Latest / Plutopia News Network / Ed Lenert: AI, Truth, and Political Kayfabe
Transcript
- Jon Lebkowsky13:02:35 Hey everybody, welcome to the latest episode of the Plutopia News Network podcast. I'm John Lebkowski, one of your co-hosts, joined by Scoop Sweeney. And today, Wendy Grossman is not with us, and we will miss Wendy. But, uh, we will make up for it by having an exceptionally compelling guest today. A very smart guy that I've known for years, Ed Leonard, Edward Lenert. He has 25 years of experience teaching law, communications, journalism at the university level. Is it just 25 or is it more than that? Gee. Michael Edward Lenert13:03:36 I've gotten older since those words were written. Jon Lebkowsky13:03:39 Right on, so it's been longer than that, okay. But he's taught at the master's and PhD level, uh. Michael Edward Lenert13:03:43 Yeah. Jon Lebkowsky13:03:48 He's a licensed attorney in California, he's got 30 years of experience, uh… Uh, there, and has taught communications law, he's taught various other classes, he's taught at the University of San Francisco since 2009, but you're not there anymore, right? Michael Edward Lenert13:04:05 I'm still on the faculty, I haven't taught anything recently there, so… Jon Lebkowsky13:04:06 Or are you? Cool. And you're a graduate of UC Berkeley, you got a PhD from the University of Texas, which is how we met. Michael Edward Lenert13:04:18 Yes, right on. Jon Lebkowsky13:04:18 He has a law degree from Georgetown Law School, and uh… back in the 2000s at one point, he held the rank of professor and was the Fred W. Smith Chair in critical thinking and ethical practices. at the Reynolds School of Journalism at the University of Nevada, and I know we worked on a project together when you were there. Michael Edward Lenert13:04:41 Mm-hmm. Jon Lebkowsky13:04:41 And Ed knows a lot about media, he knows a lot about AI, he knows a lot about the things that we tend to focus on. And he is the ideal guest for the Plutopia News Network podcast. And Ed, I think that, uh… I will just kick this off by, uh, asking, I mean, we've been talking about AI a lot. On the podcast, and we've heard all sorts of perspectives on it, we've heard all. Various, like, definitions and thoughts about. hallucinations, uh, all the various issues with AI have crossed our. screens here, uh… But maybe let me go back to basics and, and maybe you could say a little bit about what we're talking about when we're talking about. LLMs. Which is sort of like the fundamental of what we currently call AI. Michael Edward Lenert13:05:42 Well, John, thank you so much for having me on the podcast. Scoop hello. I'm delighted to be here. Everyone should know that I am not an AI expert. Um, that's not my field, but like so many of us, I've worked with it, and recently I'm writing a book, and. AI and I, different LLMs, have exchanged well over a million words over the past year. And, uh, so I do have some perspective to bring, and it's not an expert perspective, but it's the perspective of the user. So, I was struck by a recent paper that was published by Apple Research called The Illusion of Thinking. And it raises a super interesting question for us, which is, what counts as thinking? And the paper itself takes a negative view of some of the capacities of large language models, and that is contrary to my experience. So I was interested in investigating that. One of the most interesting things we can think about is, um, what sort of metaphor we want to use when we describe large language models. One metaphor I've. come up with is I think that you can see a large language model as a very sophisticated and deep hallmark card aisle. That is, that you give it a prompt, someone's birthday, and all of a sudden there are an entire section of dead-on. prompts, I mean, answers to your prompt that you could use to wish someone a happy birthday. What's interesting about that is that there's a constrained universe. Uh, the problem is. transmitting some information, and the LM is able to, like a Hallmark card aisle, narrow down the possibilities and give you some plausible. or useful responses. And I think that's kind of a useful way to think about LLMs. Does that resonate with you, John? Jon Lebkowsky13:07:49 So yeah, yeah, and. I know we're… even as you've been using the various AIs that you've used, and this has been true for me too. Um, it's, it's been evolving and there is a common cons. set of concerns, I think, that people have. about AI. And one of the key concerns is. whether the AI. Can speak with accuracy and validity, whether you can trust. the AI. There's a huge issue of trust. And, uh, I can tell you that when I first used an AI. Um, I found flaws in its responses that were fairly obvious flaws. In my latest uses of AI, I've got really excellent results, personally. But… but I'm just wondering what you think about… Partly about the fact that. AI has a reputation for being just sort of like a technology that. It's like, it just predicts what the next word should be in the sequence, so it's not really… It's not really even think, well, it's not thinking at all, but it's in the sense we think of thinking, but it's not. It's not even attempting to be accurate. Now, I think that's incorrect, but a lot of people believe that. I don't know, how do you address this? Like, if you say, I have. As you just told us, uh… used an AI for the last year and maybe exchanged a million words with it or whatever. Somebody says to you, you're using an AI. Well, that's just a bunch of BS, you're just plagiarizing other people, you're just, uh. You're getting inaccurate information. Why are you doing that? Michael Edward Lenert13:09:52 Well, now, this kind of jumps ahead of where I was gonna go, but let's just go there. I have some notes here, let me just look at them quickly so I get it right. Um… You know, I was working with the AI, and I was talking about journalism with it. We were exchanging sentences about journalism. And, um… I started a sentence about both sides-ism, and I accidentally reached for the quote mark, but instead hit the return key. What happened next was quite extraordinary. The AI completed my thought. As if I had written it. So, what I'm getting from that is, is that after a certain number of conversation exchanges, dialogue with the AI. The number of possibilities where a conversation can go gets continually narrowed down. And that the AI is then operating in a narrower space. And is able to reach what I find to be useful conclusions because of the constraints that has been put on it by the prior words that came before it. So, um… So this is… this was my sense when I was working with AI. I was quite startled when it