Game Economist Cast

E38: Economics of Game Innovation & AI's Now Proof

• Phillip Black

Share episode

Send us a text

What is the GDP-maximizing set of copyright protections? 10 years? 5 years? None at all? Chris, Eric, and I debate the relevance of patents and copyright protections and the gains to network effects of knowledge. Does the "gentlemen's agreement" to avoid patent protections on game design help or hurt the industry? 

Chris talks about Monster Hunter's lineage and woeful service, while Eric introduces a novel use of AI in game design. Phil believes the Gini coefficient is underutilized for measuring live-ops-driven revenue.

Chris K-S:

Oh on. In other news, speaking of tariffs and the eu, check out my sick new swag that I got at Aldi yesterday.

Eric Guan:

Nice.

Phillip Black:

my God. It looks

Chris K-S:

they're getting on the,

Phillip Black:

looks like they're spots.

Chris K-S:

they're getting on the Kirkland Train.

Phillip Black:

It looks

Eric Guan:

thought they were a European supermarket. What the

Chris K-S:

You don't have Aldi's in California. You just got Trader Joe's out there. Are you fucking kidding me?

Phillip Black:

California

Chris K-S:

of the best. It's

Eric Guan:

in the Midwest for

Chris K-S:

ugh.

Eric Guan:

16 years. I never saw an Aldi's. Oh. They do have'em

Chris K-S:

I think Aldi and I think, dude, I think I wanna say Aldi and Trader Joe's, they're like brothers. The owners of those companies or something. Even though I think Aldi was founded in Germany Aldi, trader Joe's relationship.

Eric Guan:

Oh, they are two brothers who founded Aldi, and I guess they branched out. Oh, so this is like the OG Trader Joe's.

Chris K-S:

Look at that. Yeah it's almost like a hybrid between Trader Joe's and Costco. So like Trader Joe's, you get like your, there's constantly a new kind of there's a fluctuating supply. You've got your like aisles that have the different pickles and the different chocolates and stuff. Except for everything is still like in a box, very utilitarian, very like. Minimal very low costs. They're super, they cut costs down by just putting boxes out that are filled with shit, and they have very few employees operating these stores. A lot of self-checkouts. So it's kinda like a combination between Costco. It's like the scale philosophy of Costco meets the small scale and variety of Trader Joe's. And they're able to provide. Products at a really cheap price,

Eric Guan:

Cool. I'll check it out.

Chris K-S:

game economist. Cast, episode 352.

Phillip Black:

five. All right, we can lead with that. Back from GDC, normal cast, normal crew, energized, refreshed, ready to talk about ai. Ready to get controversial. And also ready to talk about what we've been playing.

Chris K-S:

Phil, you're first.

Phillip Black:

so I, I have an interesting one for you guys. I've been playing a game called DC Dark Legion. This is a game by Fun plus company. I was talking about the, that other podcast where we mentioned that the founders are Chinese, but all of their games are four X games that have had a lot of success in the West, and they released a new game. That I would argue is very similar to modern four x, but also has a lot of squatter, PG elements. You're collecting characters, you're starring them up, but you're also engaging in like these multiplayer PVP maps as you would have in four x. And I was trying to figure out the thesis, and I think it has to do with basically. Restricting the size of the revenue waves that normally come with squat RPGs. So normally there's a bunch of spend when players get a character they want, or a character is the featured character for that month. It's very dependent on content and how many characters you can push per month. And four X tends to be steadier because the conflict is steadier, right? It's really about, battle for the seed of power, and that tends to have more consistent costs associated with it rather than having this huge fluctuation. And I was basically just trying to think about could we measure create a KPI ba tell us how wavy or unequal particular game is. it's obviously the genie coefficient. Like we should just use the genie coefficient for daily revenue. And what we can do is we can assign each game a genie. And if the genie is high, then we know that they've had a lot of inequality and individual days contributing to the revenue scheme. So I think a game like Chen would have a very high genie it has very wide range of revenue, whereas a game like last war would have a lot more steady'cause it's

Eric Guan:

But like what, What

Phillip Black:

What do you guys

Eric Guan:

give you knowing that day-to-day revenue is volatile?

Phillip Black:

I think it presents a novel challenge to solve. Like it revenue bouncing that high is an interesting observation that I think you can play with.

Eric Guan:

Is it good?

Phillip Black:

I think having that while of fluctuation in revenue is not, I don't think anyone wants wild swings like that, right? That's really sensitive to content.

Eric Guan:

yeah, but it might just be a reflection of the content and the val,

Chris K-S:

But what can you do

Eric Guan:

is being delivered to players. I, so I think about this. When I look at movies, for example, their revenue is incredibly spike. Spike, and then tail spike. And then tail, right? But that's just a nature What, the type of content they create and how it gets delivered. And it's not necessarily bad compared to a McDonald's, which is like a continuous stream, right? It's just a different model. Yeah.

Chris K-S:

Phil, you're saying we could create a revenue volatility index that is basically just the genie coefficient, IE spend inequality.

