← POLITICO Tech
POLITICO Tech
Rebroadcast: Joseph Gordon-Levitt on AI, Hollywood and owning your ‘digital self’
00:28:26
Loading summary
Transcript94 lines
- [00:02]
A
Well, I was down on my last dollar. Then I started saving Cause the bank said fiscal restraint is what you're craving. So I put my earnings in a high yield account, Let the savings compound and the interest mount. I'm optimizing cash flow, putting debt in check. Now time is my friend and not a pain in the neck. And we've got a little cash to rebuild the old deck. Boring money moves make kinda lame songs but they sound pretty sweet to your wallet. Brilliantly boring since 1865.
- [00:31]
B
You're not anti tech. I don't see you as anti tech.
- [00:33]
A
Okay, I'm glad you don't. Because, you know, oftentimes when I say some of the things I do, the first response is luddite anti tech. You know, decelerationist de growth, this, that, the other thing, which I don't agree with any of those positions.
- [00:50]
B
Hey, welcome to Politico Tech. I'm your host, Stephen Overle, and on this show I break down tech politics and policy with the people shaping our digital future. Hollywood has been one of AI's harshest critics. In fact, talent agents, voice actors, directors, and other film industry players have been lobbying in Washington for protections. Among the most vocal skeptics is actor and director Joseph Gordon Levitt, who I interviewed for the show back in June. For the show today, I want to bring back that conversation. Joseph, welcome to the Politico Tech podcast.
- [01:37]
A
Thanks. Good to be here.
- [01:38]
B
You know, I'm curious if you can take me back to the first moment you experienced generative AI and you thought like, oh, this is going to be a problem.
- [01:46]
A
Ah, well, it's interesting. I've had a bit of a front row seat to this technology for quite a while, honestly, because my wife works in the space.
- [01:55]
B
Right.
- [01:55]
A
So I was paying attention to it a bit before it became such a, you know, an attention magnet that it's become in the last couple years. And I remember seeing early versions of some of these large language models and saying, this is bogus. Yeah, this isn't ever going to amount to anything. And then completely eating my words a couple years later as it got so incredibly impressive.
- [02:18]
B
Well, it's changed, I mean, incredibly fast in just a few years. Was there a galvanizing moment for you? Because I, you know, I talk to creatives who have concerns about AI. Not all of them become so outspoken. What prompted that?
- [02:31]
A
Well, I think I've, I'm used to speaking about the intersection of media and technology. I've been paying close attention to it for a long time. I did this media tech Startup that called Hit Record, that didn't start at all as a media tech startup, but sort of grew organically out of this hobby that I was doing my brother, but eventually ended up as a, know, venture capital backed, you know, company.
- [02:52]
B
Right.
- [02:53]
A
And so I, I learned so much about it that I feel like I. I don't know, I guess I was sort of fluent in a lot of these dynamics, whether they're technological or economic. And so as I see this, I feel compelled to talk about it publicly because I do feel like to a certain extent there's not widespread understanding of exactly how the tech works and what that might mean economically, not only for content creators, but for kind of everybody in the future.
- [03:26]
B
What's the biggest misunderstanding you see?
- [03:30]
A
Well, even just in the word artificial intelligence, you imply that this, there's this thing, this sort of independent entity that's been created when in fact, you know, I hear some people call it collective intelligence. The truth is that the way these models work is they are fed enormous sets of data and then they crunch the numbers and these algorithms kind of find the patterns and generate outputs that follow those patterns. So there's no intelligence in there other than the human intelligence.
- [04:05]
B
An amalgamation of human intelligence.
- [04:08]
A
Exactly. And I don't know how widely that's totally understood by most the people that are using these chatbot products today.
- [04:16]
B
And so when you see something like take VO3, right, Google's a new AI video generator. I mean, what about that sort of has you the most rattled. Is it how good it is? Is it all the data and information that's gone into training it, what is it?
- [04:33]
A
Well, the first thing I would say is it's deeply inspiring. While I have concerns about it, I also am completely enamored by the dazzling potential to.
- [04:45]
B
Right, you're not anti tech. I don't see you as anti tech.
- [04:48]
A
Okay, I'm glad you don't.
- [04:49]
B
Yeah.
- [04:49]
A
You know, oftentimes when I say some of the things I do, the first response is Luddite anti tech. You know, deceleration is de growth. This, that, the other thing, which I don't agree with any of those positions.
- [05:02]
B
Yeah.