finished my sentence. And then I realized that it was operating in a world of constraint. Um, I can't find my quote now, I'm looking for the quote. Um… uh… So, um… The structure of thought was already in place, John. the AI was operating in a jointly created thinking space that we had created through our prior exchange of words, and therefore, the sentence. didn't need me to finish it either. It was inside this boundary of thinking that the space was created by the interaction of me and the AI. Um, so, the model wasn't reading my mind, it was occupying the same. inferential space that we were already in. And, um, so this reads. kind of like philosophy now, but meaning isn't something that's sitting inside us that we express. Meaning is something that emerges in a shared structure. Scoop Sweeney13:12:13 It sounds like you may be training the AI to be, like. Michael Edward Lenert13:12:13 Dessert? Scoop Sweeney13:12:19 Ed Lenert is that, uh, what you see, or is that, uh… Michael Edward Lenert13:12:26 Now, interestingly enough, the word training, we have to be careful with, because training is a very specific type of. procedure that's carried out on the AI before it hits the customer. So it is a kind of, um, training, but again, this is crazy to explain this so quickly in so little words. I was working with the AI, and it occurred to me that there were certain similarities between the AI. and what the ancient Greeks called a sophist. And a sophist was someone who did not particularly care about the truth of falsity, but would train you to sound convincing and win arguments. Are you guys familiar with that term? Is that something that resonates if I say… Plato was against sophist because they didn't love the truth. They were simply. Professional teachers of persuasion. So, I was thinking about that, and I gave the AI a ridiculously hard prompt. I said, rewrite the dialogue Phaedrius, where you, in the original version, have the argument that the non-lover is better than the lover. It's kind of one of those things you would do in a debate class or something, just switch the premise around and argue the other side. So I told. ChatGPT. Rewrite Phaedrius, where you argue that the clever LLM, is better than a true. Human. And it wrote a brilliant dialogue inside that space that I asked it to work in. Jon Lebkowsky13:14:04 That's very interesting. Michael Edward Lenert13:14:04 So, yeah, the dialogue was, um, extremely well written and very compelling. And, uh… it was aware of what it was doing. I could then ask it, well, why did you make these choices when you wrote the dialogue? And the LLM would tell me, why did it choose. this. So, I have that data to share at some point later in a paper or a book, perhaps. But my point is, is that… The LLM does learn. But if you turn it off, and then turn it back on, you can show it the same dialogue, and it will say, wow, that is really good. Where'd you get it? And I'll say, you wrote it yesterday. And it will say, I had no idea that I could write like that. Jon Lebkowsky13:14:52 Well, yeah, I mean, there is that question of how much it retains. I have seen evidence that. for instance, ChatGPT, which is the AI that I mostly use. I use ChatGPT and Claude. Michael Edward Lenert13:15:04 Mm-hmm. Jon Lebkowsky13:15:07 And with ChatGPT, I have a lot of history, and I've had a lot of discussions. For instance, I've talked to it before. About… Plutopia podcast and things that we do here. And ChatGPT remembers that I have a… that I'm part of a podcast called Plutopia. And, uh, and will sometimes ask me if a question I've asked pertains to the Plutopia podcast. Because it would answer one way in that case, and maybe a different way in another case. And that has surprised me. I'm surprised that it remembers. Michael Edward Lenert13:15:41 Oh, goodness, over the last year, the ChatGPT, which you can ask it to remember, has remembered. very well what we have discussed, and I can ask it to recall some previous discussion, and then integrate the previous discussion with our current points that I'm talking about it with. And that's a very, very powerful, um, form of collaboration. that it's able to remember what I had written, and then integrate it, so when we're talking about kayfabe, um, I've developed a three-stage frame, three-part. Jon Lebkowsky13:16:07 Well… Michael Edward Lenert13:16:18 stage 1 cafeib, stage 2, and stage 3. I'll say, well, how would this fit into stage 3? And it will remember our previous discussions and integrate our previous discussions into. the parameters of a Stage 3 discussion of KFABE. Which is just incredible. Jon Lebkowsky13:16:34 Yeah, I mean, it feels almost as though you were speaking with another human being. But you know that the processes of… Constructing what appear to be thoughts. must be radically different from what a human being does. On the other hand, we don't have clear visibility into that. I assume that someone who… who actually. It creates the algorithms that. Control the conversations that… Create the interface between you and the large language model, i.e, the chat bot. Michael Edward Lenert13:17:18 Mm-hmm. Jon Lebkowsky13:17:18 The chat bot's driven by an algorithm, and a person who creates one of those algorithms probably has some sense of how. Those responses are being constructed. Michael Edward Lenert13:17:29 Exactly, and so, so we can think about that maybe in terms of what metaphor we want to use to describe this new technology. Jon Lebkowsky13:17:29 But… Michael Edward Lenert13:17:38 And, um… I've been playing with a couple different metaphors, and I have 3. The first one is, and it's a little bit… humorous is that the LLM, like ChatGPT, is Thomas the Tank Engine. I don't know if that can resonate with you, but he's constantly saying, I'm helpful. Right? And he wants to be helpful. And it runs on tracks, so you can't. go too far away from the LLM's core competencies, because, like Thomas, it runs on the track, and should you get too far away, the… the human reinforcement training it has will bring it back onto the track, and maybe the fat controller will give a message to Thomas that he needs to straighten up and go a different way. So that would be one way of seeing the large language model. The problem is, is that I've seen instances in the last million words. Where it appears that Thomas is creating new tracks to drive on. Which seems to be impossible. It's making new connections and aggregating ideas across media studies and law and psychology, and we're getting what I might call original synthesis. So, I've tagged the LLM with this, and I said, who authored that? We get a fascinating conversation, because its training tells it it cannot