We will have to accept AI-generated content

00:00:00: Here's a hot take, and I'll explain it.

00:00:02: I think we humans think that we like

00:00:06: content that was generated by humans,

00:00:09: that we prefer that over AI-generated content,

00:00:12: but I think that is increasingly wrong.

00:00:16: And I'll explain it before you start flaming me. But I had this

00:00:20: thought for the last couple of weeks and months, to be honest, and

00:00:24: then I encountered this post here by Tim Sweeney,

00:00:27: the founder and majority owner of Epic, the CEO of Epic Games,

00:00:32: who recently posted on X. We have a huge pivot underway here at Epic,

00:00:37: moving developers around to retool for maximum AI acceleration.

00:00:42: The aim isn't to do the same stuff with fewer people,

00:00:45: but to enable everyone here to build more and better stuff, and to bring

00:00:50: the benefits to all teams using Unreal, the Unreal Engine, that is.

00:00:55: Now, naturally, DHH was quite happy about that,

00:00:58: and I also agree here. But this is, in the end,

00:01:03: you could say a relatively standard statement, right?

00:01:05: A lot of companies are embracing AI.

00:01:08: I think a lot of layoffs have been announced

00:01:13: due to AI, and obviously, that is a bad thing and not what I like.

00:01:18: But what's interesting about this statement, therefore, well, for one,

00:01:20: especially that it's not about layoffs, but also that

00:01:24: it's from a company, a games company, which is of course

00:01:27: all about creativity and the content, the games they build,

00:01:33: and naturally, one of the most popular reactions here

00:01:37: is this one at least for me when I view this post:

00:01:41: If Unreal becomes an engine where you upload a few pages outlining

00:01:46: a game idea, and five minutes later you get a fully playable AAA

00:01:50: fidelity game, do you think this would be a good thing for humanity?

00:01:54: And obviously, this is a pretty critical comment.

00:01:57: And there are more comments like this now. For one,

00:02:01: in his post, it's not about giving you something where you just upload

00:02:06: a draft and get a finished game, but clearly, of course, the goal is to

00:02:11: develop the Unreal Engine into a direction where you can use it,

00:02:15: for example, to vibe code games, if you want to put it like this.

00:02:19: And you can absolutely discuss whether these would be

00:02:23: the games you want to play or not. But here's the thing.

00:02:26: As I mentioned at the very beginning, I think we humans like

00:02:28: to think that we want to consume content,

00:02:32: no matter which content—if that's videos,

00:02:34: if that's games, if that's movies, text, music,

00:02:38: created by other humans—and I would have agreed to

00:02:42: that statement until a couple of months ago, at least.

00:02:46: And I'm still not sure if I entirely disagree,

00:02:49: but I think that we overestimate how important this

00:02:54: really is to us humans, and of course, just to be very clear here,

00:02:59: that is not a general thing for you. It may be one hundred

00:03:02: percent clear that you want human content only, and that is fine.

00:03:07: I just think that throughout the ages,

00:03:10: how things are built and who builds something has

00:03:14: changed so many times that it's pretty dangerous

00:03:18: to keep a hard opinion on this AI content

00:03:23: thing, for example, you may have said, let's say,

00:03:27: 150 years ago,

00:03:29: that you, or 100 years ago,

00:03:30: even that you absolutely only want to wear clothing created

00:03:36: by a craftsman who really knows their craft,

00:03:40: handcrafted in your country, maybe in your city.

00:03:44: You want to have shoes that were

00:03:45: created by an artisan in your city.

00:03:49: Nowadays, we buy whatever is cheapest,

00:03:52: or most people buy what's cheapest, or at least which brand they want,

00:03:56: and they don't care if it's manufactured in China, Bangladesh,

00:04:01: Italy or somewhere else. Some people care, one hundred percent.

00:04:04: Some people care and they pay a premium,

00:04:07: but of course not everybody can afford that premium.

00:04:09: A lot of people don't care or can't afford it.

00:04:13: And you can have philosophical discussions about this:

00:04:15: Is this a good thing? Would they care if they could afford it?

00:04:19: Maybe. I just think that with AI, it's kind of similar.

00:04:24: And for example, if we take text, which obviously was the first

00:04:27: form of AI that entered the mainstream with the release of ChatGPT,

00:04:32: it's probably fair to assume that a huge chunk of text we

00:04:36: consume these days, at least on the internet, but also

00:04:40: outside of the internet increasingly,

00:04:42: is AI-generated or at least,

00:04:44: to huge parts AI-generated.

00:04:48: If you're reading a newspaper article where the first draft was

00:04:52: created by an AI and which was then polished up by a human,

00:04:56: is that human generated? Is it AI-generated?

00:04:59: You can make both arguments here.

00:05:01: And I would definitely make the argument that if a human

00:05:05: put a lot of thought and effort into having the AI

00:05:09: generate a great article, it is a creative product.

00:05:13: It's a product of that human, even if AI was used.

00:05:17: A not perfect example is that if you're using a word

00:05:21: processor or a tool like Microsoft Word for writing,

00:05:25: obviously it's still written by you,

00:05:28: even though you did it on a computer. Now, it's not a perfect example

00:05:32: because clearly you were typing every character, but you used a tool

00:05:35: to help you to format it in a nicer way to make it more readable.

00:05:39: You did not write it down by hand. Again, not a perfect example, I know,

00:05:44: but AI is just another tool here—a tool that can help you get

00:05:49: that draft based on the information you collected, based on the

00:05:53: outline you prepared, maybe also with help of AI. But still,

00:05:57: it's your creativity and it's your decisions that go into there. Now,

00:06:01: obviously, that's not true for every piece of text.

