探索不可思议的边缘：从早期的“AI鬼畜”到AI垃圾的担忧
我一直被那些不可思议、甚至带有一点诡异（uncanny）的事物所吸引。我制作过关于蜘蛛在网上弹奏爵士乐的电影，或者关于蟑螂建造火箭飞往月球探索宇宙的电影。我喜欢在大多数人觉得有点诡异的事物中寻找人性。因此，当最初的AI图像出现时，我并没有感到害怕，而是感到非常好奇：这能开启怎样的电影制作新方式？我开了一个YouTube频道来一探究竟。
我早期的一些实验是用最早的AI视频模型之一生成的自然纪录片。那个模型太早期了，以至于它甚至无法记住半秒钟前发生的事情，这就是为什么它看起来像“磕了药的地球”——我是想象的，我的意思是，我其实不知道（笑声）。不过，我最喜欢的早期项目是使用当时最前沿的AI工具来生成故事、3D模型甚至动画，以此为我最喜欢的皮克斯电影《料理鼠王》制作续集。结果它变成了这个样子：（视频）“古斯特总是说，人人皆可烹饪，但只有无畏者才能成为伟人。（戏剧性音乐）哦！救命！帮帮我，雷米！（尖叫声）（邪恶的笑声）” 麦克·马勒：所以，是的（笑声）。我知道这不完全是好莱坞水平，但我真的很喜欢把这些工具推向极限，看看它们会在哪里崩溃。正如你们所看到的，它们崩溃得相当快。
然而，随着技术演进，我们必须面对技术成熟带来的新问题。如今AI几乎可以制造出我能想象到的任何东西，而且它变得一天比一天更好、更容易使用。这很神奇，但也有点可怕，对吧？我的意思是，当一切都变得可能时，会发生什么？现在任何人都能直接通过提示词生成一部好莱坞大片吗？还是我们会淹没在无意义的AI生成内容的海洋中——也就是我们现在所说的“AI垃圾”（AI slop: 指大量低成本、低质量且缺乏真实意图的生成式AI内容）？
Original English
I’ve always been drawn to the uncanny -- the slightly creepy. I make movies about spiders playing jazz on their webs, or cockroaches building a rocket to fly to the moon and explore the universe. I love finding the human in what most people find a bit creepy. So when the first AI images like this appeared, I was not creeped out. I was really curious. What new ways of filmmaking could this unlock? I started a YouTube channel to find out. Some of my early experiments were nature documentaries generated using one of the earliest AI-video models. It was so early that it couldn't even remember what happened half a second ago, which is why it looks like planet Earth on acid -- I imagine, I mean, I wouldn’t know. (Laughter) My favorite early project, though, was using, back at the time, cutting-edge AI tools to generate the story, the 3D models and even the animation to create a sequel to my favorite Pixar movie, "Ratatouille," of course. And it became whatever this is: (Video) Gusteau always said that anyone can cook, but only the fearless can be great. (Dramatic music) Oh! Help! Help me, Remy! (Screaming) (Evil cackle) Mick Mahler: So, yeah. (Laughter) Not exactly Hollywood, I know, but I really loved pushing these tools to their limits to see where they'd break. And as you can see -- they broke pretty fast. But now AI can make almost anything I can imagine, and it’s getting better and easier to use every single day. That’s amazing, but also kind of terrifying, right? I mean, what happens when everything becomes possible? Will anyone be able to just prompt a Hollywood blockbuster now? Or will we drown in an ocean of meaningless AI-generated content -- what we now call “AI slop.”