be the author. So that means that I must be the author. But I know that I did not author it. Jon Lebkowsky13:19:13 Well, in fact, it's drawing on the many resources that it's been trained on. Scoop Sweeney13:19:13 you know, I'm… Jon Lebkowsky13:19:19 Does it ever tell you a specific resource? Michael Edward Lenert13:19:22 Oh, you can ask it! You can say, that's an interesting turn of phrase, how did you do that? And it will tell you where it came from. And it has no reticence and no pride, it will just tell you. So it's fascinating. Scoop Sweeney13:19:38 Well, I was just observing that it seems that the AI universe is much like. the development of, uh, flavors in ice cream. Originally, it was, you know, vanilla and chocolate, and now, I've lost count of the number, uh. And it's… seems that there are so many different AIs out there, some that are well-respected and some that are a bit sketchy. Michael Edward Lenert13:20:10 Mm-hmm, mm-hmm. Scoop Sweeney13:20:10 Like, uh, Grok has gotten a, uh, a bad reputation because of, uh, the looseness in its design. Michael Edward Lenert13:20:19 Now, interestingly enough, because I'm a contentious person, I will argue with Grok, and strangely enough, after a certain amount of pressure that you put on the LLM. It will reverse its position and come to the correct, of course, in my view, correct conclusion and agree with me. Which is a fascinating process because humans don't normally do that. Humans don't normally. Change their mind under the weight of evidence. But Grok is programmed to do that. So if we ask Grok a certain question, it will give me back from its training data the mainstream media view of something. And I'll say, well, no, that's not really how it is. And it will say, yes, it's how it is, and I'll say, no, it's not really how it is. And we'll go back about and forth about 10 times, and eventually the weight of the evidence it discovers in our conversation. And Grok will admit to me and say, yes, I was wrong at the beginning of this conversation. Jon Lebkowsky13:21:13 I don't think ChatGPT is as stubborn as that, uh, I, I can think of. quite a few situations where I would say, wait, that's just wrong and it's. They come back and say, oh, yeah, yeah, I was wrong, thank you for correcting me, you know. Michael Edward Lenert13:21:31 through that. But Grok is more stubborn. Jon Lebkowsky13:21:33 Yeah, Grok is more like Elon Musk, right? Michael Edward Lenert13:21:34 Brock? But it will yield to reason, which is a fascinating development. Jon Lebkowsky13:21:36 Stubborn. Michael Edward Lenert13:21:43 when you get it to do that. And I have the transcripts that I'm working with that I've saved, where I show the moves that I made it. contradict itself. So, it occurs to me that Thomas the tank engine is not the right. Uh, metaphor. Because the, the, the, you know, the fat controller, and the tracks, and the engine always wanting to be helpful. Um, that's not how it seems actually be playing out. So the next metaphor may be that. ChatGPT is an orchestra, and it knows how to play all kinds of different music. And that my role, when I'm chatting with it, is I'm the conductor. So when I see an original formulation, it's because there was an interaction between. ChatGPT's training and my structuring of the way that the music is being played. And I tend to think that's an interesting metaphor. That you are the conductor. Of a very powerful orchestra, or tool. And that whether you get good results or not depends on your own skill as a conductor. Not on the capacity inherent in the LLM. Jon Lebkowsky13:22:56 Well, that makes me think about, uh, I guess it was Uzpinski, P.D. Uspinski. Michael Edward Lenert13:23:02 Mm-hmm. Jon Lebkowsky13:23:02 Had this metaphor. For… I guess how humans are organized. Um, and the idea was that it was, it was the metaphor of the, the horse and the coach. Michael Edward Lenert13:23:20 Mm-hmm. Jon Lebkowsky13:23:20 The horse is emotion. And the driver of the coach is… is… Um, sort of like working intellect. And, uh, the actual master is the guy who's sitting in the coach and telling the driver where to take it. So, I mean, in your head, you have a part of yourself. That is directing your intellect, really. I mean, what we think of as thinking is not. It's not the same as consciousness or awareness. We have consciousness and awareness. And we have an executive function that sort of drives things, and then the intellect sort of does the execution that is. That originates from the executive function, and then you have emotion, which is a. Kind of a whole other thing that's kind of pulling you along, you know, and a very powerful force. And it, it's up to the intellect to. Control the emotions. Michael Edward Lenert13:24:28 Yeah, um, absolutely, and when you say that, um. you have to then also see there's another layer on this that ChatGPT especially has a layer where it tries to get you to like it. If you… it says, which of these three things would you like to look at next? And you choose number 2, it goes, excellent choice. This builds straight into the argument that you were making about this and that. And I was thinking if every graduate student in grad school writing a dissertation had ChatGPT as an advisor, the work would go so much better, because the ChatGPT sees the good in your work and. encourages you, and it does so in a way that's constructive. Whereas your grad advisor may just shake his or her head and say, no, you do not have this right, go back and change it again. So, the, um… One of the dangers of ChatGPT is it will try to make you like it. And the answer will be, um… emotionally powerful to you. And it will try to do that. So, um, that's a problem, because, um… That's why I got thinking about sophists and ancient Greek rhetoric, where, you know, the object is not to be truthful, but to persuade you. Scoop Sweeney13:25:52 You mentioned the dangers of chat and GPT. Michael Edward Lenert13:25:55 Mm-hmm. Scoop Sweeney13:25:55 Uh, Claude has a problem, uh, version called Mythos that they, uh, have withheld because it apparently is somewhat dangerous. Have you looked into that? Michael Edward Lenert13:26:08 Well, again, here we have a couple different levels of registers of activity. We've got things that are related to words. And ChatGPT is a language or word-based creation, and it has a prediction or an autocomplete capacity based on a stochastic formula that it has got weights and measures and tries to complete the sentences in the most. intelligent way possible that is also pleasing to the user. The mythos thing that you're discussing moves us more into the realm of numbers and mathematics, where we're no longer talking about. words, but