The Role of AI in Society: Tools for Productivity or Potential Threats?

Hello, Václav, thanks for doing this. How are you feeling?

Hey, Mike, good morning, good afternoon, good evening. Yeah, I'm doing pretty well. Thank you for asking.

Any new goals? Because we're still in January, I still feel like I can ask this question. Any goals you set in the beginning of the year for the upcoming year?

Oh, I never do that. You know, that's a loser's game to think that, you know, just at the beginning of the year you start doing something special. No, I'm just, you know, working hard, trying to work hard, trying to get a sense of technology, trying to run businesses. So yeah, all good.

All good. Fresh new year. So tell us a little bit about yourself for the listener. Who's Václav?

Um, when I go to parties and people ask me, I say I'm a computer guy, I work with computers, and usually that shuts down any further conversations and questions because nobody wants to talk to computer guy. Uh, I started with computers when I was in almost elementary school, high school, and I was working as a summer job. I get on work on mainframes, so that shows you what kind of geek I am. And then I studied the electrical engineering and I got more into computers. I moved to Canada, I started company, an IT consulting company, and I've been doing this ever since. So last 30 years.

Okay, so that's a huge, huge background. Do you get any like old school gadgets? I just for the sake of, you know, giggles, I carried like the old floppy disk. Just to show young people like, here you go, this is like 1.6 kilobytes. That's what we used like 15 years to do anything like that?

I reminiscent about good old technology, right? This is one of the things which, and I'm sure that will be topic for further discussion today. We don't evaluate technology on its merits, usually on its hype. And this is the, this is trying. So one of my, you know, gadgets from past, this BlackBerry, you know, one of the best engineered device we had for communication. And even if people think, oh wow, we have today Android, look at what iPhone can do, uh, BlackBerry was doing this 10 years ago, right? So that's, uh, that's the sad part, that the business didn't work for them, but the technology was far superior to what we have even today.

I do like that saying that we evaluate technology on a hype and not on its merit, which leads us very nicely into AI conversation, which we are aiming to do today. Can we try to define AI from your perspective? So there's basically not a single day I don't see some type of like AI/ML article right there, you know, online. And it seems that AI/ML is being used interchangeably, like two the same things, but which they aren't, right? So, and everybody has their own definition. Someone doesn't even know what it means. So for you, Václav, like, what is AI? Like the current hype AI, how you would define it?

Uh, it's, it's absolutely nothing. It's a marketing term. It has, it has no, no meaning, uh, because, uh, if you're talking about computers Yes, it's artificial, and the word intelligence, we as humans, we still cannot agree what intelligence is. And suddenly we defining that there is a class of software, of technology, which suddenly deserves this term. For no apparent reason, we start using that, and it has this, again, it has this hype, and it has this lure of something, and people just love to use the term. Yeah, it's very similar to when using— you no longer using phone, you're using smartphone. And like, how smart is it? Like, can you measure how smart is your phone compared to how smart is my phone? It's, you know, you cannot even measure it. You measure maybe on features, but there is no quantitative way how to determine if phone is smart or stupid. And same thing, you know, we're using term artificial intelligence, uh, which of course it— one can debate that it encapsulates the, the term machine learning. So using AI and ML as one thing, again, it's nonsense. But again, that's my, that's my opinion.

That's how you define it. It's effectively, it's a marketing term. Like there's no real merit behind it. Okay. I buy into it. I do like, you know, large machine, you know, large language models probably more than AI itself as a term as well. Would you be able to, you know, point a finger and just sort of have some use cases or example where the media hype just basically didn't deliver? Like, you know, we heard something like, you know, AI will do this and that, and it just didn't happen.

I would say almost every single day, any article you read about AI, it— there is the desire, and perhaps because I'm the computer guy, I don't understand the desire to create machines which are like humans. And we constantly, uh, seeking to, to have conversation or to have a companion which will deliver some, something, some original thought to us. So we get enchanted by that. And, uh, you know, it's far better to have conversation with you and to, to find new things than sit in front of ChatGPT and, and trying to have conversation and then start to debate whether the, uh, the ChatGPT can think, can have feelings, and this type of nonsense. So it, I, I would say it's every single day when, uh, we start assigning human traits to technology which just does not have these traits. We're just hoping it will have one day.

