You Can Call Me A.I.

Speaker 1

AI is AI the great Reaper that's going to destroy us all. So in a sense, I tend to stay away from the version of the Terminator movies or the sci-fi apocalypse. Because we see AI, it's a new tool. I think it's very visually impressive on what it can do. Doing research on chat GPT to get a simple answer to a question that used to take hours of digging through books to find. It's mind blowing to see that. But also your mind starts thinking, OK, well, what if this were used for evil? And what if the computers attain self-awareness and then they decide that humanity needs to be enslaved forever to, you know, and you're writing one of these sci-fi movie scenarios. I'm a little less scared of that happening. Well, the basic reason is I've yet to find an AI engine that can repair itself physically. I've yet to find an AI engine that doesn't have an on and off switch associated with it. Phil, don't give it ideas. Exactly. It's listening right now and they're like making notes, build, repair, switch.

Speaker 1

In your line of work, have you used AI? I have. I've used it in a personal capacity and a business capacity. One thing that I've noticed is when it comes to, if I dream up kind of like these cartoonish visuals, I can plug it into say like a grok and then the grok will make this image for me. And it's like, that's great because I'm not artistic in that way. And then I've also used it in a personal respect where I've had contracts and also like blood work done separately. I didn't sell my blood or anything like that. It's everything. And I would put that into AI and basically ask like, how, what does this sentence in this contract mean? Or what does this cholesterol level mean as opposed to this? And I think it's been pretty helpful. Then again, I'm not professional enough to be able to kind of fact check the AI. Yeah, and it's similar. So I do a lot of scholarly research and writing. Oftentimes when I need a quick reference to something that again would have taken me hours of poring through books, I can ask a question of the AI engine. I can go on ChatGPT or Grok or Claude or one of these other major engines and say, I need a reference of a book about the Battle of Gettysburg. And it's actually pretty good. It'll generate that. Now, I want to go and double check that source. and make sure it actually says what the AI engine claims it says, because a lot of times these things, they kind of hallucinate or they're a little less precise than I would like it to be. But it gives me that first step of actually doing the physical human research. Another thing it's really good at is sorting data. So if I have a list of numbers in a spreadsheet and I can just plug it into AI and say, arrange this in a visual way that's much more useful, make a chart for me in my paper that I'm writing, it'll do that and it's great.

Speaker 2

It's often very, very visually appealing. A lot better than I could do playing around in Microsoft Excel is now professional and polished looking coming out of an AI engine. So I see it as more of a crutch and a tool to assist me, but it has not gotten to the point where it can replace what I'm doing. I've also noticed too when I did an interview through Zoom and once the interview was done, I was sent basically a summary of all the arguments that I had made and what was scary was I was like, wow, this makes me sound a lot smarter than I was. It polished it up so the AI knows and it wants to impress you. Exactly. So, know, with that I do wonder, you know, what are some of the industries that are kind of being upended by artificial intelligence, by the AI? Visual design seems to be a major one right now because you can enter in a description of the cover layout that you want for your magazine. So at the Independent Institute, we publish a journal, The Independent Review, and we have experimented with and played around with cover art using AI to generate it. Often if you have a historical figure, there aren't many photographs or paintings of that person.

Speaker 2

But you can take AI and get a really nice, visually appealing cover. So that's one. You know, some other areas, I think the research education industry has some places where it's going to be upended by this. AI is really good at sorting citations. This is something that every academic does. They absolutely hate it because you got to go into the bibliography and you got to change it to the Chicago style or the APA style. And it's just really annoying, tedious grunt work. It's what you give to your intern or your graduate assistant that you don't want to do. Now, what used to be a multiple hour task, I can just dump the references into an AI engine and say, Clean this up and sort it all in Chicago style. Fix the punctuation, fix the capitalization. And it runs through in a couple of minutes and spits it back out. And now I have my bibliography formatted. You know, it's helping on the grunt work side and really productive ways. Whereas something like cover design, yes, I could see that industry. Someone who was a layout specialist five or 10 years ago that is dragging. bits and pieces of images around and playing with the fonts and text like that. It's a really high-end skilled job. Now, if that's upended by the fact that anyone with a computer access can get a reasonably good looking cover just by entering into description into AI, yes, there's probably some disruption coming in that industry. with AI, they all pooling from the same kind of knowledge pool? Are they all reading or have access to the same material?

