为创作者构建 AI | Luma & Phota Labs

用户对得到的结果满意吗？

Are they are they user satisfied with the result they're getting?

以一种自然、不打扰的方式收集这种反馈，真的特别重要。

And can you collect that kind of feedback in an ambient way that's just like, nonintrusive way is just so important.

这个观点太有意思了。

This is such an interesting fact.

我之前刚和一个人聊过他们的Phota结果。

I was actually talking to someone earlier about their Phota result.

他认为幸福和追求是两回事。

His view was happiness and just seeking are two different things.

当他看到结果时，说：这简直就是我。

When he saw the result, and he was like, this is exactly me.

这简直是最佳的模型输出。

It's like the best model output.

相似度达到了百分之百。

The likeness is a 100%.

他白天拍的照片太胖了。

Phyla too fat in the day.

所以他对结果并不满意。

So he's not happy with the results.

这就涉及到某种筛选机制——能让用户开心的，未必符合评估标准。

So there is some kind of like, you know, filtering, like something that makes the user happy may not be the benchmark Mhmm.

你 guys 是怎么解决这个问题的？

Thing that, you know, like, you might be straight how are you guys solving that?

归根结底，我们认为用户才是最终说了算的人。

Well, like, at the end of the day, we think, you know, user is the one who is control controlling things.

对吧？

You know?

如果他们希望照片里的自己看起来什么样，我觉得这完全是他们的选择，我们得支持他们这么做。

If they want them to look in a certain way in the in the photo, I think that's that's their choice, and we need to empower them to do that.

而且，这本来就是一件非常主观的事。

And, again, this is a very subjective thing.

美化这个问题，其实已经争论很久了，涉及的范围非常广。

Beautification is has been a long, long time debate of the, like, the spectrum of of it.

每个人都有自己不同的审美。

Everyone has their own taste.

是的。

Yeah.

个性化才是关键。

Personalization is key here.

对。

Yeah.

你懂的。

You know?

我觉得个性化更多是个范围问题，而不是让每个人都能在画布上随意操作。

I imagine personalization is more of a scope problem compared to having empower everyone, everything to do things on the canvas.

面对画布上可能存在的无数种操作方式，你觉得该怎么评估呢？

How do you think about evaluation when it comes to there are so many unlimited workflows one can do on the canvas?

是的。

Yeah.

我觉得差不多。

I I think somewhat similar.

你得看看人们最终是怎么使用这个工具的。

You have to look at how people ultimately use the tool.

我们不可能在发布前猜出所有评估方式，但归根结底，得看上线之后有没有人用。

We can't come up with as many guesses on how to evaluate before we release it, but at the end of the day, you have to see once it's out there, are people using it?

人们都在哪儿用它？

Where are people using it?

人们在哪儿感到挫败？

Where are people getting frustrated?

他们在哪儿觉得控制够用，哪儿又觉得不够用？

Where are they feeling like the level of control is sufficient and where is it not sufficient?

尽量从人们使用这个工具的不同场景中收集各种反馈信号。

And trying to gather as many of the signals as possible across the different domains that people use the tool.

有意思。

Interesting.

这么说完全说得通。

So this totally makes sense.

我觉得这个工具的用途其实很广，使用场景分布很广。

I think of a very, I guess, like a broad tool as there's a distribution of use cases.

刚推出的时候，那些明显的主要用法你肯定都测试过了。

When you launch the tool, there's obvious ones that you probably would have tested before launching.

但分布的尾部可能很长，也可能不长。

And there may or may not be a fat tail on the distribution.

关键问题是，你有没有想把某些使用场景从尾部推到头部？

The question is, are there specific use cases you want to move from the fat tail to the head?

你遇到过这样的情况吗？

Have you ever run into use cases like that?

有。

Yeah.

我觉得其中一个挺有意思的是，看到品牌们是怎么统一使用自家品牌规范的。

I think one interesting one was seeing how brands would bring in same brand guidelines.

这一点我们之前其实没怎么考虑过。

It's not something we really considered very much.

确实如此。

Yeah.

但一旦他们把这些规范用起来，你就发现这其实是了解用户兴趣的非常有用的线索。

But then once they brought them in, you realized this was a extremely useful source of information for really getting to what a user is interested in.

所以这里就引出一些问题：不仅要让这些品牌或其他公司能把这些信息接入进来，还要考虑怎么让他们顺利实现这个流程。

And so there's some questions there of not just how do you enable that pipeline for these, say, brands or other companies to include that information Yeah.

但更重要的是，怎么让那些不熟悉这类流程的人也提供这类信息？

But how do you also get that type of information from others who aren't used to working with those sorts of pipelines?

没错。

Yeah.

我得说一句。

I should say.

是的。

Yeah.

对我们来说，有些用户试图使用我们的当前API，但它只支持人类和宠物的身份识别。

For us, there was some users who are trying to our current API only supports human and pets for identity.

有些用户在尝试绕过系统，想用它来做产品摄影的个性化处理。

There's some user who are trying to hack our system so they can use it for personalization for product photography.

哦，有意思。

Oh, interesting.

他们说：‘我们就是需要这个功能’，所以现在这已经列入我们的开发计划了。

And they're like, we we want this, and so that now that's on our road map.

你知道，我们确实想做，我的意思是，这很合理，但这类使用场景我们当初根本没想到。

You know, we want we want I mean, it makes sense, but so that's kind of an example where, like, we didn't see that kind of use case.

我们刚上线时，发现市场反馈后，就决定优先提升这方面的支持。

We first launched it, and there's market pool, and then we just want to increase our priority on on those.

是的。

Yeah.

你觉得人的相似度和产品的相似度之间有什么区别？

How do you see the difference between, I guess, human likeness and product likeness?

我们的大脑在识别人脸时会参与其中，所以我猜这其中的容错空间应该很小。

I mean, our brain is involved to recognize human faces, so I imagine there's Yeah.