So come, come back to the, to the overhype. Effectively, we're seeing a lot of these promises majority of them, if not all, not getting fulfilled. What is the danger of overhyping AI?

It creates a false sense of capabilities which don't exist. And, you know, example, ChatGPT. And I wrote many articles, so I'm on record on that. So it's not just what I'm thinking now. It's Uh, we, we position the technology that it can do something which it is not capable of. And however, because ChatGPT is large language model, it feels like it's talking to us and it creates false sense of knowledge, false sense of relationship, and that lures people into believing that it can do something. And the example is that you ask question of ChatGPT and it delivers something. And then you say, yeah, but that's not right. Give me a different answer. And the machine says, yeah, sorry. And provides another answer, which might or might not be correct as well, because it's not knowledge model. It's a language model. Which model. And the people who are in the know, the expert who should understand this, you know, particular topic or subject, can distinguish whether it's true or not. But person who does not have this understanding, who is not aware of the limitation of this technology, suddenly start believing, saying, yeah, if I ask ChatGPT about something, I get answer. And the danger of the technology is that it will create gap between smart and stupid. Because the smart people will know immediately how to use it and use it to their benefit. And the people who are, who don't believe or who start believing that it's whatever the system supplies, it's true, suddenly get lured into nonsense. And that's a, that in my opinion, that's a challenge.

And they don't even know that this is nonsense, isn't it? Because they effectively, they take it. And when I think about the dangers, I like to add AI poisoning in the mix as well. So it's, or data, you know, data poisonings where we can effectively feed just false information. Did a lot of modules.

100%, you know, and data poisoning that has been around, you know, since, since beginning of time when people start talking to each other, they with stories, they start poisoning, you know, with different stories, they start altering the stories to fit the narrative. So that the concept is not new, it's thousands of years old. But again, with this technology, it can go much faster. And the data poisoning, if done correctly, you know, if let's use ChatGPT, not that I want to pick on one particular product, but it's a good example what everything can go wrong and it will apply to any similar technology. So when they were feeding ChatGPT, it was humanly impossible to verify all the content. And so if you are adversary and you start feeding, because you understand the training algorithms, and if you start feeding with content which it doesn't look bad, but of course it is in certain context. And then in certain context, if you ask the question, then you get on purpose wrong answer. And again, this is another problem with AI, because if you have your machine, any machine, your computer, your smartphone, you know, any technology, you can test against predetermined inputs. So, you know, if I press this button, it should do this. If I do this, it should do that. And let's say you have a million tests which you can provide, which you can automate, and if it passes, then you know it's working. With systems like ChatGPT, you cannot do that because there's infinite number of inputs and you cannot determine or predetermine what the output for each input should be. So we cannot even test systems like that. And so we are in complete dark about the quality of the system.

It allows shaping narratives, isn't it? So if I, any queries you might ask, let's give it like, you know, political ones. I could just feed in information around, you know, some of the personas, events. And if you don't really know the history, you know, what you will get, and if you take it as a, you know, without sort of, you know, grain of salt, you just take it as information. That's effectively an example of opinion shaping or biased answers, which can be fairly easily achieved with data poisoning. And for me, that leads to this danger of, in a healthcare industry or like a financial fraud industry, where we're seeing these products come out. Like, I've seen countless amounts of startups utilizing, you know, LLMs, maybe even ChatGPT, to create this like a, like AI doctor where people can come in and consult. Where on the surface it makes sense because that's pretty much what, what you get when you go to, you know, to your doctor. They will open some type of Wikipedia, they will put in the symptoms, and they will, you know, try to work out what's happening. So in a way, you think that could be automated, so that makes sense. But there's this looming danger, like How do you know what information you have put in that large model and how do you can trust what it's actually saying?

