27 | Beyond the Hype: Generative AI's True Potential vs. Bubble Talk for Business Leaders, with Jeffrey Funk, PHD and Globally Renowned Tech Consultant

0:00Hello, and welcome to Leveraging AI. This is Isar Maitis, your host. As you know, I'm a serious enthusiast of AI, and most of my guests, and I would say all of my guests until today, were just like me, meaning very enthusiastic about AI and its potential impact on the world. And when I started seeing some of the content by Jeffrey Funk, who is a technology expert and has been for years, both as a professor in academia, as well as a consultant, he is somewhat skeptical about the immediate impact of AI on the actual economy and the business world. And hence, I thought bringing a different voice, somebody who is more skeptical, who has a different view on where AI is going would be very interesting for me and hopefully for you as well.

0:47In this episode, we're going to explore whether generative AI is another bubble like the metaverse or blockchain, and these are just recent examples, or does it have real merit and actual impact on the business world? And if it does, where are the actual places that it's going to provide business value and efficiencies that can impact businesses and the economy? Like every week at the end of this episode, I will share exciting news from this week, including the biggest news from OpenAI that just introduced ChatGPT Enterprise. So stay tuned for that at the end of the episode. And now let's discuss whether generative AI is a bubble and where can it actually have real impact on businesses.

1:31In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host, Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

2:12Hello and welcome to Leveraging AI, the podcast that shares practical ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Maitisir and I've got an exciting conversation for you today. Our guest today, Jeffrey Funk, is a technology expert. He has been consulting to government bodies and companies on how to implement new technologies in the past many years. I'm not going to say how many because I don't want to date him. If he wants to share, he will share, but he's been doing this on the highest level for a very long time. He's obviously also an expert on AI recently because that's the recent craze, And that's going to be the topic of our conversation.

2:50How much of it as a business person you need to consider as a craze and how much of it actually has substance that you need to pay attention to understand how to leverage and that can really help you grow your business versus, okay, it's another topic just like a few others in the past few years that are going to disappear in the next two years. So, Jeffrey, I'm really excited and humbled to have you on the show. We're doing this. Jeffrey is in Singapore. I'm in an Airbnb on the outskirts of the old city of Jerusalem. So it's a very exciting international conversation. I'm really glad and happy to have you as a guest.

3:25Welcome to Leveraging Air. Thank you. We live in the age of bubbles, from driverless vehicles and IBM Watson to delivery drones to nuclear fusion to solid-state batteries, EV2Ls. There's a long list of technologies that have been hyped. Bubbles have happened, and then they've popped. We've all been watching that over the last few years. And so AI has been part of this. I mentioned two of them, IBM Watson and self-driving vehicles. And now we have another one, right? We have a new one, generative AI. And a lot of people are excited about generative AI. And one of the reasons they're excited about it is because it's the fastest growing technology of all times.

4:08So at least until this week, when threads grew faster than chat GPT. So we need to be careful, right? We need to be careful on the metrics we use to conclude that a technology is important. Clearly that the fastest growing technology of all time, that's really not very useful. We need something better. And so that's what we're going to talk about some today. Yeah, that's fantastic. I think there is, like you said, there's a lot of the metaverse thing, like everybody was talking metaverse, metaverse for two years and insane budget was thrown into this. And I'll be fair, I know a lot of people who are techie like me, who are geeks like me.

4:50And I know, I don't know a single person who actually uses the metaverse regularly. To me, by the way, that's the biggest difference between at least the metaverse and generative AI. I know a lot of people, or if I'll be more specific, almost everybody I know uses Generative AI daily. So let's really talk about what are parameters that we do need to look at as good predictions or good set of parameters that we can look at to say, okay, this is something that's really here to stay. We should figure out a way as a business, as a society, as a community, whatever it is, actually learn how to use this?

5:29So I look for a real application that's being done. So think about personal computers. The first applications for them were spreadsheets and word processing. Think about e-commerce. Well, there were books and music. So there were first applications where people were using them. They were finding benefits to people. And you could see that there was a future to this. And one of the reasons you thought there was a future is because there was also something going on behind the scenes. Some underlying technology was being improved. So in the case of personal computers, you knew that they were going to get better because Moore's law was making these processors faster.

6:07In the case of e-commerce, you had faster bandwidth and those are very well documented. The data was real right there. You could see that these were growing very quickly. So I look for that type of stuff. I look for that type of stuff in every technology. Okay. So if we look today on Generative AI, there are many real applications. Like I said, I, almost everybody I know uses it almost daily right now. What buckets would you put it in as, in your world when you view this? I'm like, okay, this is a real bucket that you really should pay attention to as a real life application of Generative AI. When I said being used, I meant successfully improving productivity.