容错空间更小。

Lower margin for error.

没错。

Right.

那产品呢？

But what about product?

这方面能学到什么？

What what's the learning there?

是的。

Yeah.

首先，这两者之间有很多相似之处。

I mean, so first of all, there are a lot of similarities between these two.

首先，大家都说通用基础模型在识别身份方面已经很好了，但当你试着生成自己的照片时，问题就来了。

First is, like, everybody say, oh, the general foundation model is already very good at identity until you try to use it to generate image of yourself.

产品摄影也是同样的道理。

Same thing with product photography.

它生成特定产品看起来挺不错，可一旦你是产品负责人，想生成这个具体产品的照片，就不行了。

Oh, it's pretty good at generating this specific product until you are the product owner and you're trying to generate photo of that particular product.

模型直接就失效了。

It just stops working.

对。

Yeah.

这其实是个不太明显的问题。

So that's a non obvious one.

区别在于，身份这个东西更主观、更动态一些。

The difference is, like, identity is a little bit more, I would say, subjective and dynamic.

真的很难用语言说清楚，到底什么才是身份。

It's really, really hard to put into words what is identity.

对吧？

Right?

至于产品摄影，你得把文字、形状这些细节都做到位才行。

Like, product on the other hand, you know, you need to get the text right, you need to get the shape of the product right, to get all of those right.

所以这方面的情况不太一样。

So it's a little bit different there.

两者各自都有独特的挑战。

And there are unique challenges in both.

比如，文字渲染在产品摄影里特别关键，但在身份保留上就没那么重要。

For example, text rendering is so important in product photography, not so much in identity preservation.

我猜问题就变成了：你是希望模型全包了，还是让后续流程来接手模型能做的部分？

I imagine the question becomes, do you want the model to do it all or do you want the subsequent workflow to take some of the things that a new model gonna do?

我想听听你们俩的看法，遇到什么任务时，你觉得应该由模型来解决，什么任务你愿意交给流程处理？

Just curious about both of your opinions on what kind of tasks when you run into it, you feel like it's something a model should solve and what kind of tasks you're okay with having the workflow solving it.

我们看到语言模型出现了一个趋势，就是开始采用‘思考模式’。

We've seen a trend in language models where they've transitioned to thinking modes.

对吧？

Right?

现在要是用ChatGPT这类模型时不加思考，根本感觉不到它有多聪明。

If you use ChatGPT or one of these models nowadays without thinking, it just doesn't feel smart.

没错。

Right.

我觉得这属于一种中间状态——表面上是工作流，但底层还是靠核心模型驱动。

And I view that as kind of this in between of it's like a workflow, but it's still using the core model.

我觉得创意工具也可以朝这个方向发展，确实如此。

And I think creative tools can go in that direction too Yeah.

你可以先生成内容，然后评估它、在此基础上继续完善。

Where you you generate stuff, but you can evaluate it, you can build off of it.

可能是用户参与其中，也可能是智能代理在帮忙，但这样能帮你更好地达成最终结果。

Maybe the user's involved in that, maybe an agent's involved in that, but it helps you get to the final result.

明白了。

Got it.

是的。

Yeah.

我觉得模型和工具之间是有区别的。

I think there's a separation of model versus technology.

比如在个性化方面，其实并不重要。

Like, you in terms of personalization, it doesn't matter.

它就是一个通用模型，能解决大家的问题。

It's like a a general model that solves everyone's problem.

只要技术一样，我们就能把同样的技术用在不同用户身上。

As long as the technology is the same, we can apply the same technology to different to different users.

但关键是，你要把这个工具或者模型设计得能激发人的创造力，让人自己找到最佳的工作流程，去解决一些你根本没想到的问题。

But, again, you want to make that tool that or the model so that you can unlock the creativity in people so they can figure out the best workflows to solve a particular problem that you haven't even thought about before.

对。

Yeah.

你如何看待可控性？

How do you see controllability?

我的意思是，AI的第一阶段是生成内容，然后我们不断重提提示，直到得到想要的结果。

I mean, phase one in AI is we generate things, and we just keep re prompting the model for the thing we want.

这简直像在抽彩票。

It's almost like a lottery.

第二阶段是我们通过模型层面或系统层面增加更多控制，让结果更稳定。

Phase two is that we put more control either on the model level or on the systems level to make that work.

对于AI创作工具来说，更好的可控性到底意味着什么？

What does it mean for a AI created creativity tool to have, like, better controllability?

是的。

Yeah.

我觉得这还是个开放性问题，实现控制的方式有很多种。

I think it's an open question, and there's different ways you can add control.

但简短的回答是，你必须加入控制机制。

But the short answer is you definitely need to add control.

对。

Mhmm.

我觉得，任何真正想用这些工具来做专业创作，或者想替代现有创意流程的人，最常抱怨的就是：你不想靠打字来操作。

I think that was the number one complaint from anyone who's actually trying to use these tools for any sort of professional or or to replace any part of their existing creative workflow is that you don't wanna operate with text.

光靠文字肯定不够。

There has to be something more than just text.

没错。

Right.

如果非得用文字，那你就得有更高的控制精度，而文字在这方面很难做到。

Or if it is text, you need that level of control, which can be very difficult in text.

所以我们做了很多尝试。

And so we've experimented a lot.

最近，我们主要在做视频转视频的研究，因为觉得这是在时间维度上实现精准控制的好方向。

Most recently, we've been doing a lot of work in the video to video space because we think that's, like, a very good way to start incorporating pretty precise controls, in the time dimension.

对。

Right.