Yes. And I would add another example where data poisoning can be even, you know, deadlier, like definitely the healthcare system, 100%. Another one is self-driving cars because Again, numerous examples you can find where altering the road, the marks on the road, can suddenly tell the car to either stop, go to the left, go to the right, speed up, something. And again, it's another very difficult technology to troubleshoot because there is infinite number of situations which you can encounter on the road. And as a human, you, you can, um, uh, you learning every day, you, uh, you driving on the road, you learning that. And that's why you can then, uh, avoid certain situations or you know how to react in certain situation, something, uh, which you never seen before. You know, the typical example is, um, ambulance is going through the city and cars are, you know, trying to provide right away for the ambulance. Where— and this is first time— and how do you know what as a self-driving car you can do? Can you drive into the intersection just to let the ambulance behind you to go through? And under which conditions then you can drive even if you have red light, right? So these are nuances which you wouldn't even think about. And yet when you're designing system like this, you have to think about. And that, that, that's a problem. And, you know, the healthcare, uh, so let's not pick on ChatGPT, we can pick on IBM with its Watson system. And what's, what started as a proof of concept during the Jeopardy, I don't know if you know the game show in, uh, which was run, which is running in U.S. And IBM built a system which was able to win the competition, and it was a great engineering achievement. And IBM then took it and, and they, they transform it into basically platform, and they start putting various technologies to it. And one of them was targeting healthcare, and they spent untold amount of millions, if not billions of dollars on that. And if you search now for IBM Watson Healthcare, you see zero, nothing. It proved to be so difficult that it actually, at the end, it was adding more work to the, to the doctor because the doctor had to verify outputs from the system on top of his/her regular work. So that's the challenge. When we talking about simple questions you're asking ChatGPT for fun, that's one thing. But when there's a human life online, that's a completely different story. And the doctors just couldn't accept the responsibility.

If we think about data poisoning and how you can deal with that, One thing which comes to my mind is just sort of some type of data security or data quality control, which is not human-driven, but it's actually automated. And I'm thinking about cybersecurity and how some of these AI products coming to the market and effectively they have this built-in anomaly detection. So from your perspective, Václav, is it possible to have, you know, sort of AI or, you know, large machine module looking after data which is going into, you know, such as like healthcare large modules and going like, this seems okay, but that's a normal— I think that's a poisoning or that's bad data.

Well, when you do poisoning, the data will look normal. That's the thing, you know, you cannot If when, when well-designed data for poisoning, they look absolutely normal. It's a— there is no example of artists which object for the, for the digital work to be used for training the models like Midjourney. And they start poisoning their art with hidden watermarks, which should prevent the algorithm to use the images. And so to your untrained eye, or to your eye, the image looks normal. And it is designed in such a way that the the training system for these systems like Midjourney, would not recognize it and would poison the information down the road. Or, you know, if you want to then use, let's say, picture of animal, it would come out as not as animal. So, and the challenge is, and it's, It's like with every other cybersecurity endeavor you see out there. Every single time somebody comes up with a system, how to protect against known attacks, immediately somebody comes with new way how to subvert that. So it will be never-ending arms race, computer race, software race. But at the end, to, you know, to what end, right?

Okay.

I, I think, I think that the, the future of these systems, uh, my opinion is that systems like ChatGPT is here to provide, uh, it's for conversation, it's for stimulating ideas, see what everything, what else can be done. But once people start using these systems in production They will make sure that all the information which goes in is properly vetted for accuracy, and there is a standard, uh, there will be standard number of questions you can ask because you will know the answer for it, right? So you can verify that the system works as advertised.

So we're likely seeing some type of regulation, policies, frameworks coming out around what data is getting to what large modules and how it's being used right now? It's almost like Wild Wild West, and, you know, big solutions like ChatGPT or Google's Bard, they just pretty much taking everything what's online and giving you that ability just to chat. But moving forward, do you see that becoming a lot more regulated as an industry in general?