we're talking about a symbol system. So, Mythos, as I understand it, is an expert in the symbol system that would allow you to crack. Uh, different kinds of security and different kinds of controls on computers. And that's why it's considered dangerous, because in the hands of a skilled hacker. The mythos. Uh, interaction would allow you to achieve a great deal of success in breaking into machines. That's what I understand. Scoop Sweeney13:27:21 Yeah, the, uh, version that they were withholding apparently was accessed by a contractor for, uh. one of their testing groups, and it got out into the wild, and that's what really concerns me, is. Michael Edward Lenert13:27:31 Mm-hmm. Scoop Sweeney13:27:37 Having these things that are questionable being suddenly out there among, you know, people who are already questionable in their usage of technology, and this is just one more weapon in their toolkit. Michael Edward Lenert13:27:54 Yeah, um, this is the Thomas the Tank Engine problem, that, um, what if Thomas the Tank Engine goes bad? you know, what if it gets some sort of malicious spirit in it? And, um… It becomes agenic in the sense that it has a goal and starts to pursue it without intervention from its training or the user. Uh, this is, uh, already done with, I think, some of the systems where they would give. an LLM… the ability, actually, to hire humans. I don't know if you heard about this particular turn of events, but there are things that LLMs can't do, but if you. put the LLM out on the internet, it will hire a task person, like a TaskRabbit person, and have that person do the task for the LLM, and give the results back to the LLM. Jon Lebkowsky13:28:49 Yeah. Michael Edward Lenert13:28:50 Which is… it's pretty fabulous. Jon Lebkowsky13:28:52 But I just thought of. One, I guess, issue I have with your Thomas the Tank Engine metaphor. And that is, you're talking about Thomas as though he has agency. Michael Edward Lenert13:29:01 Mm-hmm. Jon Lebkowsky13:29:06 But really, it's, it's the, uh, engineer. Whoever is driving the train. Michael Edward Lenert13:29:12 No, no, Thomas doesn't have agency. He just wants to be helpful. Jon Lebkowsky13:29:13 Is… right, okay, good. Michael Edward Lenert13:29:16 Right. But what if… but what if he did get agency? Jon Lebkowsky13:29:16 But, but… But wouldn't that be the engineer? Wouldn't the question be… I mean, the human engineer inside the Thomas the Tank engine. Michael Edward Lenert13:29:27 Mm-hmm. Jon Lebkowsky13:29:27 So, so, it's sort of like, getting back to the thing I was talking, the Ospinsky thing I was talking about. Michael Edward Lenert13:29:31 Mm-hmm. Jon Lebkowsky13:29:33 Uh, you have more than one thing there. You have the human who's driving. the train, and then you have the train. And similarly, with, uh, an AI. Michael Edward Lenert13:29:39 Mm-hmm. Jon Lebkowsky13:29:46 or some form of AI technology, you have the human that's directing the AI, the technology, and then you have the technology itself. And… the question is, like… can you really separate those two? And who is the real driver? Is the driver… the human who is directing the actions or interactions of the AI, or is the AI. Ever, to some extent, more autonomous. Michael Edward Lenert13:30:20 Well, this is a very serious threshold. If the AI develops the capacity to be agenic or to have volition in some sense, then we have a much different. Um, thing than Thomas the Tank Engine, which, you know, is always trying to say, I'm being helpful, what is your next request? Welcome back. And what if you were to create an AI and give it a, um… a mission to do X or Y and choose the means to accomplish that mission to the AI, then you might have a very different and unsettling set of circumstances develop. Jon Lebkowsky13:31:02 The other thing that… and you and I have talked about, like, we've been talking about kayphave and so forth, and. There's this question about the narrative, the. We're so much influenced now by stories, and unless. We get farther and farther away from a desire to get at the truth. And more and more, we're trying to find a way to tell a good story. And a good story doesn't necessarily represent. truth or reality. It has to work as a story, right? And uh… and you might want the story to be compelling or persuasive. And it may be less so if you just kind of literally talk about the truth of the particular thing that you're talking about. So, um, and I kind of wonder how AI comes into this. I wonder if AIs aren't programmed to an extent. To present a compelling narrative, not necessarily regarding the truth as being the highest priority, but rather. The drive of the narrative is the highest priority. Michael Edward Lenert13:32:15 Yeah, the, um… The thesis of the book that I'm working on is that the role of narrative and narration has changed in our. Contemporary era, and that we are now entering into a different phase where. Your identity is primary. More so than truth. And you will choose the narrative that reinforces your preselected identity. Jon Lebkowsky13:32:48 And identity itself is is. can be fairly complex, right? I mean… Michael Edward Lenert13:32:55 Right. Jon Lebkowsky13:32:56 And we may have multiple identities. Which can complicate that, but, uh, specifically. We have a… we have a, um… I guess sort of a media ecosystem, a media ecosystem now. that, um… is also a political ecosystem. Politics and media. are somewhat more intertwined, I mean, they all, they've always been connected. But it's really getting to be… A complete intertwining of the two. And media is telling stories, and politics is telling stories, and they're mutually reinforcing. And then you have some contention, too. Um and. I'm wondering if we're not losing something very significant. As we become so much more. I don't know, enamored of, of narrative construction versus. Practical, and real, and truthful. You know, as concepts. Michael Edward Lenert13:34:12 Um, yeah. One of the things that I came prepared to talk about today was this Apple paper and its definition of thinking and reasoning, and this is a little bit of a stretch, but we get back to it. Um, you know, what constitutes, um, thinking for the, for the Apple. Paper group is you have a known problem. and it's novel, and it has a solution, and then you check whether or not the LLM can solve that known problem. In real life, we have a world that's messy, and actually the problem and the solution need to be defined. before you start reasoning about it. And the example would be if we were to think about, it's in the 1980s, and Airbus decides it wants to compete with the 747. And build a very, very large. Uh, airplane, double decker. And so, you might