在空间维度上，有很多有趣的可能性，比如涂鸦、点击特定区域，把这些控制方式一层层加到模型上，让用户能直接操作，同时也把这些功能开放给智能代理，让代理能按指令完成这类创作任务。

In the spatial dimension, I think there's a lot of interesting things with being able to scribble, being able to, like, point to certain regions, and so adding more and more of these controls on top of the model so that the user can do, while also opening up these controls for an agent so that you can instruct an agent to do these sorts of ads.

是的

Yeah.

我觉得这还关乎帮用户表达出他们真正想要的东西。

I think it's also about helping the user to say what they want.

很多时候，模型并没有完全做到你期望的效果。

I think a lot of the time, the model doesn't do exactly what you want.

这是因为用户的提示本身有点模糊。

It's because of, like, the user's prompt or whatever is a little bit ambiguous.

除了你得告诉模型你要做什么，我觉得还有一个被忽视的研究方向：模型能不能主动告诉你，它需要什么样的输入才能做好。

And it's also so other than the you need to tell the model what you do, I feel like this is the underexploited research direction is if the model can tell you what it needs to get your inputs Mhmm.

就是在执行之前，先主动向你确认需求。

To act acting asking for your inputs before it does something.

我觉得这一点也特别重要。

I I feel like that's also really important.

因为当你作为一个新手和创意专业人士合作时，他们经常会问你问题，你就能把自己的想法和偏好告诉他们，他们再据此调整出你想要的效果。

Because, like, when you when you are working with a creative professional, as an amateur, a lot of the times, they ask you questions, and you can you can tell them your opinions, your preferences, and they sort of achieve their preferences and opinions.

我觉得，现在基本上就是单向的。

I feel like that mostly, today, it's almost a one way street.

用户告诉模型该做什么。

It's like the user tell the model what to do.

这个观点确实很到位。

I think that's a really good point.

如果你去一家工作室，说‘给我做个十秒的视频，内容是狗在草地上跳’，他们肯定不会照单全收。

If you go to, say, a studio, and you say, make me a ten second video about a dog jumping in the grass, they're never gonna take that deal.

他们会要求更具体一些。

They're they're gonna want more specific.

没错。

Right.

是的。

Yeah.

你得想想怎么把这种思路用在人工智能上。

You have to figure out how to apply that in AI.

对。

Yeah.

有意思。

Interesting.

你觉得在模型和应用协同设计方面，是否有助于提升可控性？

Do you think model app co design has benefits when it comes to introducing more controllability?

你对此怎么看？

What's your view there?

我觉得是的。

I think so.

如果能让用户在你的应用里互动、完成他们的工作流程，你就能看出他们真正关心哪些地方。

I think being able to have the users interact in your app and work through their workflow in your app, you learn which things a user actually cares about.

你会明白哪些步骤他们必须反复重做，以及他们为什么最终选择了某个特定步骤。

You learn which steps do they have to redo multiple times and why maybe they chose the step that they ended up choosing.

对。

Yeah.

这能大大帮助模型的设计。

And that can help a lot with model design.

没错。

Right.

对。

Yeah.

我觉得另一个方向也同样重要，那就是你的产品还能帮助用户理解，怎样才是使用模型的最佳方式。

I think the other direction is also really important is, like, your product can also help educate the user what's the best way to sort of use their model.

我觉得以前这些事全都是靠写代码来完成的。

I feel like we're seeing that all of that was, like, coding and stuff.

对吧？

Right?

人们早就习惯了用 VS Code 之类的工具来做所有事情。

Like, you know, people are used to, like, to do everything like Versus Code or whatever.

现在呢，你直接在终端里敲几行命令，就能完成这些操作了。

And now, like, call code just in the terminal, you can you can do do that code.

这才是使用模型的最佳方式，因为你避开了很多其他麻烦。

And it's the best way is to sort of use that model because you cut cut through a a lot of other things.

我还是觉得，应用和模型的设计也有它的优势。

I still feel like app and model code design has advantage too.

对。

Yeah.

说到AI创意工具，我首先想到的就是它们非常直观，一目了然。

When we think about AI creative tools, the first thing that comes to mind, at least for me, is it's very visual, just being clear.

特别依赖视觉呈现。

Very visual animals.

现在我们已经习惯了和智能代理打交道。

And now we're used to working with agents.

感觉这些代理就像外星人，它们感知世界的方式跟我们完全不同。

It feels like almost like agents are aliens who don't perceive the world the same way that we did or we do.

所以我很好奇，会不会出现一种情况，创意工具主要变成后台运行的？

So I'm curious, like, is there a world where creative tools become primarily headless?

如果你想要利用这种代理循环，那会是什么样子？

What would that look like if you, you know, leverage want to leverage agent type loops?

对。

Yeah.

我觉得，再说回到工作室的场景，你其实是在和一个代理沟通。

I I think, again, going back to like if you were to work with a studio, you are talking to an agent.

对吧？

Right?

你在和一个与你协作的人交流，但他们不会用文字回复你。

You're talking to someone who is working with you, but they're not talking back to you in text.

他们会给你展示例子。

They're showing you examples.

他们会带你过一遍，说：

They're walking through, hey.

你觉得这些点子怎么样？

Do you like these ideas?

你喜欢这些想法吗？

Do you like these ideas?

我觉得这就是AI创意工具未来的发展方向。

And I think that's the direction you'll see these AI creative tools go.

是的。

Yeah.

对。

Yeah.

比如我 brainstorm 的时候，想写个代码生成一个展示图表的网页，其实我只需要一个草图就行。

Like, when I when I try to brainstorm something and, you know, you try to write a code to to, like, generate HTML page showing me the graph graphics of it, I want just a sketch out of it.

对吧？

Right?

这大概是最有效的迭代方式。

Like, that that's probably the best way to iterate.

对吧？

Right?

一张图胜过千言万语。

Like, sort of an image is more than a thousand words.

对吧？

Right?

没错。

Like Sure.

是的。

Yeah.

这真的特别重要。

So that's that's just so important.