Phillip Black:

Yep. Revenue generation inequality Days.

Chris K-S:

it's a KPI, I don't know. I don't know. I think I, I agree with Eric here. It's what, you can look at like CPI or Arp Doo and you can infer from those actions that you might take to, to try to push them up or down. This one, it's first of all it's hard to influence. Spend inequality. That's a very abstract like concept. Like how do you pull down top spend? Do you push up bottom

Eric Guan:

are you talking about spend? inequality between

Chris K-S:

do you care

Eric Guan:

day to day spend, like between days.

Phillip Black:

I spend between days, so like we're measuring a genie coefficient within. Within each title. So like some, so if you looked at Gen impact as a really high crest and trough to its revenue wave because the gotcha characters when they release do so well and the drop off after players get it is so large. I I think it's to have spiky revenue like that is. Hard for it to become predictable. Like those waves are not always as predictable. And it's always just an expensive missile. Like I think people don't want to take few shots. They wanna take a lot more shots. You get more feedback.'cause if you miss on a character you miss a huge chunk of revenue. And then your next window is next two months. I.

Eric Guan:

I think it's an interesting observation, but I would flip it on its head and say, you've got this content, which is a new Genian character. How do you extract the most revenue from that content? And if that spike decay pattern is actually the best way to extract revenue from it, then know, why not do it. Now maybe the spike suggests that you could be price discriminating better, have it high priced early on, and then like gradually make it more accessible and more cheaper later on, and smooth out that spike. But I don't see the spike as intrinsically bad. Yes, there's some planning volatility, but, and uncertainty, but yeah.

Phillip Black:

No, I think that's a fair

Chris K-S:

So this DC game is I, so it's DC as in like the ip and then it's four x squad,

Eric Guan:

about this game at GDC. It's like a mashup of genres like DC Comics,

Phillip Black:

Yeah. Yeah. Very weird. But it seems, the thing I still there, there's this great presentation, Ethan, I. Levy, who was the other ghost? Deconstructive fun a while back, had about like this tower of want. I want this because I want this, because I want this. And now high, can you build that tower? And I think this game just really struggles with having a really high tower.

Eric Guan:

flat and wide or

Phillip Black:

I, yeah it's too flat, like a four star. My characters, they get them to level 400. I'm not quite sure what I'm supposed to do after that. It just feels aimless and it's very painful. With that grind I need bigger, stronger carrots. Yeah,

Chris K-S:

I need to try

Phillip Black:

Dark

Eric Guan:

Would you recommend people play it? Okay. All.

Chris K-S:

So speaking of games that I do recommend you play Monster Hunter Wilds, and shame on you, Eric, for returning it after you purchased it. Although it's fair enough if you haven't played World, oh no, you still have your job right? But your

Eric Guan:

yeah. We got, we got

Chris K-S:

job. You guys decided to go trad

Eric Guan:

income. So

Phillip Black:

Oh,

Eric Guan:

private school is, it's a You can't cancel for a whole year.

Phillip Black:

oh.

Chris K-S:

Oh, We'll talk about subscriptions.

Eric Guan:

Liking wild. wild.

Chris K-S:

subscriptions.

Eric Guan:

Yeah.

Chris K-S:

Yeah, it's the latest Monster Hunter iteration. It's the latest triple A quality monster Hunter game to come out since Worlds, which was, oh gosh, 20 17, 20 16, 20 18. I don't know. It was sometime in there. Monster Hunter is the key contributing factor to me upgrading my system. So I went from the Xbox One to the Xbox one s, or sorry, series S when World came out. And then I just upgraded to the X when when wilds came out. Because I just can't be, I just can't be having 30 FPS when I'm fighting my monsters. It's just not, it's not tenable. Anyway. Excellent. Box price, premium game,$70. There was a bunch of other DLC types of packs you could buy. I think like the max, I could be completely butchering this. I wanna say the most expensive one that I saw was like 99 bucks. They have a pretty, pretty standard DLC system. So they release the first game, they release the game, and then a year-ish later, two years-ish later, they release a major DLC. So they've done this with almost, they've done this with their last three games, probably more games before that. Before, but that was before I started paying attention to this kind of stuff. So Worlds that came out with a big update called Ice Born sun Break, which was spec, or sorry, monster Hunter Rise, which was just for the switch. That was not aaa. That was more like. Traditional back to their roots. Low poly, not super high fidelity, not like a super complex RPG. Like Worlds and Wilds. They came out with Sun Break, which is the DLC there. And then with this game, they're gonna, clearly, they're gonna have some sort of major DLC. The game is amazing. It's classic Monster Hunter. Definitely took some getting used to after going from Rise, which was a little bit more I don't know if you like picture the super tight mechanics of a fighting game. Then compare that to the looser mechanics of an RPG, like the Witcher three. In the Witcher, when you stop moving, your horse like will take a few steps and then it stops. Versus like a fighting game, like you stop moving, your character stops immediately. It was like that where In Rise, which I was playing before, the mechanics were much tighter, much more okay, you stopped. You stopped, you're done. There's also the polygons were much, it was much lower graphical fidelity, so you've got less shit flying around on your screen. So interesting to get used to that. I was actually really surprised by how little content there was. I pretty much finished the entire game and all the quests within 60 hours, which for a Monster Hunter game is. I think that's pretty short. The main storyline is like 25, 30 hours, which again is pretty short. They just released a content update, a free content update, but that DLC is gonna be coming in probably in a year or so, and that'll be like$40. I. Is usually how they do it. The one thing I don't like about this subscription or this DLC model is for even hardcore players like me, I'm usually burnt out of a Monster Hunter game after a hundred to 200 hours. And that's within a year or two. So by the time the DLC comes out, I'm done with the game and I'm, I've never bought