- [05:03]
A
But I remember being a little kid and making videos, and if I had the tools that my kids have today, I'd be so stoked. Like, it'd be so much fun. So I completely see how this can be beautiful and fantastic, but I also see downsides that are probably avoidable if we all, you know, get on the same page about them. So what are some of the downsides? You use the Example of this new video generation tool from Google. How does that thing work? Well, it's not that they just invented this thing that can magically make videos. It's that they took everybody's videos without permission and without compensating any of those people. And that's why their new tool can make all these videos. Now, look, their tech is great, but does it make sense that they deserve 100% of the economic value and all the humans whose data they took deserve 0%? That doesn't seem to make sense to me because especially if you, if you take that basic principle and then you take it a few steps down to the future, take something, not just videos. What about, say, I don't know, professors, Right. Pretty soon there's going to be a bot where people can log on and talk to a professor bot and say, teach me college level history or engineering or whatever, and the bot will be able to interact with them like a professor. And that's going to be great. Again, I think that's wonderful.
- [06:27]
B
Right? A lot of good uses, a lot of positive uses.
- [06:29]
A
But, but okay, how are they going to make that bot? The way they're going to make that bot is they're going to take video and text and stuff from professors and suck it up into their model and then they're going to make the claim that all the money they make selling their professor bot, they get to keep 100% of the money and the professors get 0% of the money.
- [06:48]
B
Right.
- [06:48]
A
That, to me doesn't seem to make sense. Not only is it unfair, it then sets us up for a very bad incentive structure where now professors don't have an economic incentive to be good professors anymore. And that applies whether you're talking about content creators on YouTube, professors in universities, or just about any other knowledge work, I guess you could call it moving forward. I'm worried about a world where there is no longer any economic incentive to have good ideas and act on those good ideas and be creative and be excellent and strive.
- [07:24]
B
Right.
- [07:24]
A
If you want a healthy economy where people are doing those things, you want to incentivize them to work hard by compensating them for their hard work.
- [07:35]
B
And I think it also raises questions from the tech standpoint too of like, you know, if these models to date are driven by human knowledge, human ingenuity.
- [07:42]
A
Yes.
- [07:43]
B
If that stops, if that's disincentivized, can those models continue to advance just on their own?
- [07:49]
A
It's a great point you're raising. And if you raise that point with a lot of these folks in this industry, they will tell you no, our models are going to start producing synthetic training data. We already have all the human input we need, and their intelligence is going to be all we need moving forward. That's, to me, a very dark and dystopian assertion to make. The idea that at this point, humans have done all the thinking that we need to do and we don't need to do any more moving forward, that the models have got it from here, that's just, to me, on its face, insane and sort of misanthropic.
- [08:27]
B
This notion of, like, AI theft and taking intellectual property from others. You know, just a few days ago, Disney and Universal became the first major studios to sue an AI company mid journey, accusing them of using their IP to train models. The courts have. There's other lawsuits happening. The courts have to sort all this out. I'm curious if there is a solution to this that you would like to see.
- [08:51]
A
So I've been really inspired by the work of this pioneer technologist named Jaron Lanier, and he has a partner who's an economist named Glenn Weil. They both work in the office of the CTO at Microsoft. These are not like fringe theorists. These are like very prominent, successful thinkers. And they've done a lot of work on this. And in fact, they started doing the work in terms of social media, not just AI, because the social media business model that came to dominate is similar. It's all about harvesting data from people without consent or compensation. I guess they get consent when they get you to, you know, click terms.
- [09:35]
B
Right. Exactly. As we all do in 30 seconds or less.
- [09:38]
A
Exactly.
- [09:39]
B
But.
- [09:39]
A
But Jaron has written, I think, very illuminatingly on how this is a bad economic model. It leads to something more like feudalism, where the king owns the whole digital world and we are all just serfs on the king's land. And that's not a recipe for a vibrant world or a healthy economy moving forward. What you want is for people to be able to have ownership. That was the great innovation in the American Revolution. Right?
- [10:08]
B
Yeah.
- [10:09]
A
Here we are in Washington, DC. I'm feeling very patriotic today. I took my kids to the Smithsonian this morning. And so Jaron has done a lot of work on, and is continuing to do work on what it would take technically to be able to track pieces of training data through neural net.
- [10:29]
B
Okay.
- [10:29]
A
A neural net is what makes these large language models.
- [10:32]
B
Right.
- [10:34]
A
And then be able to give meaningful attribution to the most important sources that contributed to any given output.