另一方面，我觉得人们还没找到最好的方式，把信息有效地传递给模型。

On the other hand, I feel like people haven't figured out a way what's the best way to feed the viewer information to a model.

我觉得现在的模型还是有点‘看不见’。

I still feel like model today is a little bit blind.

我的意思是，它确实在变好，没错。

I mean, it's getting better Right.

通过具备高层次的语义理解能力。

By having high level semantic understanding.

但视觉信息其实挺冗余的，我觉得是这样。

But visual is just such a, I would say, redundant information.

像素之间重复太多，跟语言完全不一样。

There's, like, pixels that are very redundant, so unlike languages.

那该怎么把这种信息传递给模型，让它不再‘瞎’呢？

So how do you sort of give the model that kind of information so the model is not blind anymore?

这太关键了。

That's so important.

我觉得我们还没找到解决办法。

I I don't think we have figured out that yet.

这真有意思。

That is so interesting.

这让我想到，在创意工具里，选择合适的表达方式有多重要。

It does remind me of like the importance of having the right representation in a creative tool.

我是说，马特，你之前做过神经辐射场，以前大家用的是拉丝法，现在神经辐射场的表示方式已经形成了一整套效率谱系——高斯滑块细节表现很棒，但用起来特别麻烦，因为它不像一个优美的数学函数。

I mean, Matt, you worked on NERF and before that people were on Lashes and now there's a spectrum on how efficient a representation is on NERF and, you know, like Gaussian slide is great in details, but it's really hard to work with because it's not like a beautiful mathematic function.

你是怎么看待不同表示方式的？在你想要打造的设计工具里，未来的功能迭代需要什么样的表示形式？

How do you think about representation or different kinds of representations for, you know, the feature iteration of the design tools you want to make?

我我我

I I I

我觉得表示方式的问题，归根结底还是控制权的问题。

think the representation question ultimately comes down to the control question too.

你将来希望对它做出哪些修改？

It's what do you want to be able to modify it in the future?

对。

Yeah.

如果你在设计一张海报，知道以后要改文字，那你就该选一种不是基于像素的表示方式，不然改文字会特别麻烦。

If you're designing a poster and you know you're gonna be changing the text on it, you probably want a representation that's not pixel level or changing that text.

没错。

Yeah.

对。

Right.

如果你只是在调整图片，那用像素来处理是合理的。

If you're just modifying pictures, then pixels make sense.

如果你是在浏览三维场景，比如拍电影，那可能更需要一个三维的表达方式。

If you're navigating three d scenes, then for for, say, say, movies, you may want a three d representation.

所以我觉得并没有一个放之四海而皆准的答案。

So I don't think there's necessarily one answer.

关键要看你想要什么样的控制粒度。

It's whatever makes sense with a given level of control.

我

I

觉得这一点特别重要。

think that's so important.

我觉得现在很多人更关注模型用哪种表达效果最好。

I think right now, a lot of people are more focusing on what's the best reputation for the model.

但正如马特所说，我认为这些声誉的使用者不只是模型，还有人类。

But as Matt said, I think the consumer of those reputation is not just the model, it's also the human.

所以，你能不能以一种人类能理解的方式呈现它，然后让他们能够控制这种呈现，对它进行操作？

So, like, can you represent it in a way that human can understand and then be able to control that representation, do stuff to it?

对，没错。

Like yeah.

我很好奇，你们两位作为前研究人员，是怎么培养出为创作者设计产品的直觉的？

One thing I am curious about for both of you is how do you develop the intuition for building for the creators as former researchers?

马特，我知道你本身也是艺术家。

But Matt, I know you are also artists in your own right.

那么，这对你做研究或推出的产品类型有什么影响？

So like, how does that affect what kind of research or what kind of product you ship?

我觉得这让你对控制层面有了一些独特的理解。

I think it sorta gives you some insight in the control aspect.

我记得小时候上美术课时，老师最先强调的就是‘空白画布’的问题。

Like, one of the first things I remember being drilled in during sort of art classes when I was younger is the blank canvas problem.

是的。

Yeah.

很多人容易卡在那儿，但只要开始一点点来，仔细地一个部分一个部分推进，就好了。

There's a tendency of wanting to well, getting stuck by that, but then once starting very meticulously, like, doing one section at a time.

对。

Right.

但其实更好的做法是直接动手试试，别想太多。

But in reality, it's better to just sort of go at it, just try things out Yeah.

多折腾几遍，直到找到让你最满意的结果。

Mess with it until you get to the thing that you're ultimately happy with.

对。

Right.

我觉得AI工具也是一样。

And I think AI tools are no different.

迭代速度很快，但关键是要不断迭代。

The iteration cycle's quite fast, but you really want to iterate.

不是只输入一个提示，就能直接得到最终结果。

It's not just put in a single prompt to get a final answer out.

你需要在此基础上不断调整和优化。

You wanna iterate on top of that.

这确实会影响我们设计工具时的思路。

So that definitely impacts how to think about building the tools.

另外，我觉得还要搞清楚该怎么描述工具中哪些地方有效、哪些无效，这对评估非常重要。

And then I think another aspect is figuring out what language to describe what works or doesn't work in the tool because this can be very important for evaluation

对。

Right.

过去这么多年，艺术领域的人早就发展出了一套描述方法，比如构图、光影之类的。

Where, I mean, throughout arts over the in the past however many years, like, people have built out these ways to describe Right.

构图、光影之类的。

Composition, lighting, etcetera.

把这些经验吸收进来，融入模型的设计，我觉得会很有帮助。

So having that knowledge and and incorporating it into the design of the models, I think, can be very helpful.

是的

Yeah.

就是聊天啊。

Talking yeah.

和用户交流非常重要。

Talking to users is really important.

我们会和用户沟通。

We talk to our users.

