and Steve built in their garage and then gave to everyone kind of democratized the capability. That's what changed computing forever. Gave everybody the tools to build something on a computing platform. And that's what really made all the stuff that we see today.

And I think we're at a very similar point with AI. There's things that are occurring that are very science project like that are things that blow our mind when you think about it or hear about it, but we don't have access to it as just a normal person and we can't build upon it. So, I think the thing that's going to change AI is democratization of the capability. And that means basically anyone building, just as ubiquitous as building a website is today, building an AI bot will be in the future.

It won't take an engineer to build an AI bot. Basically, anybody will be able to build an AI bot, automate some part of their job or their company. And that will build this new digital workforce that will be millions and millions of bot employees. And I think that changes everything.

And I think that's the timeline of that is actually a lot shorter than we think. You know, I think in the end of, by the end of two years, we'll have a million bots in the marketplace doing jobs because we democratize the ability to build those bots. So, I think that's probably where we're headed in the next five years is an incredible bot workforce built by regular people. And do you see those bots augmenting existing employees or do you see that taking over for existing employees?

So, we call this concept scaling humans. So, we think over history, humans have scaled and they've scaled because of tools. Like, when we, before we had fire, we could only do so much, but once we got fire, we could do much more. Before we got the combustion engine, we could only do so much.

After we got it, we could do much more. So, that's happened so many times over history with fire, the combustion engine, the wheel, computers, the internet, the printing press. I mean, all these things have made us as a human society to be able to accomplish way more than we ever could. So, we think the next tool for that big scaling movement is AI.

So, we don't see it as a replacement of humans. We see it as a replacement of what humans do. And we think that AI will be that next big scaling moment that allows us to do things like cure cancer, you know, occupy Mars, create a new, you know, reusable energy source. I mean, whatever it is that we really need that we consider a moonshot today, I think AI will help us get there because it will replace what we do.

And all of this, we're still talking about narrow AI in today's definitions, right? We're not talking about general intelligence yet. Yeah, I think, I think broad, narrow AI will have a bigger impact in the next 10 or 20 years than we can even imagine that generally are what. I mean, even if we did achieve general AI, it would be very narrow in its application.

So I think that it's more important to get broad adoption and kind of democratized access to narrow AI than it is to even go towards the general AI path. I think that broad, like I said, broad uses of narrow AI will actually have more of an impact in the next 10 or 20 years than anything in AGI. So, everybody right now has nails and they're using rocks to pound them in. And you're proposing that we're all going to have hammers pretty soon.

Yeah, I think we'll have a nail gun. Yeah, yeah, yeah. Okay. All right, so what's the case then for general intelligence?

Are we going to where general intelligence is? This world of narrow AI, there's so many applications for narrow AI to potentially make the world better. Is there a case, a business case for general intelligence? I think there probably is, and I think we'll definitely get closer to it.

But I don't see the path as getting to AGI just by kind of leapfrogging narrow. I think that we underestimate the power of narrow AI. And I think that if you give a billion people a nail gun, they're going to build cities that we can't fathom. We can't even imagine what we can build with these capabilities.

And I think AGI will be built upon those cities that don't exist. Does that make sense in the analogy of the nail gun? We will build things that we can't even imagine today by the proper use of narrow AI. And then I think AGI will build upon that.

But if we try to go directly to build AGI today, I don't think we will be successful. All right. Let me run a question by you. Brandon and I have been trying to think about how far into the future people think AGI is going to be.

I'm going to say, whatever you want to call it, and what we've been saying is, well, what if we could use three numbers? Say that the first number is, when do I think there's a 5% chance that AGI will be developed? That might be whatever year it is. And then by what year do we think there's a 50% chance that we'll have AGI?

And then, well, when do we think it's nearly certain? When do we think there's a 95% chance? And so we've been putting numbers on those, putting years on those. Do you want to play that game with us?

Could you speculate yourself? When do you think there's a 5% chance, a 95% chance of AGI have been developed? So let me go. Let me start with a 50%.

I'll start in the middle. And I think that by 2045 there's a 50% chance. 2075. Okay.

And then a 5% chance, I think probably true 2025 only because, well, actually I'll say 2027 because we usually underestimate what can happen in 10 years. So if I'm completely optimistic on that view and saying we will underestimate what we can achieve in 10 years, then I'll say 2027 is the 5%. And should we be worried about that? I think we should be worried.