6:48So think again about word processing software and spreadsheet software. So there was a process by which people prepared documents that involved secretaries and other people. And so this process got faster. The secretary, at first the secretaries were typing up that you'd give them something written and they would type it, but they could now edit it with the word processing software. And then there were people who had to do quantitative analysis and they use spreadsheets. So there was a process that was being improved. You could see the same thing with e-commerce because there was a process. You go shopping, you go to the mall, you look around.

7:22Well, now you don't have to do that. Now you can find it online. In addition, books, the first application were very important because a lot of us, I remember when I was a young man, going to bookstores, looking for books. Oh, you don't have the book. Oh, you'd go around all these bookstores and you'd look and look. Suddenly you didn't have to do that anymore. Suddenly these weird books because a lot of people do. A lot of people like these unusual books. The long tail was one term for it. Same thing with music. There's unusual music. People like to buy unusual music. They don't want to go to 15 music stores before they find the music.

7:57They want to do it much faster. So there has to be some kind of process of being approved. And so my problem with ChatGTP is that a lot of the discussion is about individual productivity. It's about, oh, oh, I can send more emails. There's a process there. What about the guy who has to read your emails? Oh, you're able to send a lot more now? Oh, okay. I wonder about all those people have to read them. So there's a process there. And the same with coding, because you make some code and you don't care that only 1 % of the time it doesn't work. But somebody else cares. Somebody else has to fix it.

8:32So that whole process has to be thought about. You have to think about, are these processes being improved? And maybe they are because a lot of companies aren't going to tell everybody. They're going to hide it. They're going to figure out how to do something. But I have a feeling that a lot of the conversations that we're hearing are mostly about individuals saying, I can do something faster. And it's not about companies saying we can develop new products faster because of ChatGPT. And the reason why I think that is because a lot of companies are not allowing their employees to use this because they know that mistakes occur and that causes problems for other people in the process.

9:08So we want to focus on the process and understand, is the process being improved? I absolutely love what you're saying. And I want to piggyback on two specific things you said. The first is about company implementation or adoption of this technology is very different than an adoption of an individual. And I do this for a living. I help companies implement this in-house and it goes back to, okay, it's a business strategy. It's a new business tool, just like any other business strategy, like just any other business doing history. And there are a lot of moving parts. You cannot just say, oh, I can allow people to do this.

9:43So that's topic number one, which I think is extremely important. Like you got to understand if you want to bring this into your business. And from my perspective, you have to, because if you do figure this out, if you do figure out how to do this, you will be more efficient by a big spread from what you were before. And I don't want to call numbers. I don't know if that's 20 % or 50%, but you're going to be more efficient. Meaning your bottom line is going to be higher. You'll be able to run faster than your competition, cheaper than your competition. Whatever the case may be, you become more competitive and you can make more money or grow market shares.

10:15So there's merit to that. The other side that you mentioned that is very important. There are risks and some companies still do not understand the risk. And they're like, yeah, this is awesome. Guys, go use this. And then stuff breaks, either within the organization or big embarrassments outside because it delivers something that doesn't work. There's multiple examples already like that. So I like both these components. What do you think is the right way to do this? If I'm in business today, first of all, what cues do I need to look for to say, okay, this is ready? You said, okay, maybe writing code is not ready, but should I start working with this in a sandbox?

10:55What are the recommendations that you make to companies on how to A, evaluate and B, figure out what are the right building blocks to actually start using right now? Well, you want to experiment on a small scale, try things and see what happens. But at the same time, you want to think about your processes. You want to think about how can this impact on the process? What are our processes and how do we expect different tasks interact? And as you said, what are the risks? If we get to make a software mistake here, what does that impact? Are people dying or are people just inconvenienced? So in some cases, it's just a small inconvenience, whereas in others, it's a big outcome.

11:34So you have to decide on that. You have to decide what those risks are. But you do definitely want to experiment. You want to try things. Okay. So we're saying step number one is figure out the benefit versus risk for different scenarios in the company. step number two is yes, absolutely where the risk is low or can be contained one way or another, or the benefit is really high and you have a way to mitigate the risk, go and experiment. Can you give me specific examples of companies that you've seen doing this successfully on which areas or how they built these sandboxes to make this happen? Well, in AI, I don't know them.

12:14I'm looking for them, but I don't see it. What I see happening is that a lot of individuals are going ahead. I imagine there are companies going ahead, but as we talked about earlier, I think that they're probably not telling everything. They're not shouting out their success stories, or at least they're not going to shout out details because this is the proprietary advantage. If they figure out how to do something, well, it's going to help a lot. And they may not say things, even though they have some success stories. Yeah, I think one of the examples that I've seen, and I've interviewed the CEO of a rapid software development company, They help mostly startups get their product out the door faster, cheaper.