Eric Guan:

a revival event for you where you're like, Into it,

Chris K-S:

for me, and maybe it is for most people I know it works for, I know it works for them. They wouldn't do it otherwise, but yeah, it's I wonder if there was some way to push up sub DLC purchasing hey buy the all in pack and you get all DLCs from here on out for free. And then you'll have people like me who are like, oh, I'll definitely play the DLC and I wouldn't have bought it otherwise, but I do buy it at because it happens at release, you know what I'm saying? So that's Monster Hunter. Super straightforward monetization model. Nothing fancy. DLC, maybe they could push up spending by introducing the

Eric Guan:

I'm kind of surprised that they haven't evolved into a live service loop grinder kind of model, like similar to path of Exile or Diablo or whatever, because it, it's got a similar thing, right? You you grind up, you get your better gear, and like the power growth gets super slow over time and they just add more raid bosses at the end for the end game.

Chris K-S:

That was one of the worries for the community was because the wild is like much more kind of open world. It was much more of an open world live RPG, like the map is constantly updating, so it is live service in that way. So you're constantly having new monsters cycle in and outta the map. You're having different quests pop up on your radar with randomized rewards. So it's almost kinda de destiny, where you have these quests that fluctuate and then you've got like these big monsters that you can fight on a regular basis that are almost like raids, But they're not monetizing it at all. I think the community would be pretty pissed if they tried to, but it would probably make them a lot of money.

Eric Guan:

Cool. What have I been playing? So I've actually been playing this recently, but yeah, it's the AI episodes. I figured we should talk about Infinite Craft. Have you guys played this game? It was like a simple web game where you

Phillip Black:

No.

Chris K-S:

No.

Eric Guan:

You start with four words, earth, air, fire, water, and you combine two of them. And then there's an LLM like AI that a new word. So if you combine fire and earth, you might get lava. And if you combine lava and earth, you might get volcano. And volcano is added to your vocabulary and you can combine volcano and air and you might get, I dunno, tornado or something. I dunno. But it's just a fun that's all there is. It's just you combine words, it creates a new word. The new word gets added to your vocabulary. You just do this over and this was trending a couple years ago.

Phillip Black:

Wait, hold on. So I, it just, I, there's nothing that happens beyond, I get added to my vocabulary. just like a name generator with a

Eric Guan:

Uh, you wanna play? Let's play it. Let's play it. It's, yeah.

Chris K-S:

I'm on it right now.

Phillip Black:

first

Chris K-S:

if I combine

Phillip Black:

game. Economist, cast, live walkthrough.

Chris K-S:

I have a cloud. What happens if I combine wind with a cloud? I get a storm. Oh, okay.

Eric Guan:

Yeah.

Chris K-S:

This is pretty fun. This is cute.

Eric Guan:

If a game tell if you within five seconds, that's probably a good sign.

Chris K-S:

No, because I'm the type of person who likes something and I'll go super intense on it for 30 minutes and then

Phillip Black:

All right.

Chris K-S:

I can't do this anymore. So we've got a storm, let's make a firestorm.

Eric Guan:

No, just go for it.

Chris K-S:

Should I

Eric Guan:

do you think you, oh, There you go. All, let's cut. You got cloud over

Chris K-S:

I lost my clouds. I.

Eric Guan:

so you can.

Chris K-S:

Okay, let's make moreor. Let's combine Earth with volcano. I just got lava. Okay. I've got lava, I've got volcano, I've got earth. Can I? Okay. I can only ever add. Okay. I feel like I'm just gonna end up with a bunch of magma if I don't. Okay.

Eric Guan:

more creative.

Chris K-S:

and wind. Fox SA is the idea that you

Eric Guan:

doing

Chris K-S:

how a storm in

Eric Guan:

eruption water. I don't know, maybe you'll get like faucet. No.

Chris K-S:

in water. I get lava again. What if I get lava in lava? I get a volcano. What if I do volcano? In volcano? I get a mountain.

Eric Guan:

and wind,

Chris K-S:

Woo.

Eric Guan:

I do it.

Chris K-S:

Oh, I want to do, I was gonna do a cloud, but

Phillip Black:

win?

Eric Guan:

I think there's I don't know if this one has it, but there's certain goals you can work to. So Moreor Chris mentioned Moreor, after the podcast, I would challenge you to try to create Moreor in here. You can, yeah.