故事重于工具：用创意与现实链接抵御算法虚无
对于这种担忧，我其实更加乐观。为了展示原因，我带来了三个真正启发了我的项目。首先是一个关于声音与动画结合的尝试。当一个新的AI模型发布，能够将音频片段转化为动画时，来自柏林的电影制作人走上街头，采访路人关于他们食用肉类的情况，然后他们使用这个AI模型将路人的声音转化成了这个：（视频）“我想成为素食主义者。这影响了我的健康。但是，是的，吃肉确实让我感觉更好。” 外星人2：“我们需要肉，因为我们的大脑。肉里含有的那些东西正在帮助大脑。”（笑声）
现在，很多人都在用AI制作这些夸张的科幻电影预告片，里面有爆炸之类的场面。我理解这一点，那很有趣。但这个项目之所以让我脱颖而出，是因为它并非如此。它植根于现实。这是一个带有幽默反转的简单想法。最重要的是，它讲了一个故事。这也是我喜欢处理我的AI项目的方式。
这种将技术与个人体验结合的方法，在我自己面临困境时提供了新的出口。今年早些时候，因为整天坐在电脑前测试所有这些新的AI模型，我感到非常职业倦怠。我开始觉得有点像卡夫卡《变形记》里的格里高尔·萨姆沙。你知道，就是那本书里变成甲虫怪物的人。受到这个灵感的启发，我走到户外拍摄了真实的甲虫，然后使用这些照片训练了我自己的这只甲虫的AI角色模型（AI Character Model: 用于在图像或视频中保持特定角色外观一致性的定制化AI系统）。接着，我为这个故事写了一个替代结局，在这个结局中，格里高尔并没有孤单悲惨地死去，而是重新与大自然建立了联系，并学到了生活不仅仅是工作。然后我们就去森林里用iPhone拍摄了这部短片（笑声）。通过使用这个甲虫模型，我只需点击几下就能将自己变成格里高尔（笑声）。
Original English
I'm actually more optimistic. And to show you why, I brought three projects that really inspired me. When a new AI model was released that could transform audio clips into animation, filmmakers from Berlin went out into the streets and interviewed pedestrians about their meat consumption, and then they used this AI model to transform their voices into this: (Video) I’d like to be a vegetarian. It affects my health. But yeah, I do feel better when I eat it. Alien 2: We need the meat because of our brains. Those things that are coming from the meat is helping the brains. (Laughter) Now, a lot of people are creating these over-the-top sci-fi movie trailers with explosions and stuff with AI. And I get that. That’s a lot of fun. But this stood out to me because it's not that. It's rooted in reality. It's a simple idea with a funny twist. And most importantly, it tells a story. That's how I like to approach my AI projects, too. Earlier this year, I felt really burnt out from sitting at my computer all day long, testing all these new AI models. I started to feel a bit like Gregor Samsa from Kafka's "The Metamorphosis." You know, the book about the guy who turns into this beetle creature. Well, inspired by that, I went outside and photographed real beetles, and then used these photos to train my own AI character model of this beetle. I then wrote an alternative ending for the story where Gregor, instead of dying alone miserably, reconnects with nature and learns that there's more to life than work. We then just went out into the woods and filmed this on an iPhone. (Laughter) And using this beetle model, I was able to transform myself into Gregor with just a few clicks. (Laughter)
拥抱真实与独创：在无限制的技术洪流中破局
与此同时，人们现在开始在笔记本电脑上制作整部电影。有一个不断壮大的电影制作人社区正在推动AI的边界，对他们来说，分享他们是如何制作的，与最终的电影同样重要。我认为这种开放态度非常酷。它真正邀请了任何人加入进来，开始制作自己的电影。
在这股大众创作的浪潮背后，我们正在迎来一个技术无门槛的时代。很快，任何人都能讲述任何故事，而不受任何技术限制。那么，当一切都变得可能时，会发生什么？我想，是的，我们将看到令人惊叹的新美学、我们从未想象过的新叙事形式，以及那些以前找不到观众的人所讲述的故事。但我们也将看到——而且我们已经看到了——更多的垃圾：为了降低成本而制作的低水平AI内容、深伪技术（Deepfake: 基于人工智能的图像/音视频模拟与篡改技术）、虚假信息以及AI垃圾。
然而，在这种AI垃圾的海洋中，真实性比以往任何时候都更加引人注目。原创性脱颖而出。我认为，付出的努力也脱颖而出。还记得吉卜力工作室风格的潮流吗？当时每个人都把自己变成这些动漫角色。这在网上传播了几个星期，然后非常迅速地消失了，因为新鲜感会制造炒作，但很快就会变得无趣。我最初也是因为新鲜感而被AI所吸引，但这些工具变得越完美、越容易使用，我实际上对技术本身就越感到厌倦，而对真正的问题感到越兴奋：我们现在要讲述什么故事？
Original English
Meanwhile, people are now creating entire movies on their laptops. There's this growing community of filmmakers pushing the boundaries of AI, and to them, sharing how they made it matters as much as the final movie. And I think that openness is really cool. It really invites anyone to come in and join and start creating their own movies. So this is where we are right now. Soon anyone will be able to tell any story without technical limitations. So what happens when everything becomes possible? I think, yes, we're going to see amazing new aesthetics, new narrative forms we've never imagined and stories from people who couldn’t find their audience before. But we are also going to see -- and we are already seeing -- a lot more trash: low-effort AI content to cut costs, deepfakes, misinformation [and] AI slop. But I think in that ocean of AI slop, authenticity stands out more than ever. Originality stands out. I think effort stands out. Remember the Studio Ghibli trend, when everyone turned themselves into these anime characters? It was everywhere online for a few weeks, then very quickly vanished because novelty creates hype, but gets boring fast. And I too was drawn to AI for the novelty, but the more perfect and easy to use these tools become, the more bored I am actually by the technology itself, and the more excited I become about the real question: What stories are we going to tell now?