12:50They're a development house and they're doing a lot of this stuff for their own software. So they're still not using it for their clients, but then when they're developing in-house tools, that's a low risk because it's their tools. Worst case scenario, they can still work the way they worked before, but that allows them to experiment in a sandbox, but on a real project. It's not, okay, go play with this, tell me what you think. it creates a product that the rest of the company can use. So that's just one example of how people can use that. You mentioned proprietary data. Have you seen companies or reading about companies and how they've using their proprietary data to get benefits from this new set of tools?

13:32Yeah. I think that most of the benefits are going to come from training your system on proprietary data, not on public data. A lot of people have talked about this. Most software is not available publicly. It's just not available publicly. It's available. It's in our company for security and a whole bunch of reasons. It's not available publicly. So a lot of these systems that have been trained on publicly available data probably aren't going to work that well. It's probably only certain applications that this generative AI can be used for. We talked earlier about one of the examples that I have been associated with for 40 years, which is product development.

14:17So my earlier years, one of the big issues in the 80s was how to develop products faster and what kind of product development to use. And at that time, Japan was the leader. By the way, I spent a lot of time in Japan. I first went there in 85 and I lived there for about 13 years to finally leave it in 2007. because I thought what they were doing was really great. It turned out that it was only in certain areas they were really great. But anyways, they were able to develop automobiles much faster than other companies. And so there was a number of major studies that documented this. And in those studies, they found that Japanese companies were very good at developing, doing lesson learned analyses where they looked at what they had done well, good or bad.

15:01So they developed all this data on this process. So when I think about proprietary data, I think of somebody has a development process. Somebody has documented the process. Somebody has written about lessons learned and other things about the process. So there's all this data that system could be trained on. There's a lot of these proprietary processes that are in companies, manufacturing processes, product development processes, all kinds of other order fulfillment processes. And I think that companies are going to have to define those processes better, get back to it. We used to call it process re-engineering.

15:34It's a really big deal back 30 years ago. It's kind of been forgotten. I think that too many people don't even know what processes they're in. They're upset by what they do individually and, oh, I can send more emails and, oh, I can do this more faster and that faster. And they don't understand that somebody's got to read those emails or some process that we need to improve. And so I think this is where companies have to spend a lot of time. They're finding their processes and improving those processes, training their systems on these processes. I love that. I'm going to, again, piggyback on two points that I think are critical.

16:05One is to really gain from generative AI. And AI in general, by the way, it's not just generative AI. The key is having a lot of data. And if you have a lot of proprietary data, then obviously you can do stuff that nobody else can do because nobody has access to that data. So companies who will understand how to leverage the data that they've collected through the years. And like I said, it could be customer data, process data, marketing data, sales data, like whatever data you have, if you have documented that, and at least some of it is documented by definition. If you've sent, I don't know, 1 ,500 proposals in the last three years, guess what?

16:45You have data that nobody else has because you've sent proposals. Some were successful, some were not successful. What size of company you sent this to? what products were in them. You have all that data, even if it's not in a proper database, you have the data. So you can take it and put it in a place that a system can now read through and analyze. So one thing is really understanding that your proprietary data is huge. The other is documenting processes. Because at the end of the day, what AI knows how to do really well is to take segments of a process and do them very effectively if you give it the right connections.

17:20But for that, you need to understand what the process is, which components in the process AI can do well. Data analysis, data collection, stuff like that. And then say, okay, this little component I'm going to allow AI to do, put somebody at the end of it to see that it's actually doing the right thing. And then you can maybe do this. And I see you have a comment on that. But I think these are two critical points on implementation of AI in companies. Yeah. Yeah. And the proposal process is also very important. Not just to do it more with higher productivity so you can automate more, it's to do it faster.

17:54A lot of these processes involve doing things faster. So you're trying to understand which tasks are sequential and which tasks you can do concurrently. Because if you just do everything sequentially, it takes a long time. So you want to figure out how to do it concurrently. And so this was one of the most interesting outcomes from the study that compared the Japanese, the Europeans, and the Americans in terms of product development. The Japanese were doing tons of things concurrently that the Americans swore could not be done concurrently, but they figured it out. And that requires companies to really define their process as well.

18:26And it's quite likely that generative AI can help firms do that. If you have good data on those processes that do things, more things concurrently. In your eyes, when, and maybe in what industries, generative AI making significant impact that companies has to pay attention to because in the next, and you can name the number, 18 months, 24 months, three years, there's going to be significant change. The industry that's probably receiving the biggest impact right now is media because there's all this fake news out there and there's fake news sites and things like this. There's an article from the Wall Street Journal yesterday that talks about this, talks about all these news sites that are, it's all fake news.

19:09And there's YouTube videos on how to develop fake news so that you can get Google ads, get Google ad revenue. And so, of course, Google doesn't like that. And a lot of people don't like that. So this kind of bad thing that's going to be occurring. Software, I think, is in terms of a positive impact, it's going to be the first. But I still think that's going to take a long time because it takes a while for companies to work out their processes. And some of them will say they've done it, but it's probably on something not the core part of their business. They're experimenting on something that is not the core part of their business.