我们觉得，确实能感受到那种空白画布的问题。

We feel like, oh, we we definitely recognize the black blank canvas problem.

比如，你问用户：‘你希望怎么改进这张照片？’

Like, when you when you ask the user, how do you want to improve this photo?

他们也不知道。

They don't know.

但当你对照片做了些调整后，他们就开始有意见了，比如喜不喜欢你的修改。

When you make some edit to that photo, now they have opinions on, like, whether they like your edits or not.

这就是他们不喜欢的地方吗？

Is that what they don't like?

他们知道自己不喜欢什么。

They they know what they don't like.

没错。

Exactly.

所以啊，和用户交流真的非常重要。

So, you know, talking to users is is really, really important.

而且我觉得，对技术有深入的理解也很关键，这样才能在保持略微领先于用户的同时找到平衡点——这种直觉，不仅要从用户角度出发，也要从技术角度来培养。

And I also think having a very good understanding of the technology so that you can make this fine balance of, you know, staying ahead of a little bit ahead of the users, I I think is also really important to build that kind of intuition, not just from the user perspective, but also from the technology perspective.

你怎么看待过程和最终成果之间的关系？

How do you see the process versus end results?

因为在艺术创作中，如果是人来创作，很多心血都投入在创作过程中，而最终成果就是当你觉得过程已经达到满意点的时候。

Just because in art, if your human is creating art, a lot of it goes into the process of creating it, and then final result is when you decide the process come to a point of your satisfaction.

那你怎么看待智能体创作艺术的方式和这个有什么不同？

How do you compare that with how agents create art?

我觉得在很多方面，其实挺相似的。

I think in many ways, it can be quite similar.

我觉得就算是AI生成艺术，背后也还是有人参与的，因为这个创作过程本身，对艺术家来说才是最有趣的部分。

And I think even when agents create art, there's still this human involved in that process because that process is the part that's fun, I think, for artists.

你看那些特效的分解视频，就能发现最终画面背后有多少层、多少步骤的功夫。

You look at, like, breakdowns, VFX breakdowns, and you you see how many different steps and layers went into the final pixels.

是的。

Yeah.

而且这过程特别有意思。

And it's fun.

我觉得做这类工作的人，都很享受去拆解整个创作流程，看看每一步是怎么拼起来的。

And I think people who do that kind of work find it very exciting to kinda see how everything breaks down.

未来有了AI工具，分解的方式可能会大不一样，但这种分解的过程依然会存在。

I think the way things break down in the future with AI tools will look very different, but it'll still exist.

所以将来会是AI和人类分工协作：AI负责一部分，最后还是得靠人来把所有部分整合起来。

And so it'll be this kind of back and forth between agents doing some parts of the work, but then humans also ultimately tying it all together.

是的。

Yeah.

对。

Yeah.

这可能没直接回答你的问题，但一直有个争论，就是我们在做AI摄影的时候。

And this may not direct answer your question, but there's always this debate of like, okay, we're working on AI photography.

摄影会消失吗？

Is photography gonna die?

未来的普通摄影会消失吗？

Regular photography gonna die in the future?

我不这么认为。

I don't think so.

很多人拍照片，不只是因为喜欢最终的效果。

There are a lot of people who are doing photography not just because they enjoy the result.

他们特别享受这个过程。

They enjoy the process so much.

对吧？

Right?

现在的数码相机已经做得这么好了。

Like, even digital cameras have become so good nowadays.

还是有很多人坚持拍胶片。

A lot of people are still taking films.

是的。

Mhmm.

胶片摄影的过程特别有意思。

Film is such a interesting process.

对吧？

Right?

你得抓住那个瞬间，不然就错过了，而且拍完也不能马上看到效果。

Like, it has this this unique thing of you have to capture that moment, otherwise, And you don't see the result right away.

你得把胶片冲出来才行。

You have developed the film.

所以这其中有很多地方都很有意思。

So a lot a lot of it is interesting.

我觉得很多人依然享受这些过程。

So I think people are still enjoying those processes.

但与此同时，也有人只想快速看到结果，那我们怎么用AI创作工具帮他们简化流程，激发他们的创造力呢？

But at the same time, there are people who would just want the results And, you know, how do we empower them with the AI creative tools to make that easy for them, to unlock their creativity?

我觉得这也是我对AI创作工具感到兴奋的地方。

I think that's also where I'm excited about the AI tooling for creators.

确实存在大量潜在需求。

Definitely a lot of latent demand.

那些买不起高端相机、也从来没上过胶片冲洗课的人，现在找到了属于自己的表达方式。

So the people who did not have the money to buy a very fancy camera to go out there and take pictures, and they also never took a film developing class ever, like they found their medium.

我们来看一看这段历史。

Here's how we see the history too.

传统创作工具一直都在，但新进入者一上市，瞄准的就是那些从未使用过传统工具的用户。

Like the traditional, you know, creative tools, they were all there, but new entrants, you know, they come on the market, they targeted users who were not users of our traditional tool.

他们当然想学，但现在有更简单的方法能直接拿到结果。

They will want to learn, but there's an easier way for them to get there and get the results.

你有没有注意到这种潜在需求是从哪儿来的？

How have you observed where that latent demand come from?

比如在摄影方面，那些买不起相机的人，明明没接受过专业训练，却还是想要拍出好看的照片。

Like in photography, who are the people who are, you know, I don't have money to buy a camera, but yet I want some sort of results from Phota.

是的。

Yeah.

对。

Yeah.

我觉得摄影不只是艺术。

I think photography is such a it's not just art.

它更是记录我们人生重要时刻的凭证。

It's also a document of our record of your important moments in life.

这完全不一样。

So that's very different.

很多人希望把自己的生活瞬间好好记录下来。

And a lot of people, they want a good recording of their moments, their lives.

但他们不一定有拍出好照片的技巧。

They may not necessarily have the skills to take that photo.