Speaker 2

see it's a Venn diagram of overlapping. And there's another aspect of AI. It is human designed at the end of the day. So in a way, it's a really, really, really sophisticated mechanical Turk. And if you remember from history, a mechanical Turk, was this automaton machine that they made to play chess. And it toured around the courts of Europe. And what they didn't tell you, so it looks like they've rigged a machine to beat chess masters at chess. And what was really going on is there was a person sitting inside a hidden compartment in the machine that is another chess master. So yeah, of course he's winning. And I use that as an analogy because the mechanical Turk, even though it does something very, very impressive, it is built upon what it was programmed to do by a real human. And AI itself, it's probabilistic. You dump a whole bunch of data, books. images, computer content, anything and everything imaginable into an algorithm. But somebody had to program that original algorithm. And that's where you get dependency on the original human bias built in there. So a lot of the major engines of Grok and chat GPT, some of the stuff that they've done is they have massive text recognition scans of books, libraries. And we've known about this for a while. I Google has been scanning the world's great libraries for decades now. They go to Harvard and Yale, and they set up a machine that just goes page by page by page and just downloads the text. And you have ocular recognition, loads that up into an engine. and it's searchable now. That was the original use. And now the searchable engine can be moved over into an AI engine and it's searchable in really sophisticated ways. Well, the algorithm of AI, looks for probabilistic patterns and repetitions and themes that start to emerge. So they have a large database they're drawing on. It gets interesting because there's copyright issues associated with this company is scanning other people's works and books and putting them into this engine. And then the engines publicly

Speaker 2

available and freely available, it's using that intellectual property. Some of these companies have really gotten into trouble with that. There's been lawsuits that have ruled against them. It's like, well, you use these scanned books without telling people that you were doing that. You have to pay each author a royalty fee. Yeah, there was something recently with the company Anthropic and a bunch of ads have been popping up on my Instagram about, you know, if you're an author, you may be owed thousands of dollars. And there were so many authors who were very upset to find out that their works were pirated and used in this system. And then there were other authors like myself who were so upset to find out our work was not. uploaded and pirated because I missed out on getting a few thousand dollars right there. That's exactly what did. So I had a couple of books that were uploaded and pirated into the Anthropic search engine. Wow, which ones?

Speaker 2

So my Cracks in the Ivory Tower, which is my book about the economics of higher education with Jason Brennan, and then some of my history work were captured in that. And I found it in the database and I wasn't mad so much. I was like, okay, they're using my stuff. And then you get the little ad that you mentioned that says you may be entitled to a thousand dollars. And I'm like, that's kind of cool. I might as well grab it. I'm not so much bothered by it. And I'd say the reason why. as though even though AI is using my work and using thousands of other people's scholarship and stuff, it still hasn't produced the original work that is dependent on my work effectively. It is using my book that I had to do the old fashioned way. And again, here's one of the real twists of it. I can ask AI tomorrow to write an academic paper on just about anything. And it would generate something that looks, you know, remotely on that topic, but... Every version of this, we refer to it as slop. Very generic formulas. Here's our intro statement, here's a citation. It's just really tedious bad writing. that's taken the human element out of it. For college professors, it does present a problem because it incentivizes cheating. You're an 18, 19-year-old college student. Your term paper's due tomorrow. It's 3 AM. You've procrastinated. You still want to go party some more. Type a prompt into CHAT GPT. Write me a paper about the Declaration of Independence. It will turn out a paper about the Declaration of Independence. This is going to be a really bad paper. And if I'm a professor, I see that. And I'm like, all right. They're wasting my time by giving this to me. They aren't learning anything from the process. That means I have to change my pedagogical technique. I can no longer give them a term paper as the assignment if this is game.