I think we'd be worried if we don't. I think we need to make sure as we are developing that or developing towards that, that we build the entire ecosystem around it that's required. Meaning we're going to need to enforce the walls that will probably be built that don't exist today that help govern AGI. I think we'll need to build all of the protective capabilities within networks to combat or at least mitigate some of the problems coming from adversarial AGI.

Because for everything that we build, there's an adversarial side of it. There's going to be bad actors that build bad AGI. So we're going to need a police force slash military capability that combats the bad actor AGI. And it'll probably have to be AGI itself.

So we'll just need to create those infrastructures in that ecosystem. As long as we do, it's just like with anything else. Cars are on the road now. They have rules and laws and we've protected and forced those and we keep cars in check.

Still, we've accepted the risk of all the deaths that occur because of cars. But it would probably be hard to argue that the world isn't a more efficient and productive place because of cars. And we're willing to accept the risk of what happens along with it. All right, so this argument worries me a bit.

And let's talk about this a little bit. We know that there are problems with cars sometimes and we have to recall them and we fix them after the fact. We try to make them perfect ahead of time. But sometimes the brakes don't work on a particular model.

Sometimes the airbags have a problem. The same thing seems true for our software where we either have bugs that make the blue screen of death or the spinning wheel of death and we fix it later. And this is a model that we're well acquainted with. Is there, I'm worried that there may be some reason why with AGI it could get away from us such that we can't fix it after the fact.

Am I being realistic in that worry in your view? You know, I would say that the consequences of a bug in AGI could be much greater than the consequence of a bug in your email. So from that perspective, if the consequence is more grave, then there's probably be more consumer protection thought about before it goes out. Just like with cars, it may not seem like there's a lot of consumer protection that goes in.

But they look at the consequence of the wheels falling off and then they make sure that that doesn't happen most of the time before the car ever gets released and allowed, for the public allowed to purchase. I think that will probably happen with AGI. There will be some structures around what's able to be released to the public. It'll be a public good issue to make sure that there's a regulation around it.

It will be tough. I can't even imagine what the Wild West is going to look like in the first five years until we figured out. But there will be a period of time before we've gotten all the regulations in place that will be very, very interesting. I think the difference between cars and airplanes might be instructive here too that we do have consumer protection in place for both of them, but airplanes basically never fail.

More or less, we can almost say that that's true. There seems to be a different way we approach that industry. Because again, I think the consequence of failure is more grave. If a plane holding 400 passengers falls out of the sky, it's a much more grave consequence than one wheel falling off one car.

So I do think that we follow that model. So I guess with all that said, I would hope that we measure the potential consequence of the AGI and we apply the right kind of regulations to that to match it. I just hope that happens in the future. If there's really great consequences to it, then we should be pretty strict about it.

Do you think that we are being alarmist on the show when we are asking the question, is there a potential existential threat? We answer that as a yes. Do you think we're being alarmist there? It's hard to say.

I think that it's okay to be alarmist because the people who talked about the possible consequence of airplanes before they became pretty common probably led people to be so strict about it. So I don't think it's a bad thing. I don't think that the future will ever, I don't think the reality of the future will ever match the alarm because of the alarm. You know what I mean by that?

Because there's alarm, we'll fix it before it's a problem. So you'll never be right because the alarm will seem like it happened unnecessarily. But the reality is the alarm may have helped us get to that point. So does that mean that the real use of the paperclip maximizer toy thought experiment is that everybody knows it and every AI developer will be on the lookout for replicating the paperclip maximizer problem?

Possibly. I think that, again, you can never assume that there won't be the bad actors. There will be plenty of people looking out for the problems and getting ahead of it. But there will also be a lot of people who try to take advantage of the possibilities that will arise as these technologies start to advance.

So the good guys both are looking to make sure their own systems are not simply paperclip maximizers that could run away. And they're also, the good guys are also thinking about how the bad guys might intentionally use a paperclip maximizer type system and they're trying to mitigate that sounds like a defensive capability. Right. Exactly.

And just like new technologies usually create new economies, like the internet, perfect example, like it is created its own economy. I mean, there are billions of dollars that are there now because of the internet that weren't there before on the ones who use it, the ones who protect it, the ones who attack it across the board. The same exact economy will be created around AI. I mean, it'll be new, a new, you know, new trillions of dollars of economic kind of movement because of AI that doesn't exist today.

So the good guys, the bad guys, the neutral parties, everybody. Yeah. Ten years ago was when the first iPhone came out. So you putting 5% on AI in the next 10 years.