也可能没有合适的设备来拍出这样的照片。

They may not have the equipment to take that photo.

过去，他们只能试着买一部更高端的iPhone，比如最新款，指望相机性能能好一点，可他们根本不会后期处理。

And for them, they can only in the past, they try to buy a fancier iPhone, maybe the latest generation iPhone, they hopefully, they get a better camera out of it, but they don't have the skills to do editing.

这些就是我们想要解决的潜在需求。

Those are the latent demand that we try to address.

我们真心觉得，不管是AI模型还是AI助手，只要能开发出一些创意工具，帮他们轻松完成拍摄，甚至直接替他们搞定，让他们能留下人生重要时刻的精彩照片，这就是我们的目标。

You know, we really feel like whether it's AI model or AI agents, creative tools that we can develop to make that easy for them, or even do the job for them so that they can get these good photos of their important moments in their life, that's what we want to do.

是的，我们公司以前做过职业肖像拍摄服务。

Yeah, so one thing we had in a firm, which is like, we had photo shoots for people's head shots.

那种是高端服务，你会请摄影师，用专业影棚、灯光和化妆团队。

And those are like, you're premium and you you invite a photographer and then the whole studio and lighting and makeup.

最近我发现，越来越多的人虽然没法参加线下拍摄，但会用Phota来生成照片，这其实是个很有趣的市场细分——你总得有个实体对象才能拍照，不管是产品、人，还是别的什么。

And recently I found that there's more and more folks who are gorgeous like, I couldn't make it to the photo shoot this way, but I'm going to generate it on Phota, which, you know, it's like a very interesting segment of market where you need a physical presence to take a photo, either that's a product or a human or something else.

对的。

Mhmm.

你有没有注意到，新用Phota的用户也是这样？

Have you seen that too, like, on the new Phota users?

有。

Yes.

确实有。

Yes.

我觉得，Photoshop这类工具，本质上其实也是一种布景环境。

I think and also, like, photo Photoshop is if you think about it, it's a setup environment anyway.

对吧？

Right?

你得自己安排灯光。

Like, you are setting up the lighting.

你得亲自去那里拍照片。

You're you you go there and you sort of take the photos.

但很多时候，你没法完全掌控所有细节，毕竟受物理条件限制。

And and most of the time, you don't have all control of everything just because of the physical limit.

没错。

Yeah.

但如果是用生成技术来做，能做的就太多了。

When you're doing it in a generation ways, there's so much thing so many things that you can do.

灯光想怎么调就怎么调。

You can change the lighting the way you want.

角度也能随便改。

You can change the angle.

连妆容不满意了都能重新调整。

You can even change your makeup if you want.

这些便利性我觉得特别重要，确实越来越多人都在发现这些应用场景。

All of this, I think, is just so important and people are finding those use cases for sure.

是的。

Yeah.

我也很想听听真正摄影师的看法，因为马特，你拍很多照片。

I would love to hear from actual photographers' point of view too, because Matt, you take a lot of photos.

所以我想知道，你怎么看待传统摄影和AI摄影这两种方式的差异？

So I guess, like, how do you think about this old versus new world of AI photography?

对。

Yeah.

我觉得挺有意思的。

I it's fun.

我觉得这两者确实有重叠的地方，可以结合起来用，但也很有意思的是，它们本质上是两种完全不同的创作媒介。

I almost viewed there's definitely, like, the overlap where you can do them together, but then there's also this very interesting where you're, like, treated in a very different medium.

没错。

Mhmm.

当你开始完全在AI的世界里创作时，你就可以想象，一张照片如果出现在世界另一端，或者在某种你根本不可能实地拍摄的灯光环境下，会是什么样子。

When you start doing things that are purely in the AI world, you can start thinking about what would an image look like maybe halfway across the world or in specific studio lightings, situations that you would never actually be able to do Yeah.

但你觉得这个想法或者画面挺酷的吧。

But you think it's like a cool idea or image.

我有时候看到一些很棒的画面，就会琢磨怎么用AI把它复现出来。

I find myself sometimes going and, like, seeing a cool image out there and trying to think how would I actually replicate this in AI?

到底该用哪些关键词才对呢？

What are the right, yeah, terms to use?

然后一不小心就陷进去了，一路深挖下去。

And and you kinda go through a rabbit hole to get to that point.

它打开了一个我以前从来没机会探索的新领域。

So it unlocks like a new area that I haven't been able to explore before.

是的。

Yeah.

你觉得在AI创意工具的新时代里，传统工具还扮演什么角色？

How do you think about the role of traditional tools in the new world of AI creative tools?

传统工具会不会变成底层基础设施，一直都在那儿，我们只需要训练AI去调用它们完成某些特定任务？

Are the traditional tools becoming infrastructure where, you know, they're always gonna be here and now we just train agents to know how to use them for, you know, bits and pieces?

还是说它们正在被重新构建？

Or are they getting rebuilt?

归根结底，这些东西都只是工具。

I think at the end of the day, all these things are just tools.

关键还是看谁有想法，想实现这个想法，用哪个工具最合适。

And so it really just comes down to whomever has some vision and they wanna execute on that vision, what tool is right for them.

对有些人来说， Photoshop 永远是首选。

For some people, it'll always be Photoshop.

对的。

Mhmm.

对新一代来说，可能就不是了。

For probably a newer generation, it won't be.

这最终还是得看个人。

And so it'll just be kind of dependent on the person.

对的。

Mhmm.

正如你提到的，也许将来会是智能代理在使用这些工具。

And as you mentioned, maybe it'll be agents that'll be interacting with these tools.

所以我觉得很难下定论。

So I think it's hard to say.

我觉得这些工具在某种程度上总会存在，但不同用户会赋予它们不同的用途。

I think these things will always exist to some extent, but it'll find different uses by different users.

是的。

Yeah.