- [10:42]
B
Got it. Technological solution to it, because that's it. Sort of like, you know, the cliche unscrambling the egg. Once all this information goes into the models, sorting out what. Where it took inspiration from or where it literally took, you know, exact information from is a hard egg to unscramble.
- [10:59]
A
That's right. And what, what a lot of the tech companies say now is that it's impossible. What that means is they haven't built it yet.
- [11:05]
B
Right.
- [11:06]
A
And it would be expensive to build and they don't have a business incentive to build it. I think, though, there is a business incentive and it has to do with that same analogy, that ultimately feudalism is not the strongest economy, but any one company doesn't have the business incentive to do it. This is why this is a technological problem and a policy problem and ultimately a cultural problem, which is, I guess, where I come in. Because, you know, my domain is in art and entertainment and storytelling.
- [11:38]
B
Right.
- [11:39]
A
And, you know, this is why I'm speaking up, because I think a lot of this has to do with just more and more people getting the gist that we're heading for a digital world that's not going to be good for most people.
- [11:51]
B
I was going to ask you what the counterweight is to these tech companies, because especially here in Washington, they have huge influence. Is it Hollywood? Is it policymakers? What is the counterweight that will apply pressure to make that change?
- [12:05]
A
I don't think Hollywood is the one that's going to do it. Frankly. I'd love to say that, yeah, we Hollywood people are going to stand up and make it happen. Like, I don't think that's quite realistic, but I do think people at large, a greater understanding as people become more and more sort of versed in what this technology is and how it works, that's ultimately what's going to do it. Because, not to sound partisan, but I always liked that one thing Obama used to say when he used to say, you have to make me do it. And I think that's. That's on us. That's, you know, the quote, unquote, we the people, we have to care about this. We have to understand that the Internet as it's set up right now is disempowering to ordinary people. And it's putting all the power, more and more concentrating into the hands of these few gigantic tech companies. And it doesn't have to be that way. This episode is brought to you by Rumchata, a delicious creamy blend of horchata with rum. It's best enjoyed over ice or in Your coffee delivering vacation vibes anyway, or anywhere you drink it. Find out more@rumchata.com Drink responsibly. Caribbean rum with real dairy cream, natural and artificial flavors. Alcohol 13.75% by volume 27.5 proof. Copyright 2025 Agave Loco Brands, Pojoaquee, Wisconsin. All rights reserved.
- [13:42]
B
I knew you're feeling patriotic. You're here in Washington. You signed onto this open letter to the White House around copyright protections for entertainment, right. We've seen AI companies pushing for a very broad license that would sort of exempt them from needing to abide by some copyright laws and protections. What can Washington do? What could policymakers do that in your view would help address this problem?
- [14:08]
A
Make a law that says your digital self belongs to you.
- [14:11]
B
You just have ownership of your. What is it? Your image or data? All the above.
- [14:15]
A
Yeah, that if any, that any data that gets produced by a person, whether that person wrote a thing or uploaded a video themselves or was just tracked by some location sensor in their phone, all of that data has information that you, you the human helped produce. And I think you, the human deserve some amount of any economic value that comes from that data. I'm not saying you necessarily have to have a hundred percent. It makes sense that the companies who built the technology that then tracked that data deserve some of the economic value. I'm just saying they probably don't deserve 100% of the economic value. There should be a balance. It should be negotiable. And look, it's not a simple thing. It. This requires a big, you know, robust ecosystem to figure out. But like that's how an economy thrives. When the king owned all the land back in, you know, the Middle Ages, that was real simple. And the idea of no individual people are going to own little pieces of land and then other institutions are going to own other pieces of land and there's going to be this thing called the real estate market and we're going to have all kinds of rules and regulations about how that works and you're going to have to have a deed and blah, blah, blah. You know, that all. We take it for granted. Now that all sounds obvious, right?
- [15:36]
B
But it had to be built up and over time and. Yes, through institutions and demand and everything you're talking about.
- [15:43]
A
That's right. So nowadays in the digital world, your data is. In a certain way, it's like building your house, you know, and if you're, if you're creating on one of these platforms owned by one of these gigantic tech companies, I've heard it said. This isn't my phrase. It's like building your dream house on rented land. You shouldn't do that. You should buy the land first before you build your dream house. Well, we should all just be born owning our own digital selves.