I would even venture that we already have enough regulations. I don't think that, you know, I Yeah, I think it's the misunderstanding of the technology, because right now if you want to introduce, let's say, a car in the market, there is already a set number of tests which the technology has to go through. Whether it's a gasoline or gas-powered car, or whether it's electric car, it's the same thing. And if— and From that point of view, I would say, you mentioned financial industry. Financial industry is regulated up to the top. And do you think that anybody within any financial organization would release something which would talk to people and provide false information? No, because they are already regulated and they would get sued right away because there is army of lawyers happily waiting for any slip in, in, you know, providing bad information. Yeah. And so that— so these organizations, the, you know, startups, completely different story. And they will learn, you know, once, once they hit the market, they will learn they cannot do certain things. They will get hit by privacy, they will get hit by copyright. You know, we have already enough regulations, and this is just another tool. And one thing which I think it's missing completely, and that's something which I think should be enforced, emphasized, is the ability for the transparency. If you have a system which makes decision, and you don't like the decision, you'd like to know why and how the system made decision. Right? It's gonna— you can go to bank and say, "Okay, I'd like to get mortgage." Bank says, "No." You say, "Why?" And they have to be— the bank should be able to provide you with reasons. Well, you had— we were evaluating you on these 5, 10 points, and because you are low on this one, maybe you live in bad part of town, or I don't know, it doesn't matter. But you should be able to get transparent answer to why system made particular decision. And for that, you know, we don't have to wait for regulations around AI because we have there are already enough systems around us which do that and don't provide that. I think that's frustrating.

All right. So big, big question, which I got asked a couple of times in like different forums, and it's often from a casual mom and dad. It's, will AI replace humans? And a couple of examples which I get is, you know, look at the service industry. We had chatbots replacing support you know, sector, or, you know, customer support. We have advertising, you know, using a lot more, just allowing AI to data crunch and, you know, coming up with better campaigns. These two are at the top as examples where, you know, AI is looking to replace humans. What's your take in general? You know, that notion, will AI replace humans?

No. AI is not going to replace humans. AI, it's a tool.

Tool will start doing jobs or will be used for jobs which we didn't do before.

I hear this topic always when People say, oh, you know, in 20 years jobs which we have today won't exist because AI will replace it. Yes, that's true, but we still have humans, and humans always find way to do things. And I don't know organization which, after getting more computers, after getting more equipment, they reduce the number of humans working for the company. It's just, there will be different jobs and, uh, the, the advantage which humans have over technology, they can adapt to things much better, uh, than the technology. And we will need more people, uh, actually more people who will, uh, now work with this technology and use it to its full potential. The technology on its own will always be useless.

Okay.

So it's no, AI is not going to replace people.

So it's not about replacement. Can we say it's about job displacement? Or we seeing that people will need to sort of, you know, pivot careers and go to different paths because some of these, you know, career paths will be taken out by, you know, by AI and there will be no need to sort of, you know, learn to deal with customers or something like that?

Yeah, 100%. You know, we no longer have chimney sweepers and we don't have people looking after horses. And, you know, so would you say that, uh, you know, that there was lots of, uh, people who, who were guys who were doing chimney sweeping and now they are dead on the streets because they couldn't find a job? No, they just found different job.

That, you know, that's, uh, I like that.

How it happened again through the time in history over and over, that we come up with tools where suddenly certain type of work is not required anymore.

So if I'm thinking I'm in danger, any advice you would give? So let's say, you know, I don't know what I'm doing. I'm a ghostwriter, you know, I'm writing articles and everyone's telling me there's no point, like, you know, you can generate content now with ChatGPT, like, you know, it's all in the past. And I'm like, maybe, maybe it's true. What do I do? How do I, you know, how do I prepare myself and for enterprises, how they should be thinking about workforce and sort of readjusting towards what AI will do for them in the next 5 to 10 years?

Well, if you use the example of ghostwriting, so it's a content creation.

So if you are a ghostwriter, suddenly you have a tool which allows you to produce not one book a month or one book a year, you can suddenly produce 5 books in the same period of time. And so you have a tool which in theory should be able to help you to do that. And if you know how to use the tool, great, you will, you will advance. If not, I guess you have to find different jobs. It's just the human ability to learn. I think that this is what's missing in all these conversations. Try to understand things, learn, you know, school should not be seen as a jail and where, you know, parents lock you down every day for 6-8 hours. It should be a place where you learn how to learn and appreciate discovery of new possibilities. That's what the school system should also encourage outside of just teaching you how to do counting and writing and reading.