对。

Yeah.

我很喜欢这种说法。

I really like this framing.

这并不是说我们遇到了一个全新的问题——比如新工具和老工具的对立。

It's it's not like we're having this new problem of, like, oh, very new tools, very old tools.

我觉得工具一直以来都在不断迭代。

I feel like the tools have already always be iterating in the past.

总会有新工具冒出来，不管是不是基于AI的。

Like, there's always new tools come out, whether it's AI based or not AI based.

新工具一直都在不断出现。

There's always new tools come out.

我们现在可能正处在一个工具迭代速度特别快的阶段。

Just we're probably at a stage where the tools are iterating so fast.

所以我觉得，有些工具将来可能真的用不着了。

So I I do think some some tools, you may not need them in the future anymore.

它们会被取代，但很多工具还是会留下来。

They're gonna be replaced, but a lot of tools will still exist.

就连我们现在用的AI模型，未来也会被新的模型取代，这本来就是一个不断迭代的过程。

And even the AI models that we have today, they're gonna be replaced as well, like new models come out, so just an iterative process.

把自家的模型整合到你开发的工具里，有什么好处呢？

What is the benefit benefit of having your own model as part of what you put in the toolings that you build?

开源一个模型，有什么优势吗？

Like and is there a benefit in open sourcing a model?

我觉得拥有自己的模型最大的好处之一，就是能基于它继续开发。

So I think one of the biggest benefits of having your own model is the ability to build off of it.

对。

Mhmm.

归根结底，你的产品有一群特定用户，你得为他们做到最好，而最好的方式就是定制化，做一些调整。

At the end of the day, your product has a certain set of users and you wanna do the best thing you can for those users, the best way to do that is by customizing it, maybe making adjustments

到

to

那些对通用开源模型来说没必要、但对你产品来说非常关键的改动。

the model that don't make sense for, say, a general open source model, but makes sense for your product.

没错。

Right.

这一点在现有模型基础上往往很难实现，只有你自己开发的模型才能提供这种专属的优化。

And that's one thing that is not always easy to do on top of existing models and may require some specialization that only you can provide in your own models.

没错。

Right.

是的。

Yeah.

对我们来说，有点不一样。

For us, a little bit different.

我们并不直接做基础模型。

We we don't we don't directly do foundation model.

我们主要做个性化服务。

We mostly work on personalization.

原因是现在有很多很棒的公司，比如 Luma。

The reason is because there are so many cool companies out there like Luma.

我们觉得个性化应该是用户自己掌控的。

We really think personalization is something that user should own.

那是用户自己的模型。

Like, that's the user's model.

对用户来说，那是他们的模型，但最终他们是在使用这个模型。

It'd be the user's model for them, but at the end of the day, they're using that model.

他们完全掌控这个模型。

They have full control of that.

所以，针对这一点，我们想开发一种技术，帮他们实现这个功能，但模型的所有权属于用户。

So, like, for for that particular thing, like, we want to build a technology to enable them to do that, but they own their model.

对。

Mhmm.

我们还希望他们能把自己的模型和任何喜欢的基础模型结合起来使用。

And we want them to be able to combine that with any foundation model they they like to use.

对吧？

Right?

比如，他们可能需要一个专门针对某个特定场景的模型。

It could be a specialized model that they must specifically be for for, like, a particular use case, for example.

对吧？

Right?

所以，真正把个性化和基础模型区分开来，正是我们的目标。

So and to to really disentangle this personalization versus foundation is sort of the goal that we have.

随着AI工具的发展，你觉得在未来几年里，顶尖艺术家和其他人最大的区别会是什么？

With the advancement of AI tools, what do you think will separate the best artists from everyone else in the next couple years?

从我自己的经验来看，已经很明显了，能看出一个人是不是用了我们的工具，甚至用的是任何AI工具。

I can say from my experience already, it's very obvious when when people use our tools or not even our tools, any of the AI tools.

你很快就能分辨出，谁只是随便丢个提示词就等着出结果，谁是真正思考过整个创作故事的。

You can tell pretty quickly someone who has who who's just throwing in a prompt and getting a result out versus someone who has thought more about, like, the whole holistic story.

我总是对那些有创意的艺术家能用这些工具做出什么感到惊讶。

I always get amazed at what very creative artists can do with these tools.

很多时候，你会看到一些作品，根本不像这些工具能做出来的，但艺术家们就是找到了办法让它成为可能。

Many times, you see things that just don't seem like they would have been possible given the tools capabilities, but artists have figured out how to make it possible.

所以我觉得，这才是真正的关键所在。

And so I think that's really the thing that separates it.

关键在于，你怎么用现在能接触到的工具，做出独一无二的东西？

It's how do you make something unique with the tools you have access to now?

对。

Yeah.

我觉得很难说谁是好艺术家、谁是差艺术家，但大概可以分成优秀艺术家和普通艺术家吧。

I I think I mean, it's it's hard to say there's good artists and bad artists, but, like, they're probably good artists and, like, average artists.

我觉得AI工具确实拉高了底线，让你我这样的人也能用这些工具做出挺不错的作品。

I feel like with AI tools, they're definitely raising the lower lower bar of just just so that, you know, even someone like like me can create something pretty, pretty good with the AI tools.

对。

Yeah.

但我认为，顶尖艺术家和普通艺术家之间的差距会变得更大。

But I also think that the gap between the very best artist and average artist is gonna be even bigger.

就像马特说的，那些清楚自己想要什么、怎么去实现的人，现在能更容易地表达创意，再借助这些强大的工具达成目标。

Exactly because what Matt said, you know, like, they're so so much easier for people to express their creativity when they have a very good understanding of what they want to get, how they wanna get there, and they can use these powerful tools to do that.

他们能做出以前根本做不到的东西，所以这个差距只会越来越大。

They can achieve something they wouldn't be able to achieve before, so the gap is gonna just get bigger.