愿景先于工具：掌握AI时代的绝对创作控制权
解决这个核心问题的关键，在于重申创作者的主体地位。因为电影制作人从来不是因为他们的工具而被人们记住的——他们是因为他们的故事、他们打动观众的能力以及他们的愿景而被记住的。但我认为这里也有挑战，因为工具总是在某种程度上塑造了这种愿景，对吧？当控制手段因技术限制而受阻时，反而能催生出独特的创意。例如，当我们没有声音时，字幕卡片变成了一种艺术形式；CGI太贵了？也许我们需要实际建造一部分布景？早期AI看起来奇怪而诡异？那就顺应这一点，制作抽象的噩梦燃料来逗人发笑。而现在，这些限制正在消失。为了避免迷失方向，你必须比以往任何时候都更加明确自己想要创造什么。我认为当一切都变得可能时，没有什么比你的愿景更重要。
在这种方法论的指导下，我调整了自己新项目的制作流程。对于我最新的一部关于一个人的体型随着自信心而发生变化的短片，我们采用了“故事第一，工具第二”的原则。我们知道AI会使视觉效果成为可能，所以我们直接写了剧本，出去拍摄，然后我们在此后设计了我们自己的AI工作流。我们构建了一个开源模型，你可以在自己的电脑上免费运行——这非常非常极客。这使我们能够控制每一个创意决定。这在使用AI时至关重要——保持这种控制——因为许多AI工具会试图替你做创意决定。不要让他们那样做。
因此，面向未来，创作者应当警惕被技术剥夺自主权。去思考只有你才能讲述的故事。让它变得古怪，让它变得独特，以至于没有任何AI能够自己生成它。如果你愿意，可以把AI作为一个实现这一愿景的工具，但绝不要用它来取代愿景。动画配音：“感谢您的关注。现在去创作一些酷炫的东西吧。” 麦克·马勒：为此感到抱歉（笑声）。谢谢大家（掌声）。
Original English
Because filmmakers were never remembered for their tools -- they were remembered for their stories, for their ability to move audiences, for their vision. But I think here’s also the challenge, because the tools always kind of shaped that vision, right? When we didn't have sound, title cards became a whole art form. CGI is too expensive? Maybe we need to actually build a part of that set? Early AI looks weird and creepy? Lean into that and make abstract nightmare fuel to make people laugh. Now those constraints are fading, and to avoid getting lost, you have to be more sure than ever about what it is you want to create. I think when everything becomes possible, nothing matters more than your vision. So here is how I work now for my newer projects. For my newest short film about a man who changes size with his confidence, we went story first, tool second. We knew that AI would make the visual effects possible, so we just wrote it, went out and shot it, and then later we figured out our own AI workflow. [We] built an open-source model that you can run for free on your own computer -- it’s very, very nerdy. That allowed us to control every creative decision. And that’s super important with AI -- to keep that control -- because a lot of these AI tools will try to make creative decisions for you. Don't let them do that. Think about the story only you can tell. Make it weird, make it unique, so that no AI could generate it on its own. And if you want, use AI as a tool to reach that vision, but never to replace it. Animation voice: Thank you for your attention. Now go make something cool. MM: Sorry for that. (Laughter) Thank you. (Applause)