Chris K-S:

So is that an outcome that could occur because it's an

Eric Guan:

like Donald Trump, nine 11 in here.

Chris K-S:

dude, they should, they should have like objectives over here, you know what I'm saying? Like they should have get to blah, blah, blah in,

Eric Guan:

Yeah. Yeah. There's a I think, you know, people make up their own version and there's probably clones of this that have done added meta progression, stuff like that. But yeah. This is, oh, snowman. Yeah. See anyway, you can get all sorts of, because it's an LLM, you can generate pretty much anything. It's like kind of that promise of like scribble knots or whatever. You set the snowman on fire, it melts it to water.

Chris K-S:

Shit. Yeah, dude, that did not work in my favor.

Eric Guan:

Uh, yeah, it really is like an open-ended sandbox

Chris K-S:

I thought I'd get like a carrot.

Eric Guan:

Yeah I, I bring it up because I think this is one of those few examples. We talk a lot about AI in terms of oh, it's gonna take away software development jobs or artist jobs. But this is an example where AI is being used as a game mechanic in and of itself. And it can actually create new gameplay and new game designs. As a result. I will say after this came

Phillip Black:

this is

Eric Guan:

This

Phillip Black:

this. Totally convinced me. This is totally convincing. Fucking Eric again. I, this is exactly what I wanted to see. It's such a simple proof of concept.

Eric Guan:

Yeah, I will say I haven't seen anything like this in the last, like year and a half since this came out, but it is pretty dope and I'm sure there's designers out there thickening in this space.

Phillip Black:

It just needs some light, gamification layers. So why do you always drag it on top? Can you make a

Eric Guan:

yeah, you can drag multiple objects on your anything until you

Phillip Black:

what is, oh.

Eric Guan:

yeah.

Phillip Black:

what do you can set up recipes almost.

Eric Guan:

and I'll say the recipes aren't fully deterministic again'cause it's an LLM, but but yeah.

Chris K-S:

dude, I'm convinced everything just results in volcano and laba. That's what it all

Eric Guan:

Chris, let's see if you can get more, do later today. I guarantee you it's possible.

Chris K-S:

I'll send it

Eric Guan:

but yeah, ai, you We're, today we're mostly gonna be talking about it as a productivity tool, like a, capital technology as opposed to a new design. But yeah, this is, I think there's some space here.

Phillip Black:

All right, let's fucking go. AI debate. I published a piece last week and I'm basically taking aim at all what I'm calling the copyright. And there may be some amongst us even today. And my criticism of copyright they assume in a lot of their arguments that AI is copyright infringement. Therefore, claim that artists must be paid royalties. We're never really, it's never really specified how the royalties are to be paid. That's A TBD, that's a big TBD. my problem with this is that it's not clear to me that it's copyright violation. So for instance, if we use the analogy of an artist who trains in the style of another artist, we would never consider that copyright infringement. And as we know, these LLMs cannot produce the original work. But they do train. Is that really that different than a human artist training on a style? They'll be criticism number one. And criticism number two is really a copyright in general, which is that we've almost never had copyright enforced in games. We had this nemesis system which launched and that was patented by WB games and everyone decry that. We don't want to see patents in games. Patents are bad. They inhibit our ability to build on top of each other's innovations. And so we see this the best illustrated in games once again. And think that applies here. I think human beings are gonna create no matter what. So I think I take aim against All altogether.

Chris K-S:

do you see a difference between mechanism copyright and IP copyright? So if I were to. Create a piece of art that's in the exact style of Studio Ghibli. Is that the same type of infringement as me creating a all in hole game where you suck things into your hole? Are those two things the same? Is that the same like wrong? Assuming that we're like trying to disprove that this is wrong, are those two a part of the same universe?

Phillip Black:

For the purposes of this conversation, I would say yes. A process. A and an ip.

Eric Guan:

So what about, uh, Pokemon? Can I, should I just make, if I wanna make a Pokemon fighting game, I can just make it, and Pokemon should, companies should not have the rights to pursue me and say, Hey, you're stealing RIP.

Phillip Black:

So what I would say in situations like that is that I think consumers will crave the original content. Like it's not just using the characters, it's other things like we know they're like seals of approval, between something that's authentic and educated and original and something that's not. so I think that while that content will exist, the original pieces of content will still exist and still continue to be profitable, even in the face of it, just like we see with

Eric Guan:

So you're saying no, that It, they should not be able to protect the Pokemon intellectual property. Uh, what about consumer confusion? This is the classic trademark, counterfeit issue. What about consumers who think that they're buying an authentic Pokemon game, but they're buying, Eric's Pokemon knockoff?

Phillip Black:

S So I would like point to like sneakers, like this is a big thing in like the world of sneakers, if you're familiar with selling Air Jordans, is that there's a lot of counterfeits. so this is like given rise to all these third party marketplaces where you send the shoes to them first they authenticate it and then they resell it. And part of it is they have the buyer guarantee because they authenticated originally. And so I think the market just solves problems like this pretty naturally. And so I don't, I think

Eric Guan:

So, like a radical, like anti-intellectual property stance.