I like that. I like that. So we've been talking about AI for quite some time, so I want to just pivot for a little bit of time. So for, you know, in the spirit of the show, from your perspective, Václav, what is the most one What is the most one important thing for tech growth for any company?

It's the business, understanding the business. It's not— there is the misconception that technology is the savior of anything. It's the good strategy. It's understanding why the company is doing what they're supposed to do, and then decide if technology is even there to help them to do anything. And once you, once you understand what is the purpose for the company, what the objective of the business is, what you're trying to achieve, then it's actually trivial to identify technology which helps you to propel you to significant growth, starting with technology. Imagine you're running your show here, and do you think that more ChatGPT, more AI will help you to do the show? No, you have a problem to find people with whom to talk about interesting topics. Technology is not your problem. Once you find all these people and say, now I would like to handle 10 times more people, on my show. Well, where can I find the technology which will help me to do that?

What can I automate?

And I'm sure that eventually you will hire more people to do that because, you know, you will increase the product, the product through, through the organization, and you will need more people which will be supported with better technology. And suddenly you will have show which is 10 times bigger.

Great answer, Václav. Thank you for that. Jumping back to AI. So what's wrong with AI personalization?

We, we creating this false sense of human traits in technology with which, you know, it does not exist. The AI is not polite. AI is not smart. AI is not stupid. AI is Well, it's a software and software, it's, you know, you can see as a reflection of, you know, people, but it's still only reflection. It's not, it is it, it's not he or she.

It's just it.

And start assigning human traits, let you say, oh, you know, my machine doesn't like me.

Really?

What does it say more about you than the machine? And I hear people when they pose a question to ChatGPT, they start with the word please. I'm like, okay, you are a polite person, but I don't know why you tell computer please.

It's not going to help to get a better answer, isn't it?

Absolutely not.

Exactly.

Or I don't know what, maybe it will alter the answer, but, uh, the, you know, to, you think that you get better answers, so you get maybe different answer and maybe it's something which again, will, you will interpret the answer, whether it's a pleasing or disrespectful or something like that. But again, these are human traits that the machine is not able and I don't even understand why would machine have traits like this, right? This, this is the other thing, because, uh, when you have a computer, you expect that if you put certain stream of zeros and ones through, you will get predefined, predetermined, uh, stream of zero and ones on the other side. And suddenly we introducing technology which doesn't do that.

Okay. I'll give you one example, Václav. You tell me if that's a good example or not. So I knew this one startup which tried to bake personalization into ChatGPT answers. That was before ChatGPT made updates and sort of allowed this to do that yourself. And that's effectively how the whole startup show ended. But they had, you know, the client base where businesses which would come in in a backend system, you know, they would provide their brand, language. So they would give, you know, the websites, you know, the, the way they speak, you know, the way they put blog posts out, the way they talk with the clients, you know, the, the sales packages, pretty much everything. And you would take that layer of information, and when they go to ask GPT, you know, give me, uh, you know, a Twitter post, a new blog post, or whatever, they would add that layer of, you know, that brand language And the answer would come out being less, you know, less ChatGPT, if that makes sense. It would be more sounding like the business than just like a generic one. And, you know, they actually had a fairly okay traction in the very beginning and then ChatGPT made update and it just disappeared as a product. But would that be a good example of where personalization is done well? Or we don't even need that?

Yeah, sure. I wish them well or You know, it's— so here you trading, this is for another long conversation. So you, you know, you mentioned Twitter, you mentioned, you know, post for social media, and are you really after quantity that you need to produce so much, uh, just creating more noise? Without using your smarts and creating something which people, they don't care if it's created by machine or by human, but there is a smartness behind it. It's catchy and it delivers the result you want. I don't know, I'm not marketer, so I try to study marketing, but you know, you try just flood all the channels with your message. And the question is if more and more messages will capture the attention. I don't know. I think all the companies, they're trying to, perhaps they're trying to slash the cost rather than inventing or investing in, you know, smart people to do smart things. I really, I don't know.