- [16:14]
B
I think that's such an interesting concept, and I think one, frankly, Washington is catching up to in some ways, they're so probably far behind. I'll say. The argument I hear so often tech companies make in D.C. is that winning the AI race, beating China. China, China, China, China, China. Exactly, exactly. You know where I'm going with this? Like that is so important that everything else is secondary. Right. And. And it's a very effective argument for them. I can see a counterargument to that on behalf of entertainment and some other industries. I'm curious if you see one.
- [16:47]
A
Yeah, sure. Well, the first thing I'll say about that is I don't think it's entirely bs Right, to make that, to bring that up. I do have concerns about autocratic regimes like China and the tech they're building. And I do think that we Western democracies need to stay competitive. So I think there is that. However, I do think there's some dishonesty going on. Part of this is based on this idea of the quote unquote, fast takeoff is what they used to call it in AI circles that that phrase has sort of gone out of fashion now because people don't think it's probably going to happen anymore. But what they meant by the fast takeoff was at some point there's going to be this threshold and they, they call it AGI. You know, listeners out there maybe heard this term artificial general intelligence, and it's this sort of mythic thing without an actual definition. There's no specific definition of it. But the idea was that one of us is going to all of a sudden achieve AGI and it's going to take off real fast and have this super intelligence. And whoever has the AGI will win it all, control the entire world. And. And then the comparison gets made to the, the arms race, you know, in the Manhattan Project during World War II, when. When Nazi Germany and the US were racing to build the nuclear bomb.
- [18:12]
B
Right.
- [18:13]
A
And in that moment in history, it really did happen that way. Whoever built the bomb first kind of won the war and was able to sort of set the rules for the whole global dynamic.
- [18:24]
B
It was either dystopia or utopia, depending on the outcome.
- [18:27]
A
Yes. But I think more and more it's clear. This idea of AGI is not really analogous to all of a sudden we built the bomb, we can drop it on Hiroshima and we can now win and control everything. And we, we can't let the bad guys do it. It's going to happen and bit by bit it's going to happen in a complicated way. And I don't, I don't see how someone getting a better AI model six months earlier or six months later is really going to make that big a difference. I think ultimately this is a thinly veiled disguise for, for saying we just want to make maximum money, we want to win. And, and look, they do have a business incentive to get there first.
- [19:14]
B
Right?
- [19:15]
A
But let's be honest. Do we think that these companies are prioritizing national security and the public good? No, they can't. And it's not because they're bad people. It's just not what they're set up to do. They're set up to prioritize their business interests.
- [19:30]
B
Right? Profitability is paramount.
- [19:32]
A
That's what it is. That's what they have to do. If you don't do that, then your competitors do. That's why we can't just be leaving all of this up to private businesses. And it's not to say that we should ignore issues of national security, we should be paying close attention to those competitive dynamics, but we shouldn't leave that up to for profit businesses who have no accountability. The Manhattan Project was run by the government. It was run by patriotic generals, by soldiers. Soldiers whose ethos is I'm gonna put, put my life on the line to protect the safety of the American people and ideals and. Yeah, exactly. But do we think that these business people in Silicon Valley and other places have that same mentality of a soldier who would die for the safety of the American people? I just, I don't see that these are businessmen and they are forwarding a business agenda and that's why they should be forwarding their business agenda. It's good to have ambitious, fast moving businessmen. They just also need to be, you know, complemented by guardrails that are put up by the public right through the government.
- [20:40]
B
Is there a guardrail you would really like to see? I mean, I know you mentioned giving people ownership of their data and their digital image. Like, is that enough? Is that the starting point? What's, what's the first step to actually putting some parameters on these all powerful tech companies?
- [20:55]
A
Yeah, well, one, one thing to mention is there's a couple bills being put forward in New York State right now. Actually one just today Passed in the legislature, I believe it's called. It's called the Raise Act, Right. In New York, and hopefully the governor will sign it into law. And it's just putting some basic sort of safety and transparency protocols into place for the biggest AI labs. There's another one in New York called the. I think it's called the AI Training Data Transparency act, which gets it exactly what we've been talking about. It doesn't necessarily implement everything we just spoke about, but it's a good first step that these companies should have to be open about the data that they are frankly stealing and have stolen to train their models.
- [21:46]
B
I mean, there's a lot of products that are subject to regulation and transparency and safety reporting.
- [21:51]
A
That's exactly right. Almost every major industry has some rules and regulations they have to follow. Follow. Of course they do. And we all stop at red lights when we drive.