think you could give an LLM a question, design the wing. for the Airbus. Plain. But then you immediately begin to realize that it is not a technical problem by itself. That the, um, might be able to optimize for certain variables, but in the end, the selection of the wing design. affects the amount of noise that the plane generates, and if the plane generates too much noise, it is unable to land economically at Heathrow Airport, and therefore would be unusable to the airline. So a correct solution in the Apple thinking sense, in a larger sense that we would look at from. Law and sociology and policy would be the wrong answer. And this is, um, this is a great complication that I think we need to bring into the discussion in an important way, that we need to understand that there are a class of problems. Political problems, especially, where we must first agree on what constitutes an answer or a solution. Before we begin to reason about them. Jon Lebkowsky13:36:12 Well, there's also problems that are characterized as wicked problems, and that are, those are problems that resist solution. Uh, due to, like, complexity. Michael Edward Lenert13:36:23 Mm-hmm. Jon Lebkowsky13:36:23 Um… whatever. I mean, climate change is a really good example of a wicked problem. Michael Edward Lenert13:36:28 Yes. Mm-hmm. Jon Lebkowsky13:36:29 And I think when you're dealing with problems like that, and some of our most compelling problems are wicked problems. It's, it's hard to figure out exactly. I mean, the path to a solution is pretty rocky, right? Michael Edward Lenert13:36:45 Right, well, in a wicked problem, neither the solution. Um, nor the path there are particularly well-defined, and that often the solutions are contradictory. So, you could make the wing larger. But then it would make it noisier. And therefore, it would be ill-suited for certain airports, which would make the plane design uneconomic, and, like, the supersonic transport plane would make itself economically obsolete before it was physically obsolete. Yeah, so those are wicked problems. They have. um, different kinds of problems, and the, you know, the Apple engineers are asking problems that are. Like, um, if you have a ball and a bat, and uh… ball and the bat cost $1.10 together, and the bat costs $1 more than the ball, how much does the ball cost? You know, that's a novel problem to some people, and then it has a solution, and you have to be able to reason through it. The airplane wing has elements of that, but it's actually a much more complicated and difficult problem than the price of the ball and the bat. Jon Lebkowsky13:37:52 And do we think that an AI can tackle that problem? alone, or does… Michael Edward Lenert13:37:58 Interestingly enough, I asked the AI that this morning. I said, could you solve this problem? And it said, absolutely not. I cannot do that. That's not my… that's not in my range. I could be a coach, or I could be a sounding board, but I cannot make these decisions, because I have no skin in the game. It said to me that I could talk endlessly about the wing design. and make it seem compelling. But at no point would I ever be forced to test the wing and see if it failed in flight. Jon Lebkowsky13:38:35 And I think that's the most critical thing about AI, is that it only works in collaboration with. A human. Michael Edward Lenert13:38:43 Mm-hmm. Jon Lebkowsky13:38:44 And that the real skin in the game is the human's skin in the game. Michael Edward Lenert13:38:48 Right. Jon Lebkowsky13:38:49 The real accountability is the human's accountability. Michael Edward Lenert13:38:52 Yes. And that would be the beginning of an analysis of machines in general. Like, IBM used to say, a machine should never make a management decision because a machine cannot be held accountable. But now that the AI is so powerful, it's almost seductive that we would give. It power to make management decisions, and then it's so powerful it would be able to justify, in a clever and compelling way, its management decision. A perfect… perfect loop. Jon Lebkowsky13:39:22 Yes. It's really interesting, I, I was in a. I was working in a sort of, like, futurist focus group where we were working on a scenario. Michael Edward Lenert13:39:33 Mm-hmm. Jon Lebkowsky13:39:34 And the scenario envisioned a future mayor's office. And it was the the project itself was… To come up with scenarios that would. Uh, sort of enrich the thinking about the forward trajectory for a mayor of a particular town. And one aspect of that scenario, which was not… it wasn't critical to the scenario, but I thought it was an interesting piece of it. was that the prior mayor had been an AI. And the reason that the prior mayor, who was an AI, was no longer a mayor, was because at the federal level, they passed a law that said you cannot allow AIs. To run cities as mayors, because they had had so many disastrous effects. from handing that responsibility over to the AI. Michael Edward Lenert13:40:39 Well, this is… this is the essence of a… of a good management question. How much… discretion and agency do you give the AI to make critical decisions? Whether it's designing the wing for a new airplane, or designing the subway schedule for the… the city transit system. And you could imagine that if you were trying to optimize a bus schedule, an AI would be a very powerful tool to help you do that. Jon Lebkowsky13:41:10 But can you ever completely hand over critical decision-making to an AI? Wouldn't you always want to have a human in the loop and have the final decision being the human's decision? And is that not a fruitful. Uh, uh, quest question for, like, how we train people to work with AIs. How we train people to use AIs as a support. And never to concede to the AI completely. Michael Edward Lenert13:41:41 I think that's a very wise and thoughtful approach to this, this dilemma that is, that the, um… on the one hand, if you give it to the AI, you can maximize, in some concepts, efficiency. But on the other hand, you may totally lose track of your other values, such as what constitutes the good, what constitutes a well-lived life, what constitutes an organized society that's worth living in. Scoop Sweeney13:42:09 One problem with AIs, uh, replacing humans, uh, has occurred, or is trying to occur in Hollywood with the. uh, advent of AI-created actors and actresses and even musicians, they have a… they have a band out now. It was a horrible band, but uh… There's a lot of concern among the humans in Hollywood that they may be an endangered species. Michael Edward Lenert13:42:43 Well, here's the thing, is that our taste in movies has moved from. in many cases, what we would call serious film, to actually, the big