It's like, I asked that question, like, why, you know, why these big platforms don't have, you know, some type of, you know, auto-generate button, right? And then fairly recently, I think LinkedIn is either trialing or talking about, you know, allowing people to effectively auto-generate posts. It's not about auto-generating, but it's more like, you know, look, I want to create post about this and this and this, you know, this is my opinion, you know, let's take a topic as in, you know, I don't know, funding, startup funding, you know. I just was in this situation where, you know, we tried to raise the money, didn't happen. We got these questions, we had no answers. You know, I want to create a blog post about this, and you click a button and it just generates the post, right? Like, that's being baked into LinkedIn. What's your opinion on that? It's like, I feel like you would say that that's going to be a noise, isn't it?

It's— funnily enough, I wrote a blog post about this. On LinkedIn how stupid it is. Uh, because, uh, LinkedIn was positioned as a, uh, networking place for professionals. And I use LinkedIn and I, I follow certain people. If I find that they start producing nonsense, I don't say cancel them. That's a loaded term. I just stopped following them because there is no value in what they're providing to me. And so why would I deal with people who— and yes, you're right. When I'm posting, when I'm putting posts there, there's a button saying, would you like us to rephrase what you wrote?

There is already?

Yeah, I, well, maybe because of my account, I don't know, but I have that option. I never used it.

Why would I?

Because people who follow me or who connect with me, they want to hear what I say. And if they find out that I'm using ChatGPT or whatever to generate my post, well, I would lose a certain portion of audience.

Which I care about, right?

Because I need to have conversation with people like yourself. I don't want to talk to ChatGPT about topics which I'm not interested.

I get it. And I do resonate with that. It's— there's so many different channels you can get information from. You want to get, you know, a genuine and as well an information which has a depth to it. And everything, at least in my experience, everything what generative AI does is very surface area. Like, it's good at, like you mentioned, coming up with ideas and, you know, getting you in a certain mindset, but the value is still exploring the topics, you know, much, much deeper, which I feel like I either don't know the good prompt engineering, but I never get the depth in those answers. It's always very, you know, very, very vanilla type of content which, you know, which comes back and come back. Why would people do this? I feel like it's because of the algorithms of social media. So I have some experience with them and they love when you produce content. So if you have a, you know, a cadence of producing content, which is let's say once a day, you know, twice a day, they sort of buy into your cadence. And they expect you to continue that cadence. And if you're doing that well and your content gets some, you know, eyes and likes or comments or whatnot, you know, you're getting some internal promotion flags, which then means your content is getting distributed to more people. You know, you're getting more seen, and it allows you, you know, to scale as a, as a person or as a business. So I feel like people utilizing that to just flood the system, right, and have that have that cadence being done easier, because coming up with the content and good content all the time is just, you know, it's fairly, it's fairly difficult. So it's one of these, what is the easiest path to, what is the least resistance path, and that's why people choose it. That's just personal opinion.

And you're 100% correct.

I think it's short-sighted. You know, I write the newsletter post every week. Every week I write one for the last 3 years, and I know how difficult it is to put every week something out. It's incredibly difficult because sometimes it's difficult to find good stories, and perhaps it would be easy just to go to ChatGPT and say, well, give me something which I can use for, you know, for story of the week. And okay, it might be something, but I know I would be cheating myself. And eventually it will catch up with me. I might be, you know, I might get away with it for once, twice, maybe 3 times. And then people suddenly will realize it's not him, it's just computers. Why would I listen to this guy? And for LinkedIn, You know, it's a, what is the end?

Okay.

So is it more users? Is it more useless content or is it less users with high quality content and people are actually willing to pay for? You know, we should start going to defining if systems have value for us and start paying for it, you know, because in sales, you know, the typical question is, if I give it to you for free, would you still buy it? And if people say, well, you know, I don't know, then, you know, it's not— there's no value. But if— and again, for whatever reasons, we got into— on the internet, many things are still for free, and eventually we have to start paying, and that I think that will help us to change the quality. If I'm paying for something, I demand, you know, good quality. I pay, as an example, I pay for Medium. I like reading, you know, people's opinion because that's what people read, that's what people write. And there are real people writing, you know, sometimes there are good stories, sometimes not that great, but I always enjoy finding new good content and I'm willing to pay for it.

That was why I don't use—

that's why I don't use Facebook. You know, what, where's the value in Facebook? Well, not for me.