If we think about Uber and Lyft and Airbnb and all these companies, then all this stuff that's been created as a result of the smartphone, it really is quite astounding. And that doesn't even take into account the exponential on how the next 10 years could be much faster than the last 10 years. Well, I agree with you. But I think we're still two years away from our iPhone moment with AI.

I think that, and when I say iPhone moment, I mean that ubiquitous tool, that platform upon which every single person, not just an engineer, but like every bomb pop kid, whatever, can build an AI bot, I think that is our iPhone moment. And when that happens, like the 10 years after that point, so where we are today, 2017, will be something that will be very difficult to even imagine. And do you think that that 10 years will be less than 10 years because of the exponential? You know, good question, quite possibly, yes.

In a lot of this, because we know now what a model of exponential growth looks like. So we're likely to beat the last one. So if you think about where we went with an iPhone from 2007 to 2017 with the iPhone, that's not so foreign to us now. We understand that we understand how venture capital works.

We understand what creating new business, how the speed of that, how we can do it at scale. So if you take all those things and use that as like our own human learning, we will probably exceed the rate of growth of the last big wave. And I think that, yeah, that's fair to say that it probably happened less than time. All right.

So now if we're thinking about somebody looking at either a career change or they're just about to go to college or they're going to skip college and do something else, what would you recommend for a person who wants to make an impact in this new AI world? I think it is about finding practical applications for what we're calling narrow AI. So going down the path of trying to create AGI or help towards that path, I think find a way for narrow AI to have a practical application into every single nook and cranny of the economy and the world today. If you find your niche of like, hey, some simple machine learning or simple AI, an AI bot could do X.

If you figure that out, you're going to be set. And I think that building of that broad platform of narrow AI is going to be the key. So I would recommend that people start looking at how to create an AI bot that has a practical application. I actually just signed up for a bot the other day.

I guess it works in Facebook Messenger. It also seems to work in my email, but it's looking at my bills and offering to negotiate with my cable provider. And it'll take 25% of the annual savings to get me a lower bill. Yeah, that's awesome.

That's great. I didn't even know such a thing existed. And now there's a bot that's talking to me in Facebook, Facebook Messenger, that's trying to make money by saving me money. Yeah.

Yeah. I think most people will end up interacting with hundreds of bots on a weekly basis. You'll have your set of bots that you use and you'll be managing those and then you'll be able to spend more time thinking about harder problems. Yeah.

All right. All right. Well, let me ask you a sort of final question. What questions should we be thinking about as we on this show, thinking about the potential existential threat and thinking about where AI is going?

What should we be considering to help people think about this future? Yeah. I think the reality that we do with today is you have four or five major tech companies that control most of what we know and experience in life. So from Amazon, Google, Facebook, Apple, a lot of our interaction with AI today is really controlled by those organizations.

So the question is, do we want them to control all that experience for us moving forward? And are we okay with one organization, for example, having all of that power capability? Or do we think there will be a new player or two or three or four that come on to become that AI, kind of who we look to for AI? And that's a question I ask.

Like, is it going to be one of the big ones that exists or a new one? And how can you, as a consumer, kind of experiment with other companies to see how you could use their AI products to see if there's a new thing to leverage versus what we all know is great tech companies? And then the other thing would be just asking yourself, what if I had automation, if I had an AI bot doing some of my work today, what would it be? And if it was doing that, then what else could I do?

And if you think of the world through that lens, I think you will see that there's probably a lot of things that could be automated. And there's probably a lot of, there's a lot of harder problems that you'd want to tackle. So there's the questions I kind of ask people often. Awesome.

Sean, this has been really fun. Thank you so much for taking the time out of your schedule to share with us in the concerning community. How do people either find out more about you or cross checks or get in touch or how people find you out there on the web? Yeah, so cross checks, C-R-O-S-S-C-H-X.com.

From there, you can pretty much find anything about our company. You can find me, my email address is on that website. You can look me up on YouTube, Sean Lane, S-A-N-L-A-N-E. I have a lot of, I'm out there.

So if you want to get ahold of me, it's pretty easy. Beautiful. So I hope our hope our listeners will take you up on that. This has been a really fun conversation.

I really appreciate you taking the time. And hopefully, as a result of us all being engaged in this question, we can all have all the good stuff from AI and avoid it as a claimant. Thank you. No problem with that.

Thanks so much. I appreciate it. It's fun. All right.

Take care. Bye.