当然，他们还得真正掌握这些新的AI工具。

And, of course, they also need to have very good understanding, mastering all these new AI tools as well.

我想再补充一点。

I think maybe one final point.

你肯定见过那些生成得不好的AI作品。

You've seen AI generations that are bad.

对吧？

Right?

我觉得我们都见过不少差劲的AI生成内容。

I think we've all seen them that are bad.

我敢说，那些产生糟糕作品的工具，也同样能生成让你惊叹的效果。

I can guarantee the same tools that made those bad generations, they've also made generations that you would think are amazing.

没错。

Exactly.

是的。

Yeah.

对。

Yeah.

今天早些时候，我听人说，他们觉得AI生成的都是垃圾。

Something I heard from someone today earlier was they assume AI to be slop.

好像只要是AI生成的东西，都是垃圾。

Like, anything AI models have would be slop.

能不能做出好东西，还得靠人来加工。

It's up to the humans to create something out of it.

比如你只是说，我要一张咖啡的图片，就普通一杯咖啡那样。

Just because if you just like say, I want a picture of a coffee as a like a cup of coffee.

第一次生成的结果通常都很普通。

The first generation is just mediocre.

这种图在任何图库网站都能找到。

You can have it in any stock image website.

没什么特别的，也谈不上个人风格。

It's nothing unique or it's nothing personal.

对。

Right.

所以最终能不能做出属于自己的东西，关键还是看使用者怎么去打磨，加入自己的风格。

So it really is up to the user to kinda make it like theirs to kinda put their mark and style on there.

你对个人风格怎么看？特别是从个性化这个角度来说？

What are your thoughts on personal styles, like, in in terms of personalization?

比如说，风格到底指的是什么？

Like, what comes to the style?

对用户而言，什么才算真正的个性化？

What constitute personalization for the users?

我觉得这个问题真的很难回答。

I feel like that's such a hard question.

说真的，我几乎觉得除了从数据中去学习，根本没法给它下一个准确的定义。

Like, it's almost like it's I almost feel like there's no good way to define it other than you can learn that from data.

这不光是看用户过去的数据那么简单。

So and it's not just like, okay, the user's existing data.

还得看用户使用时的反应，他们喜欢还是不喜欢。

It's also like when the user use it, what's their reaction to it?

他们对它是什么感觉？

How do they like it or not?

因为就算你问用户他们自己的风格是什么，他们填的那些关键词也可能不完整。

Because even if you ask the user what's their own style, I mean, they can put some keywords there that's not complete.

我觉得，如果解决问题的方式是先明确他们的风格，再建模，那我可能不会这么干。

So I feel like, you know, if the way to solve the problem is you really to kind of define their style and then model their style, that's probably not the approach I would take.

从研究的角度来看，我觉得你应该直接看数据，从数据里找出那些潜在的模式，这样反而能更自然地发现他们的风格。

Even from a research perspective, I feel like you look at the data, you find the attribution that the the data shows you, and then you probably have figured out that style.

而且即便到了这一步，这种风格依然是隐性的，因为它已经融入模型本身了。

And even at that time, it's still implicit because it's part of the model.

是的。

Yeah.

我明白了。

I see.

有意思。

Interesting.

我觉得这东西还会随着时间变化。

I think it's also something that changes over time.

是的。

Yes.

就像一个人也会随着时间变化。

Both as as a person changes.

我知道五年前我觉得有趣的东西，和现在完全不一样。

Like, I know the stuff that I thought was interesting five years ago versus now is very different.

而且有时候，这周我在做一个项目，下周又换另一个项目，自然就需要不同的风格。

But then also, you know, I'm working on one project one week and another project another week, and I inevitably need a different style.

所以也许我们可以这样定义：针对具体任务的风格，和用户整体偏好的个性化风格。

And so maybe that's how you define, like, the styles for specific tasks and maybe personalization, sort of the holistic, like, what you as a user tend to like.

对。

Mhmm.

不过，我觉得这两者有点混在一起了。

But, yeah, I think it's it's a bit mix.

太棒了。

Awesome.

那今天就到这里，非常感谢你来参加这个播客。

And with that, well, thank you so much for coming on the pod.

这次对话真是太有趣了。

This is such a fun discussion.

是的。

Yeah.

真的很感谢你。

I really appreciate it.

对。

Yeah.

没错。

Yeah.

谢谢你。

Thank you

感谢你邀请我们。

for having us.

感谢收听这期的a16z播客。

Thanks for listening to this episode of the a 16 z podcast.

如果你喜欢这期节目，别忘了点赞、评论、订阅，给我们打分或写评价，也推荐给你的朋友和家人。

If you like this episode, be sure to like, comment, subscribe, leave us a rating or a review, and share it with your friends and family.

想听更多节目，欢迎前往YouTube、Apple Podcast和Spotify。

For more episodes, go to YouTube, Apple Podcast, and Spotify.

在X上关注我们@a16z，并在a16z.substack.com订阅我们的Substack频道。

Follow us on x at a sixteen z, and subscribe to our Substack at a16z.substack.com.

再次感谢收听，我们下期节目再见。

Thanks again for listening, and I'll see you in the next episode.

提醒一下，本节目内容仅供参考，不构成任何法律、商业、税务或投资建议，也不应用于评估任何投资或证券，且不针对任何a16z基金的投资者或潜在投资者。

As a reminder, the content here is for informational purposes only, should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security and is not directed at any investors or potential investors in any a sixteen z fund.

请注意，a16z及其关联方可能持有本节目讨论的公司的投资。

Please note that a sixteen z and its affiliates may also maintain investments in the companies discussed in this podcast.

如需更多详情，包括我们的投资组合链接，请访问a16z.com/disclosures。

For more details, including a link to our investments, please see a 16z.com forward slash disclosures.