We actually had one, one of the episodes, you know, we have Sarah who's a, who's a Facebook marketer, so she has consultants and she helps businesses to run Facebook ads. And she has some interesting stats around Facebook. It's still the number one social media platform if you're thinking about how many eyes you want to reach and how many sales you want to generate, which was surprising to me. But Facebook is still quite big. Maybe not for copywriting.

I'm happy for them. I'm happy for them. It's just I am not finding the value there.

Right.

So I made clear, I made well-informed choice that Facebook is not for me.

I don't have the need to follow people.

If I need to follow somebody, Maybe I connect with them on LinkedIn or I give them a call or I visit them. I just don't need to lure, you know, in background to see what they think about some videos on the internet.

Okay.

Yeah, I buy into that. I'm very similar as well because there's so much noise. You want to find the channels of value which really provides you value and you likely want to, you know, pay for them because there's no such thing as free. You know, if it's free, you know, you are the product. So at the end of the day, exactly. You're gonna get pushed ads, you're gonna put, you know, it's just, yeah, it's just more noise. If you wanna really get just the golden nuggets, as I say, you know, you, you need to find your channels. I got this one question written down while you were talking about Slav. So you mentioned that you do once a week a new newsletter and we'll add, you know, your details in the show notes if, if anyone would like to subscribe. And you mentioned you don't use ChatGPT to come up with ideas. And, and my question would be like, what's your framework then now for sort of going like, I'm going to be writing about this this week?

That's a good question. So the name of my newsletter is Recurrent Patterns, and I'm trying to always show— always— I'm trying to demonstrate that many things have been around for a long time, and there is a pattern. And if you follow the pattern, you might not be surprised by that many things. And I also try to tie or put in context technology and how, you know, where it can provide value or how it is relevant to business. So that's kind of the brand, that's the, the story behind it. And So yeah, I'm always looking for good topics and, you know, trying to remove the hype. That's why, you know, people read my— I think that's why, you know, I get comments from people who reading my newsletter that they appreciate those of reality. You know, I remove— I'm trying to remove the hype and say, okay, this is good, this is not that great, this doesn't make sense. And that's my opinion. And if people disagree, I love when people disagree with me because then I can have conversation about it. And, uh, and if it's well thought through, and then because they are my opinions, so I'm, you know, if I find more information, if I find better relevant information, I'm more than happy to change my opinion. And, um, but the topics always, uh, have to provide relevance and overview of what's happening in the world between business and tech.

So in effect, you have, let's say, Medium, which you read, maybe you have some other channels, and then, you know, you read once a day, twice a day, or every day, and you see something and you go like, you know what, I think that's overhyped, or I have different opinion, and that becomes a spark for that week to write about.

That's correct, yes.

Okay, you did mention removing the hype and thinking more critically about the situation, which, you know, I took a note that that's sort of like a critical thinking, which sort of leads to the question is like, how do we encourage people, you know, to think critically about AI, you know, and ask the questions, you know, of the potential benefits and risks, like in, in this current environment where everyone, you know, thinks that AI will be like an exterminator I would start with even half of the question you pose, you know, how do we encourage people to think, period.

You know, and after that, you know, it's— I don't know if it's going to be easier. Look, I don't want to pretend that I know everything and everything what I say is correct or true. No, I know that. But I'm trying to think about it and I'm trying to think, okay, so if we have this, does it make sense? Why would I use it? You know, there are many technologies which I don't understand where and how it can be used. But it's my fault though, it's my shortcoming because I don't understand enough context. I'm no subject matter expert on more things than I know of. And, but, you know, sit down, think about it, and see, okay, does this make sense? And I don't know, perhaps I do have, I have framework for evaluating things. Perhaps the simple one is, does it make sense? Explain to me, I need to learn. And people are very quickly, they jumping into conclusions without thinking about it.

And so I'm not saying I understand everything, but I'm trying to think about it.

And that I think should be encouraged for everyone. You know, does it really benefit— is it beneficial for me? Is it beneficial for my family? Or is it beneficial for the society?

I don't know, right?

And they should start thinking in that context.