- [21:59]
B
Right. One thing I know you've talked about in the past, and I think it's an important point because obviously you're a Hollywood celebrity, but this is not just about the rich and the famous. The impact of AI is going to be felt across every industry, every sector, every worker. We covered a lot at the early days of this podcast, the strikes that the writers and actors did in 2023, and some of the protections they won. Do you. I'm curious how you think that's going two years later. And do you think that in any way creates a model for other industries to protect their workers as AI becomes more ubiquitous?
- [22:34]
A
I think especially SAG did a pretty good job for what they are. They're a, you know, a labor union of a certain size. I don't think any one labor union is really going to be able to. To make this right.
- [22:52]
B
It's.
- [22:52]
A
It's gonna take laws. The lawsuits that are pending are also important. The Copyright Office just weighed in. You probably heard about this. You know, they released a report saying, like, hey, it seems, as you know, according to our opinion here at the Copyright Office, that the claim that these labs are making, that it's quote, unquote, fair use to hoover up everybody's data and monetize it at scale. That's really not the spirit of fair use. It's maybe covered for some particular cases, but maybe not most. You know, the Copyright Office weighed in, and with that report, the very next day, the head of the Copyright Office was fired. The Executive branch would not specify a reason why the head of the Copyright Office was fired. It was literally the next day after this report. So, you know, we Definitely have an uphill battle ahead of us. But that's why I think it's encouraging that state governments like New York State are. Are picking up the mantle and why it's so important that, you know, there's. I'm sorry if I'm getting like, too in the weeds, but not at all. You know, there's this provision we get.
- [24:04]
B
In the weeds here.
- [24:05]
A
Okay, good.
- [24:05]
B
You're welcome in this place.
- [24:06]
A
There's this provision in the big Beautiful bill that says states should not be moratorium to. Yeah. To regulate AI for the next 10 years. And I just read this. Even Marjorie Taylor Greene said, oh, I didn't know this provision about states not being allowed to regulate AI was in the big beautiful bill. If I had known that, I wouldn't have signed it. So let's hope that the Senate takes that out of there before they sign their version.
- [24:37]
B
Yeah, that's. People are starting to pay more attention to that. They should listen to our podcast. We talk about it because, yeah, it is starting to get some political pushback. Ten years is a long time. Long time to not allow state laws.
- [24:48]
A
Yeah.
- [24:49]
B
I've interviewed a lot of lawmakers about AI and I like to ask them, like, what shaped their initial understanding of artificial intelligence? The answer I most commonly get is entertainment.
- [25:01]
A
Yeah.
- [25:02]
B
It's Terminator, it's Minority Report, it's Mission Impossible. And it speaks to kind of an irony here where really, Hollywood's power of perception is so influential. And oddly enough, recently Google is starting to fund film projects that portray AI in a more positive light. Coming from that world, I wonder if you see that as like an underutilized resource in this conversation, shaping people's perception and understanding of this technology.
- [25:33]
A
I've thought about this a lot of how to make positive science fiction. It's a hard, creative challenge if you're going to make entertainment, because dystopias are just a much easier story to tell if you're trying to, you know, entertain an audience. But I think it's possible. Probably the best example ever is, you know, the early or earlier Star Trek series, the original. And then I grew up in the 90s on Star Trek the Next Generation. Yeah, this is a pretty positive, optimistic view of the future. And it was widely popular. And so I do think it's possible. One thing I'll say about Hollywood's portrayal of AI that I think is important to note, what Hollywood will almost always do is personify the AI. Yeah, there's a reason for that. It's because Hollywood's trying to keep you engaged and move you emotionally.
- [26:28]
B
They need a character story.
- [26:29]
A
Yeah, right. So, so it, it, it's really easy to make, you know, this scene where Arnold Schwarzenegger is shedding a tear because he's going to miss, you know, the kid that he's been protecting in Terminator 2, even though he's a robot or, you know, beautiful movie that just came out last year, the Wild Robot was sort of all about this robot that ultimately had feelings and a lot of what we attribute to humanity. I think that the personification of this technology is dangerous and that we should be very careful about letting ourselves slip into the delusion that these products are people.
- [27:18]
B
Well, Joseph, thanks for being here on Politico Tech.
- [27:20]
A
Yeah, thanks for having me.
- [27:23]
B
That's all for this week's Politico Tech. If you like Politico Tech, go ahead and subscribe and recommend the show to a friend or colleague. And for more tech news, subscribe to our newsletters, Digital Future Daily and Morning Tech. Our producer is Nirmal Maliko. Pran Bandy made our theme music. I'm Stephen Overlea. See you back here next week.