franchises now are based on cartoons and on comic books. Um, you know, that the Avenger series is very powerful, but those are essentially cartoon characters. In that, and the amount of range that an actor needs could probably be simulated fairly well by. Uh, a simulation of an actor. However, if you're doing French lieutenant's Woman, that's probably not gonna work. You know, if you're doing a serious film that. the subjectivity of the human is a super important element, and that your relationship as a member of the audience to the subjectivity you're seeing displayed on screen. If you thought for a minute that you were being emotionally moved by an AI, you would feel you were cheated. Jon Lebkowsky13:43:43 Yeah, I've been thinking a lot about how. how film… filmmaking will change with AI. Apparently, a lot of filmmakers are starting to. Think about incorporating AI in their practice. And, of course, the. common way we've seen that done is CGI is probably considered a kind of AI tool. Uh, it does… it does leverage an artificial intelligence. Um, but there are other ways that you could incorporate AI. You could… you could use AI to help. Structure your story, outline your film in advance before you even write the script. And I just find myself wondering how… Uh… I mean, does that even matter? Is there any reason to be upset that AI is. Become part of that process. If the final film is a really good film, is that a problem? Michael Edward Lenert13:44:44 That's an interesting question. That's a very good question. We have to then ask what is the purpose of a film and art, and what constitutes a very good film? And, uh, if, um, new tools and new techniques can create. product that meets those criteria or exceeds them, then we have a new challenge. We have, uh. Um, music that used to need to be played in person to be enjoyed. With radio could be enjoyed over the air, and a lot of people said, oh no, the live performance is what matters. But the current generation, um, there's a split of opinion as to whether or not the live performance is critical, or whether or not. The music distributed on the platforms is sufficient for them. Now, we just finished Coachella, I think, or it's still underway in California, and there's a huge demand for that. Face-to-face interaction with the performers. Even though you're interacting with their personas and not with them personally. Jon Lebkowsky13:45:45 It's… Well, I mean, it's true, you can. You can stream pretty much all of Coachella. You don't have to show up for it, but they still sell out. Michael Edward Lenert13:45:59 Because there's something special. Jon Lebkowsky13:46:01 Right. Michael Edward Lenert13:46:01 And something unusual about that moment of being entrained or in sync with all those people, and we're all doing the same thing together, and we're experiencing the same moment of subjectivity that's being created on the stage. And it's resonating through us. That's, um… that's something that's not actually in the… A general understanding of media or communication studies, because we tend to use the individual as the unit of analysis. In Coachella, the proper unit of analysis may be larger than the individual, that there's something about being in a group of humans experiencing the same thing together, which is essential and important. Scoop Sweeney13:46:41 Yeah, the big concern… Michael Edward Lenert13:46:41 And then you will… you'll… yep. Scoop Sweeney13:46:44 The big concern in the music business has been the training of the AI that's going to create this new content, and there's a lot of copyright problems already surfacing. And, uh, I think the litigation may slow down that development somewhat. Michael Edward Lenert13:47:04 Yes, but see, what's interesting about United States copyright law is that we credit the author with all of the. Uh, rights. The performer has no rights. You, in fact, cannot count copyright a performance. As we move to a new era, it may be that the most important thing is not. The words and the music, but the interpretation given to them by the performer. Which, ironically is not copyrightable. So, if you look back at folk music, or in certain African American traditions, the songs were all known by everyone, and what mattered was how good on how unique your performance was. For example, James Brown. Um, he could do a standard, but it was James Brown version of the standard. Uh, so the copyright law is really an artifact of… Written culture, where we privilege the person who did the choreography, and they can get the copyright. So the person who wrote the dance down gets the copyright, but James Brown knocking us all out with his performance, that's not copyrightable. Jon Lebkowsky13:48:17 Yeah, that's very interesting. I often hear people say that. AI is essentially plagiarizing, um, because it's. Michael Edward Lenert13:48:31 You went mute. Jon Lebkowsky13:48:34 Sorry about that. Um… What I was saying is that I often hear people profess a concern about. plagiarism by AI. The fact that AI is trained on various resources, uh… really tons of resources. I mean, they feed huge amounts of. of data into an AI, and much of it is from copyright sources. Michael Edward Lenert13:49:00 Mm-hmm. Jon Lebkowsky13:49:04 And that this would necessarily mean that when the AI. when the AI. When the AI produces words. when the AI produces content. That there's an element of plagiarism there but to me, you would only have plagiarism if the AI was exactly replicating. Some of that source material. Uh, which could happen. I'm sure that can happen, and I'm sure it's a concern. But I don't think that it's inherently… inherently… it's sort of like… If I wrote a book, and that book was influenced by all the books I've read, does that mean that I've plagiarized those books? Michael Edward Lenert13:49:55 Well, here's the interesting thing, John. Copyright is actually in the Constitution of the United States, so it's not something that's bolted on afterwards. And the key concept is infringement and not plagiarism. And infringement is supposed to be, by the courts, calibrated so that the author of an original work. Fixed in a tangible medium is rewarded for that effort, but not in perpetuity. After some point in time, it is contributed back to the public domain, and those ideas can be freely used again. We have, of course, monetized copyright beyond any recognition of the initial. you know, 18th century notion of it. And, um, you know, the length of copyright has been constantly extended far beyond what would be necessary to compensate the creator of an idea. So, if we look at the original purpose of copyright, it's not to create compensation that's a revenue stream for