Phillip Black:

So I, I used to be, I used to be, when I was on Team econ team hardcore U Chicago econ, I was, I heard this argument before and I was like, no fucking way. Like it reduces the incentive to create, right? That is the huge problem with getting rid of this. You can't be that stupid. I think that when you balance that against the success of building on top, that happens in innovation, like it happens in trees that you have one innovation, which is based on another innovation that when you slow down that process because you have copyright, that is such a huge deterrent to growth, that it overwhelms the incentive effect. Like ultimately, human beings always create like it's pretty inelastic. in demand or quantity supplied from a change in price. I would argue.

Chris K-S:

So I, I think that. The best way to approach this is from damages. Like a cla If you were litigating this, you would just look at like the damages that were dealt to Nintendo or Pokemon Company, and you would evaluate oh yeah, how much do you owe them? For copyright infringement. This is like kind of the Pareto efficiency or Pareto model where you like try to solve such that every, nobody could be made better off. And I think that the problem with. So that's how I think you would be able to say whether it's good or bad is what would the benefit, what would the overall social welfare of the economy be if there was no copyright laws versus if there are copyright law, right laws, and it's not really a counterfactual you can actually solve or run. So I think that it doesn't hurt Pokemon, it doesn't hurt studio Ojibwe. It doesn't hurt. These incumbents, but does it hurt, Eric's small indie company who's trying to produce, Pokemon number, two, and I think that it's less obvious when you're talking about a non-incumbent that doesn't have this brand moat

Eric Guan:

The, can I frame the sort of the, the intellectual property thing from the eco the thing you alluded to earlier, Phil, which is. Is there's this explore, exploit, trade off, basically in intellectual property development. So you think of when I say that it means like everything from patents to new game mechanics to art inventor creates an invention. I. And to get the most social value out of that invention, you want as many people to have access to it as possible, as many people to build on top of it and utilize it as possible and as freely as possible. However, you still need that invention to be created in the first place. And the incentive argument is that, the inventor will create the invention if they know they can profit from it. But if the inventor knows, they'll create an invention and then immediately someone else is gonna steal all the profit and they're gonna lose money from it. They won't create the invention in the first place. And so the copyright argument says that you should give the inventor some. Temporary monopoly control. Some ways they can extract revenue from it before it gets released to the public. So that they continue inventing. And this is, I think, the crux of the intellectual property debate. And as Chris mentioned earlier, it, it varies by sector to sector. So for example, for game mechanics, it seems like we're extremely low on the, the, that incentive to create. But in a super high r and d industry, like pharmaceuticals, like the upfront r and d is and massive, but the actual marginal cost of production is negligible. And so they have extremely strong cop intellectual property protections there. And and I think. This really boils down to those two things, like how much do you need to incentivize additional production versus how much do you want to free up access so that as many people can benefit as possible.

Chris K-S:

And if you just look at the way it's handled in the US like we do that we punish people for, I just watched the social network for the first time

Phillip Black:

where you

Eric Guan:

Did that come

Phillip Black:

have you been, Chris, for the last fucking decade.

Chris K-S:

I can't believe that movie came out 15 years

Phillip Black:

film. The film of our generation, I would argue.

Chris K-S:

It was really good. And I only realized how old it was when they showed the market cap of Facebook At the end, it was like, Facebook is now worth 25 billion.

Eric Guan:

Nothing.

Phillip Black:

child's

Chris K-S:

like, holy shit. When was this made? I wish I'd seen this movie. I would've invested. No. It's like the Winklevoss twins, they're all pissed off that their idea was stolen and they did get compensated for it. And I think that was. I think that was right. Yeah. And then they went and bought Bitcoin and they're fucking 100 billionaires now. So that seems like a just system to me. And I disagree with this idea that an LLM is no different than an artist honing their skills. Think that you do need to get into the philosophical discussion on whether or not LMS are conscious. I don't think that they have the ability to actually think and create the way that humans are. I don't think that LMS are humans. I don't think that AI is human. And so a human creating is fundamentally different than an AI creating. If I use, we talked about this a couple days ago. If I take a picture of the Mona Lisa, I did not create that image. I didn't create the Mona Lisa. I took a picture of it using a piece of technology. It's a tool that allows me to imitate reality. That's no different than an ai. It's a piece of equipment that somebody built, an engineer built that emulates the stuff that you're, the context that you are putting it in. That is not creation. I'm not creating the picture of the Mona Lisa. In my opinion. And so I, I just think that it's, I think that there's a fundamental flaw in saying that LMS are creating and iterating in the same way that a human does. A human that goes and trains for 10 years honing skills, trying to understand how can a classic example is Hermes, they're basically, they trained for 10 years so that they can perfectly emulate the previous craftspeople so that they can, produce this thing. That's a very different. In my opinion, those that is not the equivalent to an ai being trained to create a bag. And indeed, I think we get into the bigger discu, or we also touch the, this topic of the creator matters. The person who created the thing matters. I want a Hermes bag that was created by a, an artisan. I don't want an air mass bag that was created by some machine that stitched it together.