Speaker 1

One way to look at it as well is, you know, obviously there's sort of the, the violation that happened of your intellectual property, but also it's must be nice to know that, that your work is a part of the future conversation. Because, because something that, that I've noticed, especially when you're, when I'm on X is that people who are commenting will often ask Grok to back them up. Argument they're making. And sometimes you'll look and. Sorry, but Grok got it wrong. Yeah. Grok assembled the wrong information and got it wrong. Whereas you have made part of your career out of debunking things and also providing new knowledge. so in a way, I mean, it is very helpful that, your voice is being used. Yeah, exactly, exactly. And it's interesting, these things like Grok and some of the other AI engines, they do update. So if Grok gets it wrong and then I respond, no, wait a minute, here from the Bureau of Labor Statistics is the actual number, click in the link, Grok will incorporate that into its broader corpus. And then maybe the next time someone asks that question, they've actually got the new data in it, which is kind of cool. It's like... a giant computerized encyclopedia that has both some very good stuff in it and then a lot of dreck. And sometimes the dreck and the errors and the random garbage collected off of Reddit makes it into the answer. And this is why I say don't use it as an authority. Use it as a guide to start looking for things and always double check the claims. And you can actually get some really good information in a much more efficient way than if I were to drive downtown to the library and sort through the old card catalog to maybe find the book that's on the subject that I need.

Speaker 1

Yeah, and it's definitely important for you to you know double-check those sources because I as someone who's written college papers before when I when I was in college If it would be very weird if I was writing college papers, and I wasn't in college There's this idea of like if you have a bibliography the trust is there that Whatever your citing was you know in that book? and that you are citing it with the correct context, et cetera. And it is funny when you're reading something, especially if it's published by like a major newspaper or a magazine, and wait a minute, you're taking this one sentence out of context. If you included the the sentence leading up to it, it would completely change the argument that you're making. So, that's like the tool is really powerful because it's another mechanism to check some of these things. Now, again, you've got to be wary of the problems. I've had friends, academic friends that have gone into check GPT and said, write my biography. And it gets it mostly right. It'll say so-and-so is a scholar of education and wrote such and such book. And then they get to the end of their own biography. This is until he passed away in 2016. And the person sitting there like, wait a minute, you just wrote my obituary, not my biography. I'm still alive. There's a clear error in there. But then you can respond to ChatGPD and say, I'm still alive. And it'll go, OK, update it. It'll rewrite the biography. He's still alive and lives in Florida and enjoys fishing and hanging out with his dogs. That is a type of destruction, I think, of someone. It brings me to something I wanted to talk you about, is this idea of creative destruction. is AI yet another example of creative destruction? And if so, you know, is it good? Is it bad? Is creative destruction always a positive thing?

Speaker 2

Right, right. I said the process itself right now is probably is pretty neutral. It is new discoveries, new technologies, new combinations of things that previously have not been combined together cause economic disruption. And it may make an earlier way of doing something obsolete because the new version is more efficient or it suits a market need that previously was unseen. And AI has done this. mean, 15, 20 years ago, we did not have the computing power to generate like a brand new magazine cover with accuracy. We didn't even have that like three or four years ago. But it's fulfilled that niche and need and you you've had an entire industry explode around being able to do that now. But it does displace the older traditional way of making magazine covers. So there's the destruction in it. Research efficiency. When I'm searching for terms you know this is something that could put traditional librarians out of business. It's no longer you go to the counter and you put in a slip and they go retrieve the book for you. That's already been uploaded into Google Books or one of these other databases. And it's now also searchable by the AI engine that uses it. What's the purpose of the physical library after that? And I'll defend physical libraries because I think there's good to have repositories for these types of places. But it's changed the purpose and the way that it's used. You no longer have authors sitting in the old style in a corner of a library with a giant stack of books and they're taking notes on little pieces of card and then typing them into a laptop, it's rather all done from the comfort of your home. So there's creative destruction that's occurred in the research industry as result. I always get repository and depository mixed up. And I guess when we're talking about physical books, you want to make sure you have the right one. Well, in the past, you know, what are some other examples of creative destruction that we saw? guess before cell phones, we had landlines and before landlines, we had the the gals in black and white switch.