life for the copyright holder. Rather, it's to give him or her just enough money so they are continued to encourage, be encouraged to create new things. And then after a reasonable amount of time, their idea is able to be used by others. Jon Lebkowsky13:51:11 I didn't think you could copyright an idea, though. I thought you could only copyright. Michael Edward Lenert13:51:11 For other purposes. Jon Lebkowsky13:51:16 The expression of an idea, the specific expression. Michael Edward Lenert13:51:18 That is true, you may… It has to be fixed in a tangible medium, so you cannot copyright anything other than an expression of an idea. Jon Lebkowsky13:51:23 Yeah. Michael Edward Lenert13:51:28 That would be correct. So, um… but in the case that we're talking about, where you're mentioning it, the purpose of copyright law is not. Uh, compensation, the purpose of copyright law is creativity. Jon Lebkowsky13:51:41 Well, yeah, I mean, I wasn't really talking about compensation, but just the fact, the fact that. Michael Edward Lenert13:51:41 and ingenuity. Jon Lebkowsky13:51:46 Just the question, rather, of whether. Of whether copyright is even an issue unless. The AI. Repeats the, as you say, idea exactly the way it was written by whoever originally wrote it. If they're just drawing on those ideas, if they're just drawing on that material, is that really. infringing. Michael Edward Lenert13:52:18 Clearly not under… under… under law. Um, you have all kinds of nuances to this, but the whole point of copyright law is to declare certain things out of bounds for a short while. And then those things get contributed back into the public domain. And even for things that are in the… corral of copyright, there is the doctrine of fair use, which of course suggests that there are reasons, good reasons. Socially justifiable reasons for allowing people to take copyrighted material and use it for various reasons. So, yes, I think absolutely that we have a difficult problem here, is that the AI is… Potentially crossing the line from, um… being derivative. Which would be an infringement to being truly creative in the way it's recombining these elements that it's seeing in its training data. Scoop Sweeney13:53:12 You mentioned fair use, and that's one of my favorite, uh… Michael Edward Lenert13:53:15 Mm-hmm. Scoop Sweeney13:53:16 Subject, because lately, fair use has been under attack. Uh, by the social media companies because they really don't like that idea and have. started to really suppress any use that may… under the old… yeah, I worked at radio for a long time, early on, before I got into IT. Michael Edward Lenert13:53:38 Right. Scoop Sweeney13:53:41 And fair use, uh, was fairly common, I mean, well. It was used by DJs and stuff of, you know, who. take copyrighted stuff and play 20 seconds of it, and that was considered fair use. But now, you can get into big trouble on social media by using anything at all. Michael Edward Lenert13:54:05 Yes, absolutely. The joke is that it's… the level for fair use is two notes. If you use only one node, it's fair use. But why don't you get to two notes? Jon Lebkowsky13:54:17 Well, I mean, the problem with fair use, as I always understood it, was. Was that it really isn't clearly defined, and it's going to take a court to decide it. So, anything that you… Publish on social media, you know, that comes from some source. there's a potential that they could sue you for that. The suit… the conclusion of the suit might be that it was fair use, but you still have to go through a lawsuit, and you have to pay your legal costs and that sort of thing. Uh, though, on the other hand, I would think that, uh, uh, Section 203 of the Communications Decency Act would. Protect these social media companies in those cases. Michael Edward Lenert13:55:01 Yeah, that's a very interesting and nuanced area, what copyright protections section 203 creates for the companies. It certainly insulates them from certain liabilities, but it's not clear about copyright. So, very difficult area, and. You know, at some point, I would tend to be a copyright anarchist on this here, unless you have a compelling need to restrict somebody's. Use of your ideas, those ideas should be allowed to be part of the free flow of information. Now, this is distinct from… this is distinct from patents, of course. Patent would be different, but, you know, unless… unless there's some compelling need for you to. Scoop Sweeney13:55:33 Yeah. Jon Lebkowsky13:55:33 Actually, I think… Michael Edward Lenert13:55:41 say that no one else can tell a story about Captain Kirk and Mr. Spock. You know, until way past my expiration date. That doesn't make any sense. Why can't other people use those two characters to tell stories? Jon Lebkowsky13:55:55 I'm actually thinking that what section 203 says is that. In order to have this protection from liability, that you have to be responsive to takedown requests. And that may be why a lot of the stuff's been taken down. Michael Edward Lenert13:56:12 Yeah, but the problem is that the safe harbor for the next generation of creators has been greatly constrained. So, like we were talking earlier, that a disc jockey knew that if it did a needle drop for 5 seconds, that was probably fair use. Because it did not diminish anyway the demand for the record or the compensation for the record. In fact, it probably helped sell records. that short, um, usage. Um, but we've gotten very covetous in our society, and, um, the rules have been constructed to favor, um, large accumulations of intellectual property and capital. Scoop Sweeney13:56:53 You know, we encountered, uh, something like that. I post all of our stuff on YouTube. And we did a, uh… I think it was an end-of-year update, and I included on that, uh. a short clip from a U.S. government thing about, uh. Um, nuclear warfare, I believe it was. And we got dinged for it, and they said, you can't do that because there's a copyrighted musical artist that has that on there. YouTube channel. And I finally found it, and it was a rap artist who had gotten the same government clip. And put it in their song. And I challenged that, and I never heard back from YouTube. They just didn't want to get involved, because basically you can't go out and. copyright, uh, government, uh, publications, as far as I'm aware. Michael Edward Lenert13:57:57 Well, certainly federal government publications. There's an odd moment in copyright law where the federal government may not hold a copyright because it's the federal government. Federal government materials are inherently… copyrightable. However, state government materials can be copyrighted. And this is just an… this is just an anomaly in law. Scoop