Phillip Black:

I agree. I agree with all these frameworks. Like I think this makes sense. I think we're after the same thing, which is like, what is the profit maximizing framework?

Eric Guan:

Crazy thing you said to me is that. production is totally price and elastics, you said, right? That you're saying people are gonna create stuff no matter the profit motive, which just sounds non

Phillip Black:

in inelastic. I wouldn't say perfectly inelastic. Oh, I can't argue that means I wouldn't be at the margin, but I would say like relatively inelastic. I think human beings will, like, let's think about it this way. What does copyright 70 years right now? What do we what would we bet the change? The change in relative output is? If we brought it down to, let's say, 40, that's a pretty radical

Eric Guan:

there's been regulatory capture and corruption and like we can get into that more, but

Phillip Black:

So what do you think the optimal number is

Eric Guan:

I give it like 20 years.

Phillip Black:

I dunno come around to

Chris K-S:

dude, I would actually say five

Phillip Black:

Oh.

Chris K-S:

If you can't, if you can't get yourself a moat and you can't build the brand and you can't basically what we want to avoid is like intellectual

Eric Guan:

like okay.

Chris K-S:

I. Yes. Yeah. So we want you to be able to build a brand like Pokemon before somebody else can come and steal it. I think that's what we know that nobody's gonna be able to come in and compete with Pokemon in creating Pokemon, right? Pokemon is the best at doing that. What we're trying to avoid is a situation where some unknown artist gets their, they're art stolen or something like that. And I'm of the belief that if you're not able to create a brand for yourself, Such that somebody can't come in and do what you're doing better after five years. If we were to choose a time period, I don't understand why we need these 40 year time horizons.

Eric Guan:

That there's been a regulatory capture, and Disney famously keeps extending the copyright deadline. Keep keeping Mickey Mouse, under their intellectual property protection. At this point, Mickey Mouse is like almost a hundred years old. And famously Steamboat Willie went into public domain a few years ago. But they finagled it so that Steamboat Willie is not the same as Mickey Mouse. And so Steamboat Willie into pub went into public domain, but Disney still owns Mickey Mouse. And it's, I dunno, it's deeply hypocritical.'cause they built their empire off of public domain ips snow White and Cinderella, right? And Sleeping Beauty. And now they're like, oh, no one else is allowed to touch the stuff we made. That, Hans Christian Anderson or whoever the fuck made ages ago,

Chris K-S:

would the world be better off if they didn't have these 100 year long protections? Would Disney be better off or

Eric Guan:

I I think we're all on board that it's too long and that the would be better off if, especially at least art and literature, intellectual property was weaker, much weaker than it is today.

Phillip Black:

And I'm not as convinced that Disney would be worse off, as you may imagine, Chris, like I, I think it's a little bit harder in the entertainment sector, but if we imagine like technology and getting rid of patents. Another company having a breakthrough again because you had a patent, because they stood on top of your patent, comes back and people were cursive value back to you. So there's a lot of positive externalities that happen in these exchanges too. All the time.

Chris K-S:

Oh, Ima like pharmaceuticals, It's imagine if we had like unlimited supply of certain pharmaceuticals that are patented and only allowed to be produced by specific government issued

Phillip Black:

like, I, I would argue like very little from what I think we understand from economic research of the total value of an invention goes back to the creator. Like it's a very small percentage of total revenue of their invention. I, what I'm saying is I think if you guys think the incentive if you're skeptical of my inelastic supply argument, they should capture a lot more of their value.

Eric Guan:

Wait, how?

Phillip Black:

A different schema. I would argue that if you think that the supply is is very elastic, like we wanna know what those effects are.

Chris K-S:

I think that like founders in Silicon Valley and VCs, the people who are really generating, I don't know. We're not, I'm not talking about art, but those people are pretty inelastic. If the incentives were to decrease, I,

Phillip Black:

I'm saying like if we're only, let's say we're only giving them 3% of the GDP, they create. I think if you guys are true copywriters and you think this supply side effect is large, I think you want to give'em 10% because they're gonna make so much more shit if you give'em 10% of GDP rather than five. So my, I throw it back

Eric Guan:

Sorry, can can you say that again? I didn't, I didn't

Phillip Black:

you guys should increase it. Argument that I was making previously is that the supply side response to a change in the expected wage of the copyright issue, like me creating an IP is not gonna do much to the total supply of IP that's created. That's my argument'cause it's relatively inelastic supply curve. But if you disagree with me and you think the supply curve is elastic, what I'm proposing is that we should actually extend the copyright protection because we will get even more

Eric Guan:

so I think there's an optimal point, right? Like I think 70 years on copyright for art is way too long. But I think zero is probably there's no way, the optimal point is at zero for every industry, right? And I think it depends a lot on, the, how much upfront investment is required. I can use pharmaceuticals as an example, Because the ratio and the payoff horizons are so much longer and it's more skewed. They need more protection than game mechanics, for example, might be much quicker. Like you make a hit game, it's popular for a year, you've captured most of your revenue in that first year before someone else can copy it. maybe Don't need protection for that reason. You're talking about spike in decay, curves Revenue earlier. Yeah.