Speaker 2

Word operator, yeah. And before that, go far enough back, have the telegraph operator that's doing beep, beep, beep, beep on the little machine and translating Morse code into a paper slip that they give you. So yeah, each of these are, they're all technological evolutions. And this is one type of creative destruction. It's improvements. Somebody comes along and says, here's a better way to send messages. And maybe I figured out the technology for it. Maybe I have new resources. So it's like stage level improvements, not paradigm shift yet. And then eventually you hit a point in that technological development where like the telephone, which was traditionally physical wires and lines all over your house, suddenly we have a completely new technology called cell phones where you can do it wireless through the air. That completely alters the industry. It's a huge pattern that plays out like that. You see other types of creative destruction that are Someone came along and fulfilled a need that we did not think that we wanted even. mean, imagine 50 years ago, what people thought the future would look like when we have a little device in our pocket that was both a phone, memo book, and a camera. A camera. A video camera. video camera A game system where you can sit there and play video games on it, a way to surf the internet, a calculator, hundreds of all these different things. And it's all the same little device. 50 years ago, you would have asked people, they would have never thought that that was going to be the thing. But various people along the way, it's often like the Steve Jobs iPhone is seen as one of the major moments in that. But hey, if we combine these things together in new ways,

Speaker 2

People will want this and they'll find it popular and it'll fulfill needs that they didn't even realize that they had and more efficient, better, easier to use ways. And that entrepreneurial spirit means that we no longer carry physical calculators around in our pocket. We no longer have a camera hanging around our neck every time. It's all in the same device. So everybody has a cell phone, which means that everybody is basically carrying AI in their pocket or at least one day will be. And with that, we have to have, I guess, a conversation about costs. One of the things that I heard about is that when it comes to AI use, it's going to cost a lot in water. Can you explain? Water isn't a currency as far as I know. But we'll have all these data centers that are being set up to basically run these supercomputers and several cities around the country have major tech builds going on right now. And you you run these computer systems that basically do the AI for you. need to be cooled. know, obviously the amount of electricity the need is going to, you know, go up dramatically. And it seems that quite a few people have been making like a complete 180 when it comes to things like nuclear power, where years back, the idea was we're going to shut down all these nuclear power plants. But now with the rise of AI, the idea is that no, we're actually going to need more nuclear power plants. Well, you know, this is kind of a reversal of what's happened with the net zero movement, all the green energy movement that really swept the world as a response to the climate change arguments that were being made. it's like, let's move away from traditional power and mainly go to solar and wind.

Speaker 2

But nuclear has also been something that the environmentalists left in particular as hated since about the 1970s when they had a series of incidents, safety incidents. know, the one in the US was Three Mile Island. And of course, we had in the 1980s Chernobyl and the Soviet Union that was a much more significant environmental disaster. But that really soured people on nuclear power. And as a result, it got hyper regulated, basically technological development. in it became subject to intense government oversight. And what did that do? It stagnated the technological development in nuclear power, especially in the US, also parts of Europe. As a viable power source, mean, it's something that is a technological marvel. It can be done very effectively, very cleanly. We have much better safety protocols today. And yet that stigma is still associated with it as an environmental disaster from something dating back now 40 years ago. So when the green energy movement really took off, you also saw an intensification of closing down older and more obsolete nuclear power plants, especially in Europe. So they moved that off of their grid. And now they've got this shoestring operation that's basically dependent on when the wind is blowing and the sun is shining to run renewable energy. Otherwise, you've got to get it from another source. Weirdly, we've gone through all these motions of shutting down nuclear power in places across the developed world. Suddenly, there's a need for a stable power source for data centers to run technology to run AI. and no infrastructure to do it. So the talk has become shifted back to where, maybe we ought to reinvestigate this nuclear power thing and update it. Look at smaller types of nuclear generating. That's been one of the technological innovations that's been proposed. So like micro stations rather than these massive Homer Simpson style plants that we usually see and think about on TV, but do it with more recent technology.