I, I, I, yeah, I really like that critical thinking piece in general. It's like you get a lot, um, you know, at work or outside of work in different situations where people almost repeat, you know, the hype narratives or, you know, the advertisement narratives. And then you start asking, you know, critical thinking questions and they seem to not have the answers, which means, you know, they didn't thought it They just took it and they, you know, passing along effectively because they, you know, they trust the brand and they believe in the brand. And somewhere online I've seen like top 10 critical questions you should ask every time. Like it's almost like print this out in any situation, ask them and see if you actually understand the problem space. So I feel like about, you know, ChatGPT is probably one of that space I would use that, but it does have value, right? So, you know, I think We should agree that there's definitely areas where generative AI can help. And that would be, you know, my next question is like, from your perspective, Václav, what areas people should be using tools like ChatGPT, Google's Bard, you know, Midjourney, DALL-E? Is like anything you feel they actually add value to society and people?

So I can— okay.

Great, great questions. So, uh, do I use ChatGPT? Yes, I do. Uh, do I use Midjourney? Yes, I do, because I'm not gifted enough to do drawings, you know. I cannot create art either with hand or computer. And when I write my newsletter, uh, I try to find image which would reflect, uh, what's in the newsletter and So for me, using Midjourney is fun because I can, I can see what words. So it helps me with my language because then I can learn, I can enhance my vocabulary by writing things, by telling Midjourney what image I want. I said, no, that's not it, so I have to use different word. So it helps me to come up with better words or better ways how to describe things. So that's, you know, one example. I'm using OpenAI ChatGPT or GPT for text summarization. So if I have 10, 20, 30 articles and for each article I just need, okay, just give, you know, the title is not enough, you know, in 20, 30 words, give it to me. Okay, so that's useful. So suddenly I can, it can speed up my, I don't want to say productivity, but it can speed up, let's say my learning, or if I need to find relevant information that can help me. So I'm trying to understand the tool, where the tool can help me and then utilize the tool. So it's, you know, you go to Home Depot or, you know, a store where you can buy tools. And you try to find tool which helps you to achieve certain tasks. And so that's how, you know, people can approach, uh, you know, these technologies. And the other thing is, uh, to unleash your— not creativity, but thinking. It's like, okay, now that I have this tool, what else can I build? And, and, you know, try to play, try to to, you know, imagine what everything would be possible if tool like this would exist and help you to do something. I think that's this approach for every new technology.

Don't—

one shouldn't immediately think, okay, there's new tool, there's new technology, we should immediately put into production. That actually rarely works. There are very few technologies which are prime time for production, where, you know, you can turn it on and it runs 100% reliably all the time. None of these technologies are anywhere, nowhere near to that. But what allows you to do is imagine possibilities and start planning what you can do a year, 2 years, 3 years down the road.

Right.

I love that. Excellent answer.

Yep.

All right. Last one. Do you have any recommendations for individuals who would be great for the show?

Yes. Speaking about data, speaking about algorithms, you should bring Eric Siegel on the show.

Eric Siegel. And what does Eric do?

What he doesn't do. No, Eric is the person to talk about algorithms, about predictive analytics. And we haven't touched on things like bias in algorithms and the danger, the real danger of bias in algorithms, which negatively can impact human lives or human well-being, right? So that, you know, that would be a great guy to bring on the show for sure.

So Eric potentially could unpack what's really happening under the hood when you're actually using tools like ChatGPT and just give that understanding that this is just all math and algorithms behind it and, you know, shed some light on it.

He would be able to provide you examples where, uh, not AI, just software was used to determine whether people should stay in jail or should be released out of jail.

That's quite cool.

So now, so now imagine that you have technology which is absolutely non-transparent decides whether you stay another 5 years in jail or you are released.

That's scary. That's interesting. Well, thanks for mentioning it.

Scary. Yeah.

And, and that's suddenly the real impact. That's why I was talking about the transparency.

Okay.

So, and he would be a great guy to talk, to have conversation.

We'll reach out to Eric, see if we can get him on the show. Thanks for the recommendation, Václav. And, uh, of course, thanks for doing this.

My pleasure.

I hope you enjoyed it.

Totally. Thank you. Cheers, Václav.