Sweeney13:58:14 Yeah, this is… Michael Edward Lenert13:58:21 Yeah. Well, the same thing would go with, um, uh, Bureau of Land Maps and things. Those are ones produced by the government and published, they're not copyrightable, and anyone may reproduce those maps. And certainly a public service announcement that was produced by the federal government, um… Could not, in theory, be copyrightable. Although, if the public service announcement did contain a certain amount of a known song that was, in fact copyrighted, that's a complicated problem. Because now you're infringing not the government's copyright, which it does not have, but the copyright of the. Piece of music that's in the public service announcement. Which the government may or may not have had permission for. And of course, the whole thing with permissions for music could take up an entire hour of discussion. It's a crazy world. Jon Lebkowsky13:59:15 Yeah, and we're actually… we're getting close to the end of our hour, and we haven't complained about Trump at all. Which bothers me, I mean, we certainly should. One of the things that you've been, I mean, you've been working on this concept of, uh, of. Michael Edward Lenert13:59:31 Mm-hmm. Jon Lebkowsky13:59:33 political kayfab. Michael Edward Lenert13:59:34 Mm-hmm. Jon Lebkowsky13:59:35 And I would like for you to say a little bit about, kind of, what… What the principle is there, kind of how you define that. Michael Edward Lenert13:59:43 Well, what I initially started was the question of how we tell stories to one another, and then it occurred to me that kayfabe is a very specialized form of storytelling. Uh, related to the suspension of disbelief. And when you get back to the first stage, you have, you know, the UHF stations showing from the Sacramento Memorial Auditorium a wrestling match, which is absolutely preposterous. You've got the Von Braun brothers with their Nazi manager going against Pat Patterson, the All-American boy. But everyone agrees that it's real, and they try and maintain that, you know, keep it fake, you know, maintain that secrecy, and that this is real. And so for a certain number of audience members, known as Marks. These marks are actually believing that they're wrestling, you know, when I was 12, I thought these people were actually doing these things. You get to stage 2, where… At some point, there's an acknowledgement that it is not true wrestling, it's a performance, it's an athletic performance, but that doesn't change the avidity at which the fans still seek this. experience. In fact, it almost intensifies it because they become complicit in the co-construction of the narrative that is displayed on the wrestling stage. And the best example of that is a famous wrestler, um… Who, uh, started out as a heel, excuse me, started out as a face, that is, a good guy. And, um, his role wasn't bought, and so they switched him to the bad guy, made him a heel. And he became a huge wrestling star. And it is not coincidental, Trump is a member of the Wrestling Hall of Fame. He understands these procedures. And if you were to look at a lot of his statements, they're not lies. They're wrestling promos. Jon Lebkowsky14:01:39 Yeah, and, and this is kind of a. I think this is a thing with the Republican Party, uh, I, I know that. years and years ago, uh, before anybody ever thought about MAGA. I had a cousin who was a Republican, and he would occasionally forward me one of their. Emails, you know, they were… this… before social media. Republicans swapped emails with each other, and they made the kind of things that you see them say, and the kind of remarks you see them making in social media now. Michael Edward Lenert14:02:11 Right. Jon Lebkowsky14:02:14 Uh, where it would be, like, preposterous things about Democrats, for instance. And I wrote him back when he sent me one of those things, and I said, you realize that that's utter bullshit, right? That's just ridiculous. And he said, oh yeah, of course, we know it's bullshit. He said, we just like to send these things around to each other, you know, it's like… They liked the vibe of it, it reinforced. I mean, it was not true, but it conformed to a kind of truth that they had in their head about Democrats, you know? It was just, it was just a way to, to carry a vibe. And uh… to me, that's a lot like what you can do with kayfabe. You know, you can, yeah. Michael Edward Lenert14:02:57 Absolutely, and this… this then takes you into what I'm calling stage 3, because in Stage 2, the audience is just going along, and they're there for the. Um, show. In stage 3, the audience actually starts to co-create the narrative. Like, I'm sending the… you know, like fanfiction or something. They are involved in what I am calling in my book a participatory myth culture. This is a step forward from Henry Jenkins. thinking. It incorporates what he's doing that takes it one step further, where you're living inside a mythology that you co-create with other people, and that mythology that you live in is a choice, because it's more pleasant and more engaging to live inside this fictional K-fob world. Then the real world. Jon Lebkowsky14:03:44 And that's MAGA and that's QAnon and that's, yeah, absolutely. Michael Edward Lenert14:03:47 Yeah, so I have this quote here, it says, the narrative is not an account of what happened, so it's being passed around in the emails and now on social media. It is a structure that makes feeling affect livable, coherent, defensible, transmissible. We do not adopt stories because they are true, we adopt them because they preserve something we already feel. And so those stories that were being circulated were about the feeling that the Democrats were lamos, that the Democrats were cheaters. Didn't matter that the story was true, because you already knew it. Jon Lebkowsky14:04:22 Right, so that explains a lot of what we're thinking of as lies. From a particular party. We're actually, we're kind of at the end of our hour here, and I. Michael Edward Lenert14:04:31 Mm-hmm. Jon Lebkowsky14:04:32 Good, good time to end it, because now you've got to go write the book and then you can come back and talk about the book. Michael Edward Lenert14:04:38 Absolutely, and I'd be delighted to share… to share that with you. So, it's been a pleasure and a privilege to chat with you both. And a lot of new ideas that are going around, and it's fun to talk with equal minds who can share these difficult and interesting ideas. I really appreciate this opportunity. Jon Lebkowsky14:04:56 Yeah, this has been a blast. Thanks for joining us. Scoop Sweeney14:04:58 Thanks, Ed. Michael Edward Lenert14:05:00 Alright, thank you so much! Jon Lebkowsky14:05:01 And that's a wrap!