Chris K-S:

I think that there's probably like a reasonable. Midpoint in terms of protections. I reject the idea that LLMs are no different than that creation process is the same as a student iterating on a master's work or something like that. But, I think that there's a bigger, there's probably a bigger issue with royalty payouts, especially in the context of ai in that there's no way to properly attribute the output that an AI gives to, or even the revenue that an ai produces to an individual contributor without just straight up doing some sort of a split share. Hey, we trained on, 10% of the data we trained on is from Studio Ghibli, and we made a million dollars. Here's a hundred thousand dollars or something like

Phillip Black:

I this as a problem. this is a problem in almost every sort of like network thing or something where there are multiple factors of production. This is a problem in mixed media mod modeling. For instance, when you have people who like May, they may have seen an ad on tv, I. And then they purchase something. How much do you attribute to the ad by raising brand awareness versus them seeing like maybe a click through another I, very complicated stuff. And also like on UGC projects, sometimes people will use other people's pre-made objects and they'll add on their own things. And people that I've. Talk to you about the problem of trying to figure out whether or not we should allow that and what the complication should be, which by the way, makes this a very real policy. This is not just like a pie in the sky thing, guys. Like we could actually run experiments in UGC games where we set different institutional policies, but whether or not you're forced to share, and by the way, some UGC platforms do this, like high pipe, the one out of Finland that's doing like tech TikTok for mobile, you by default have to have remixes or sharing open. So other people can copy all of your work.

Eric Guan:

Is there any required.

Phillip Black:

and yet it's

Eric Guan:

Okay. So it's just everything is public domain all the time, basically.

Phillip Black:

So it does have the effect that you mentioned a really long time ago, Eric with Fortnite, which is that like people cloned it and almost was like voter median theory, but because there's CCU problems and a lot of other things, everyone flocks to one version.

Eric Guan:

Yeah,

Phillip Black:

there doesn't end up being like all these copies that survive.

Eric Guan:

tell that story

Phillip Black:

that horror story doesn't

Eric Guan:

Uh, smart ping. I like that one.

Phillip Black:

Oh, and then, yeah, that was another one. So to give you an example, apex Legends had a ping system where if you had your mouse over an object, it would recognize that it was an object and you could click a little mouse button and it would say to your teammates, Hey, there's this object here. So you could tell them go get it or do something. You could ping it. And Fortnite, after that feature launched, apex had that in their game, and I think about two weeks we started to see an early version of it. And so that didn't seem to slow innovation either in games, again, like more people just got all of this really great innovation. It was dispersed

Eric Guan:

so I think the the positive version of the story, the anti copyright version, is like, Hey, this innovation, everyone is able to use it. All games are better. And that's awesome. the flip side You see this fast follow strategy where people are like, Hey, r and d is hard. Let's just watch other people and fast follow what they do. And if there's, if the in innovate versus copy incentives are skewed too much, you see a lot of fast follow behavior. And I think we do see that.

Chris K-S:

I am I, not to put you guys on the spot, but can we think of like an example of fast following where the imitator ended up? Getting the better end of the stick than the, innovator Fortnite. Yeah.

Phillip Black:

PG. That was a really great fucking example, Eric.

Eric Guan:

League of Legends, doda team fight tactics, Doda There's like dozens of these. Yeah.

Chris K-S:

yeah. But would

Eric Guan:

Yeah. I, I think

Chris K-S:

much worse off if Fortnite hadn't come in?

Phillip Black:

I think PUBG is better

Chris K-S:

Yeah, and I guess that's. That's what I'm saying I'm not saying that there wasn't an imitator that was successful. I'm saying that there was an imitator that killed the innovator and you might even argue because Fortnite popularized or popularized that mechanic that, like battle royale, maybe they actually brought people into PUG. So I guess so I think that's a great example, but I think it's actually in favor of Low, no barriers.

Eric Guan:

I'm happy with how things turned out. So I guess going back to the original statement, do we think the games industry needs more or less inte intellectual property protection? I actually, I signed with Chris I signed with Phil on this, which is I like the fact that game mechanics aren't copyrighted. I like the fact that these games can see each other and innovate off of each other. And yeah I do think games are better off for it. Yeah.

Chris K-S:

I'm on Phil's

Phillip Black:

Oh my Jesus. What? What the

Chris K-S:

we're all, I think we're all like, I think we're I, my to be clear, Phil, my main issue was that you were trying to say that a machine is a human. And I've been saying this for honestly since before AI was a big deal. I worked with an AI person a couple of years ago and they were like, there's no difference between a machine and a human. They're the same. And I was like, that is. Bad shit. I still don't have a super compelling argument for why, but I just,

Phillip Black:

I I, I think like that's the problem is that we got the naming wrong. To me, it's not artificial intelligence, it's aggregated intelligence, like it's the best versions of ourselves. Like it's been trained like a great master on thousands of years of human history. It is the most human thing we could possibly imagine.