19:40And that makes sense, right? Like the example you presented earlier there, it's something that doesn't have the bad impact that looking at adding GeneraVy to another aspect of the business might have. I oftentimes wonder when we'll see an impact on generative AI on the productivity numbers for Israel or America or for anyone. Because people have been talking about this for a long time. So Eric Benjolson said this in his 2012 book with Andrew McAfee that this was going to be a big productivity miracle. And there were people who wrote articles saying we're still waiting for the miracle. And so the 2010s had the lowest, had the lowest improvement of productivity ever in America, ever.

20:26And the last five quarters have been negative. Now, a lot of this is from the COVID, working from home. COVID, yeah, working from home. And now there's been all this change in industries and things. Manufacturing facilities closed down, all kinds of things. But still, I think that a lot of people who talk about these new technologies fail to set themselves in the right context, which is productivity growth has been so slow. Therefore, it's highly unlikely you will see an acceleration. Yeah, we might see some kind of improvement, but even if we see some, if it goes up to 1 % a year, that's pretty good.

21:012 % a year. When I look at all these technologies that we're supposed to diffuse and didn't, you know, some of them I mentioned earlier, and you mentioned the metaverse and crypto and Web3, there's a whole bunch of these. and you look at the numbers, because I've looked at the numbers for these things, they're not very big for these new technologies. Even AI is not the biggest, like$60 billion this year, even though people were projecting that it was supposed to be$15 trillion by 2030. So it's nothing$40 billion. And yet if you look at previous decades and look at technologies like PCs and mobile phones and e-commerce grew much faster, they had much bigger impact much earlier than the latest technologies.

21:44So when we think about what's going to happen to these technologies, we need to keep in mind that, well, things haven't been happening as fast as we thought they would. So probably this one also won't happen as fast. We don't know, right? All this is probability. Everything, it's all about probability. It's all about, we want to make, take actions that have the highest probability of success. It's not that, oh, this isn't going to happen or never going to happen. What actions will have the highest probability? And so you made the comment of we talk about experiments and things like that. So we do experiments so that we're ready, so that if things do happen, if things do seem to be taking off, then we can be there with them.

22:22But we don't want to bet everything on this technology and risk everything. Okay, awesome. So I think what you're saying, and I want to try to sum it up, is that there's, don't go into this because of the craze, because there'd been multiple crazes before that either didn't pan out at all or just pan out much slower than everybody else thought. Number two is figure out the overall company process, meaning where does that fit, which blocks that is replaced or enhanced or whatever, and then experiment. But it's multiple processes. It's multiple processes. There's not one process. Yeah. Of course.

23:00Okay, go ahead. Yeah, we mentioned multiple before, whether it's your sales process, whether it's even simple processes like internal processes of how decisions are made or how the handoff happens between marketing and sales or whatever the case may be. And then figure out how to experiment with your internal data as much as possible, because that will give you benefits that nobody else has to gain some benefit that you didn't have before while still keeping the risk reasonable. Is that a good summary? No, I agree. I agree. I also think another thing that you mentioned that is a big deal that I think a lot of people are projecting wrongly and I've never done that.

23:46I've never drew the line like you have. There's a big difference between personal productivity, the generative AI can do magic to. right now, like even without improving to company productivity, which is something completely different because it's a lot more complex, a lot more people, a lot more processes, and a lot more risk. And so I think what a lot of people are doing today, myself included, is taking the benefits in personal productivity and trying to apply them one-to-one to business productivity, which is not the same thing. Yeah. No, I agree. That's a good thing. So I got to ask you a follow-up question because of what I, now I think I understand what you believe.

24:29Do you see risks to specific jobs, like specific kind of titles, professions, or skills that are at risk in the two-year timeframe that people need to start paying attention to? If you look at employment in different professions, I did this a few years ago, So back about five years ago, I was very optimistic about AI. There was another form of generator AI that people talked about for engineering, where it helped you generate new engineering designs. And so I was very optimistic. So I studied this impact of AI on accounting, law, all of these. And I, first thing I did is I looked at how many graduates were they increasing each year and how many people, how many jobs, how many people got jobs.

25:21and we look at the percentage of it. And there was no problem. There was the number of lawyers getting jobs kept going up. Everyone kept going up. Lawyers, doctors, engineers, accountants. The one that didn't was journalists. So journalists have been very badly impacted by the end of subscription newspapers and things and all this free information on the internet. So they don't get paid so well. So there isn't so many jobs. So a lot of journalists have now become like context creators and content managers and trying to get more likes or views for their articles or for articles in general. So that's one of them that hasn't done very well.