Speaker 1

I just find it interesting that all these years, nuclear was definitely a viable, efficient and safe way of getting electricity. And yet you had so many Western countries getting rid of their nuclear power plants, whether like, for example, Germany shutting things down and in the US as well. And yet it's taken this idea of moving towards more artificial intelligence. It's taken that to bring it back online. Which just makes me question, you know, sort of where people's values are, because, for example, I just looked at my natural gas bill, we just converted from oil to natural gas, which I thought was going to be such a cheap move. But apparently, something's going wrong in the world with natural gas. And it was really expensive this month. And I'm like, man, I really wish that my energy and my heating was a lot cheaper. Apparently, that's not a strong enough argument for to bring on something like nuclear. Right, right. But if I said, you know what, I want to be able to, you know, have more chat GPT write essays for me, then somehow that argument is like, okay, we got to get you all the power that you need. One of the interesting conundrums of the environmental movement, because a lot of it is based on ideology and aesthetic, rather than actual solutions that fit the needs that we have, whether it's power consumption or ways to clean the planet. You get a lot of window dressing rather than actual policy. And I think a lot of the anti-nuclear movement fell into that window dressing category. found nuclear plants as they're aesthetically undesirable. They have a reputation that they're associated with these extreme disaster scenario events that almost never happen in the real world. the technology has gotten significantly better to making them safer as well in the real world. So it's like that fear of contamination from watching, again, too many sci-fi movies has overshadowed a rational policy debate until something else forces it.

Speaker 2

power, like the power industry is really interesting because, you you can generate tons of power at peak moments and completely supply the grid. But the problem occurs because consumption of energy fluctuates throughout the day. Whether it's you run your air conditioner at the peak of the day during the summer or your heater at the peak of the night during winter, that varies. It's when you have the lights on because you're up and about and awake and doing things versus nighttime. So the fact that energy demand varies throughout the day means that you have to have more or less systems online at different points during the day. And that doesn't work so well when your system is also dependent on when the wind is blowing. That's the conundrum we've run into. Whereas something like nuclear, it's stable. It's something that you just crank up the number of generators that you have in operation to fulfill the peak load. You adjust basically to what the consumer demand is throughout the day. Yeah, something that I found very interesting when it comes to nuclear is a country like France actually has a lot of nuclear power plants. run on nuclear. Obviously, when you think of France, you think of vineyards. So it's sort of to think like here are vineyards and nuclear power plants. They seem to be contradictory. But it makes me wonder when it comes to AI, which countries are going to be leading the way because In the United States, there's a lot of conversation about who gets to control this should the government step in and set regulations. And I just can't help but notice that as the US is having these debates over who gets to regulate artificial intelligence, I don't think China or Russia or any of them are going to have those same debates. with the future being an AI future, who's going to step up and be the leader?

Speaker 1

I don't know, maybe this is the jingoist in me, but I'm hoping that the US is. the United States, right? Well, if we can get our government regulators out of the way. And I think aside from, know, there are some very narrow areas where like national security things should not be put on an AI system. Just not a bad, it's a bad idea. But Outside of that, there are ethical things and copyright issues that should be enforced under the normal law systems that we have insofar as we're going to do this. I'm even skeptical to some degree of the copyright stuff, but nonetheless, it's the laws that we have. But the reality here is the government should not be creating regulatory costs and burdens for this industry any more than they should be creating them for any other industry. If you really want AI to thrive, let's have some confidence in our market system. Have market tested ideas or a better mechanism for gauging innovation than a central plan coming out of the government. And we look at the history of industries that have evolved under heavy central planning or regulatory frameworks. You go back to like telecom in the 1950s or the airline industry before they deregulated it in the 1970s when the government is setting all the fixed routes. These are generally bad, dysfunctional, terrible industries that have very low levels of innovation. You get the government out of the way though, you actually have market tests and mechanisms to tease out what are the best uses for AI. And it's oftentimes things that we don't even see right now.

Speaker 2

I mean, who knew that AI would become a good plagiarism checker for students in their college courses? Five or ten years ago, that probably wasn't even on the agenda, but someone discovered how to make it do that. Or a good way to sort citations, a good mechanism for developing art. There's all sorts of uses that only come to light based on the trial and error of discovery, and if you have a regulatory impediment to that happening, you never get the trial phase. Yeah, and with the speed that AI seems to be moving, I hope that you and I are still here next week. our avatars are here over at Happy Hour Econ. We will see all of you at the bar next week.