Eric Guan:

I would call it super human.

Phillip Black:

But I would say that's the struggle, Chris, is like when people don't like that argument, I'm like where does the analogy fall apart?

Eric Guan:

so one point I wanna get into is so let's say we actually did want to implement intellectual property protection in games. I think there's actually a huge enforcement problem. One is like the attribution issue that you guys are talking about. Like even if you wanted to say, this artist gets this amount of credit for the AI generated art, like that problem itself is really hard. And also just. like regulatory capture and corruption like you guys know about this inhaler cap lawsuit. So pharmaceuticals, They invent some drug and then it gets a certain amount of lifetime as a patent and then it goes into public domain and you get generics. But pharmaceuticals have been doing all sorts of tricks to extend patent deadlines. There was an inhaler company, they made some inhaler like, a hundred years ago, whatever, that helps with asthma. They invented. Air quotes a special kind of cap on the inhaler. There's like a little button you press that lets it pop off more easily. And they use that little plastic cap as justification for extending their patent and continuing their, like monopoly status on this thing, preventing it from going public domain. And Pharmac schools do all sorts of shit like this where they. Do dumb ass remixes claiming it's a new invention and extending patents. There's a big FTC lawsuit about this. But the point being even if you have a well-intentioned regulatory apparatus, there's like high chance that it gets corrupted, gets captured, gets abused somehow. And so even if you intended to have, like x 5%, copyright protection, it ends up getting abused into like much higher.

Chris K-S:

Yeah. So I think that's interesting in terms of the enforcement on the enforcement level from like a political and legal landscape. Yeah. I think it's a mess, an absolute mess. Convincing judges and politicians that this is stuff that's important that they should rule, in, in favor of the, looser restrictions or paying the, the royalties, whatever it might be. But there's also just like a pure technical limitation in Unenforceability. Like, how the fuck do we do not know what the, all the data that was used. And Chris and our in our game Economist cast discord channel pointed this out. Do we even know like what Chatt PT has been trained on at this point? We're talking about billions and billions of parameters, which is probably trill hundreds of billions of data points, if not, hitting the trillions. It's first of all, we'd have to go back and figure out what these things were trained on in the first place. Second of all, like how do you regulate what people are using unless you have like very strict government regulations on the data sets that, the databases that these guys are pulling from. I. There's no centralized market for this data. There are many markets for data. AWS has a marketplace for data. You can subscribe to a data stream and literally just have that data stream come into your environment in real time, every single day, train your model on that. But that's not necessarily where all of the traffic comes from. So it's technically there's no real way to figure out what data was used in the first place. It's obvious in the case of chat, GPT using Studio Oji stuff because it's okay, they were very overt about it. It's obviously using cha studio Oji bli. It wasn't trained on anything else because it's like not coming up with stuff that looks anything other than studio Oji, bli. But there's a lot of cases that are not as clean and cut. And I could not, I would not be able to investigate or figure out what chat two PT has used for their 3.5 model or their 4.0 model. It's like, how do you attribute.

Eric Guan:

I think there's like a laundering component here where imagine, like you had to pay for every Studio Ghibli image you fed into the machine. But you could train one machine on some set of Studio Ghibli. You pay the upfront cost and then you have the Studio Ghibli trained one to generate new images. And then you just bootstrap yourself by training on those Studio Ghibli copyrighted things, but not produced by Studio Ghibli. So there's a whole enforcement issue.

Phillip Black:

This is a synthetic data point that like this other guy was making about, like you would never be able to copy, you would never really be able to make Studio Ghibli originally if you didn't train on fan art. And fan art in and of itself is a copyright violation, but you could still, I'm sure, argue that we could get, aI to make something very close to Studio Ghibli, again, by an artist or someone customizing or prompting it in such an extreme way that it would produce it, in which case you now have synthetic data that you can train it on, right? So now you gotta figure out, okay, is the fucking synthetic data copyright violation?

Chris K-S:

That's why I don't think you can do it on, so I almost see there are two way, two places where you can employ the, or calculate the tax. You can either do it on the input side before it's gone into the model, or you can do it on the output side. Basically, Hey, you made$10 million this year. You're supposed to pay 10% out to your contributors in royalties. So you pay$1 million split across all of your people. That's like the output side. Input side, Do you pay per row of data and then you've gotta, you've gotta price that somehow. And the problem with this data is Studio Ghibli, and I still dunno if it's Ghibli or Ghibli Studio Ghibli doesn't they don't like. Produce content 24 7. They have these big spikes. Okay, here's a million frames because we just came out with a new movie. Alright, three years later, here's a million frames. So it's like, how do you price that? So it's input and output. There's just no clear way that you could possibly price this stuff.

Phillip Black:

I agree. All right. Do you wanna wrap it up? Sounds good. I'll talk soon guys.

People on this episode

Head Shot

Phillip Black

Host
Head Shot

Chris Kaczmarczyk-Smith

Co-host
Head Shot

Eric Guam

Co-host