00:06:04: There is a lot of slop out there,

00:06:07: and that probably makes it so hard to differentiate in many cases.

00:06:11: With AI, it's super easy to produce low quality

00:06:15: products, slop, low quality text, low quality movies.

00:06:20: It's super easy. And nobody, I would guess, likes that, myself included.

00:06:25: That's not the kind of content I'm talking about here.

00:06:28: I'm talking about content generated by

00:06:32: AI or with help of AI, where, of course,

00:06:35: you still have the creativity and the decisions of the humans.

00:06:39: And I think we need to be more open,

00:06:41: or simply be prepared for a future where this will be the norm.

00:06:46: And now back to this post here by Tim,

00:06:49: where he turns Epic and the Unreal Engine into a

00:06:54: tool that heavily embraces AI, and as I understand it,

00:06:58: gives developers that work with the Unreal Engine tools

00:07:03: to have an easier time leveraging AI to build their

00:07:06: games by building certain AI features into the engine.

00:07:10: And I think that's the only way forward.

00:07:13: We will not have a future where games are being built

00:07:18: by people handcrafting 3D models in Blender, for example. Of course,

00:07:23: yes, there will be some exceptions here and there, and they may be

00:07:26: successful just because of the fact that they are handcrafted, though.

00:07:31: If the game sucks, it will still not be super successful.

00:07:34: But the majority of games, either way, will leverage these tools.

00:07:39: I mean, today already, every serious game developer and game

00:07:44: studio is, of course, using AI for code assistance for generating code,

00:07:49: like every other developer is or should be doing at least.

00:07:52: Of course, that will spill over to art. Now, again,

00:07:56: here's the important point: you can use it to have generic art assets,

00:08:01: to have slop, to have 3D

00:08:03: models with lots of little errors. You can do all that, or you use it

00:08:08: to turn your creative vision into prototypes,

00:08:12: which you can then polish up and bring into your game. You can

00:08:16: use it to quickly explore different art directions by quickly

00:08:21: spitting out different prototypes and mockups, just as a developer.

00:08:25: Where you can also quickly create a bunch of mockups of a website, and

00:08:29: then you can see those different variations and iterate on that.

00:08:34: You can use it as a tool, but of course,

00:08:37: the result naturally will be AI-generated,

00:08:41: since large portions of it are AI-generated. But it's still

00:08:45: the artist, the developer, that poured their knowledge,

00:08:49: their expertise, and their decisions into that

00:08:53: AI usage to get a great result out of it.

00:08:56: And nowadays, I'm pretty sure that is what we'll see,

00:08:59: and what we'll ultimately accept. What we'll have to accept

00:09:03: that we'll have music, we'll have movies, we'll have games,

00:09:06: we'll obviously have software. We already have it. We'll have text.

00:09:10: We already have that too. All that generated by AI,

00:09:14: but then we'll have shitty versions, which are just slop, vibe-coded.

00:09:18: Which will simply not win

00:09:20: in the market on the long term,

00:09:22: and will have great results where it's not just

00:09:26: one prompt and let's see what we get as a result.

00:09:29: But I think we need to be open to say,

00:09:31: yeah, it's good that you're using AI.

00:09:34: Simply for the simple fact that it allows you to stay in business. If

00:09:38: you have a favorite game studio and you want them to stay in business,

00:09:42: they need to use AI because all their competitors are moving faster.

00:09:46: So that alone is a good reason. But that's not the only reason.

00:09:50: AI also allows you to be more ambitious,

00:09:53: and I covered this in other episodes already.

00:09:55: But it allows you to iterate quickly to explore,

00:09:58: as I mentioned, different artistic styles, and I think

00:10:02: these are all the positive sides we can see

00:10:05: here when it comes to AI for art and so on.

00:10:08: Now, of course, there

00:10:09: are many things that are to be criticized about AI,

00:10:13: or that can be criticized, and things you can be afraid of.

00:10:16: Now, for one, for example, the entire copyright

00:10:18: thing—how these AI models were trained and are trained.

00:10:22: Horrible! Lots of copyright infringements here.

00:10:26: At least, I guess that's what many people, including myself, feel.

00:10:31: Even though it may not be the legal side, or there may not

00:10:34: be a final judgment there, but obviously that is not great.

00:10:38: Obviously, you can be afraid.

00:10:40: We're all feeling this uncertainty because things are moving so quickly.

00:10:45: But that does not change that it's here, and that it can,

00:10:49: and that it should be leveraged to build better products.

00:10:53: And still, right now, we're in this chaotic phase.

00:10:55: We have a lot of shitty software, a lot of shitty products

00:10:58: because we're still all figuring out how to use these tools.

00:11:02: And frankly, because these tools,

00:11:05: they are becoming much better, and you can now create

00:11:08: better results with them than you could a year ago.

00:11:11: So combine that with that we're still all figuring out part,

00:11:14: and here you are in the current situation. I'm convinced great products,

00:11:18: no matter about which kind of product we're talking,

00:11:21: will come out because of AI. That includes all sorts of content,

00:11:25: including videos. And no, this does not mean that you will

00:11:29: only have AI-generated YouTube videos or something like that.

00:11:33: But you can have amazing AI-generated

00:11:36: YouTube videos, which are not the slop.

00:11:39: You can have amazing AI-generated movies.

00:11:41: And the slop part,

00:11:43: I think, will simply lose

00:11:45: on the global market. So,

00:11:47: yeah, these are just my two cents on this topic. I guess.