Tech Visionary Says the Big AI Labs Don’t Get What People Want
Tech Visionary Says the Big AI Labs Don’t Get What People Want
科技远见者：大型 AI 实验室并不了解人们真正想要什么
Tim O’Reilly’s yardstick for measuring the worth of a company, person, or society has long been: create more value than you capture. It’s no surprise that O’Reilly—publisher, internet pioneer, VC, conference organizer, and dispenser of tech wisdom—is applying that metric to the way people design and use AI. Specifically, he’s pushing for a future where open-source AI is an elixir for the masses. He worries that, like Microsoft in the 1990s, today’s hyperscalers are trying to lock users into their products. So he is promoting efforts to open-source AI technology—not only making critical technical details such as neural-net weights accessible, but unlocking the whole stack of an AI system, giving control to designers and users.
长期以来，蒂姆·奥莱利（Tim O’Reilly）衡量一家公司、一个人或一个社会价值的标准一直是：创造的价值要大于获取的价值。作为出版商、互联网先驱、风险投资人、会议组织者以及科技智慧的传播者，奥莱利将这一标准应用于人们设计和使用人工智能的方式，这并不令人意外。具体来说，他正在推动一个让开源 AI 成为大众“灵丹妙药”的未来。他担心，就像 90 年代的微软一样，当今的超大规模云服务商正试图将用户锁定在他们的产品中。因此，他正在推动 AI 技术的开源——不仅要公开神经网络权重等关键技术细节，还要解锁 AI 系统的整个技术栈，将控制权交还给设计者和用户。
O’Reilly sees AI as a new creative medium, which he uses extensively—and even has a blog about his chats with it. During our conversation, we discovered that we disagree about AI’s role in producing original content. Guess who took which side.
奥莱利将 AI 视为一种新的创作媒介，他本人也广泛使用 AI，甚至还专门写了一个博客来记录他与 AI 的对话。在我们的交谈中，我们发现双方在 AI 在原创内容生产中的作用上存在分歧。猜猜我们各自持什么立场。
STEVEN LEVY: You are all in on open-source AI. Make your case. 史蒂文·列维（STEVEN LEVY）：你全力支持开源 AI。请陈述你的理由。
TIM O’REILLY: First, let’s make sure we’re talking about the same thing. When most people talk about open-source AI, they’re really just talking about open-weight models. It’s much bigger than that. In the ’90s, when everybody else was focused on open-source licenses, I was like, “No, no, it’s about the architecture of the system. Does it enable participation?”
蒂姆·奥莱利：首先，我们要确保讨论的是同一件事。当大多数人谈论开源 AI 时，他们实际上只是在谈论“开放权重”模型。但开源的意义远不止于此。在 90 年代，当其他人都在关注开源许可证时，我的观点是：“不，不，关键在于系统的架构。它是否促进了参与？”
Why is that needed? 为什么这很有必要？
The big labs are reading the future wrong. They have told themselves a narrative where having the biggest, best model is the key to the future. Big models like Claude are optimizing for particular use cases, but they aren’t necessarily the use cases that people want. I want to embed my own special sauce. The most important thing is having a clean separation between the model, the harness, and the application. And right now we’re not getting that. They built an architecture of control rather than an architecture of freedom and participation, so they have the ability to track you.
大型实验室对未来的解读是错误的。他们给自己编造了一个叙事：拥有最大、最好的模型是通往未来的关键。像 Claude 这样的大模型正在针对特定的用例进行优化，但这些未必是人们真正想要的用例。我想要嵌入我自己的“独门秘籍”。最重要的事情是实现模型、工具框架（harness）和应用程序之间的清晰分离。而目前我们并没有做到这一点。他们构建的是一种控制架构，而不是自由与参与的架构，这样他们就有能力追踪你。
Isn’t it against the interests of big companies to give up control? 放弃控制权难道不违背大公司的利益吗？
Oh, it’s totally against their interests. But that doesn’t mean that they’re making the right strategic decision. For a long time, the latest and greatest models were really better for everything. Now they’re better for some things and worse for others. People are talking a lot about how Fable and Sol are worse writers than the lower-level models. [Note: Anthropic and OpenAI would disagree.] The breakthroughs that we’re getting in so-called frontier AI are actually pushing us further away from what ordinary people are going to need. We could win frontier AI here in the US, and China will kick our ass because they have lower-level models diffused widely through society. The goal is to give people the ability to innovate freely, to paint outside the lines.
哦，这绝对违背了他们的利益。但这并不意味着他们做出了正确的战略决策。很长一段时间里，最新、最强的模型确实在各方面都更好。但现在，它们在某些方面表现更好，而在另一些方面则表现更差。人们经常讨论 Claude 和 GPT-4（注：原文此处可能指代 Claude 和 GPT 系列）在写作上不如低级别模型。[注：Anthropic 和 OpenAI 会对此表示异议。] 我们在所谓“前沿 AI”领域取得的突破，实际上正让我们离普通人的需求越来越远。我们可能在美国赢得了前沿 AI 的竞争，但中国会打败我们，因为他们有在社会中广泛普及的低级别模型。目标是赋予人们自由创新的能力，让他们能够打破常规。
People worry that open-source software can be a security risk, since bad actors will be able to jump the frontier model’s guardrails. 人们担心开源软件会带来安全风险，因为恶意行为者将能够绕过前沿模型的护栏。
All of the cybersecurity incidents we’ve seen are from the frontier models. So risks like cybersecurity and the ability to develop pathogens are actually an argument for slowing down the frontier models more than an argument for restricting open-weight models.
我们所见过的所有网络安全事件都源自前沿模型。因此，网络安全和开发病原体能力等风险，实际上是放缓前沿模型发展的理由，而不是限制开放权重模型的理由。
So you feel that with open source the AI powers will no longer be dominant? 所以你认为有了开源，AI 巨头们将不再占据主导地位？
I don’t predict the future. but I will say the world is proceeding the way that I hoped it would. What if the big frontier models end up like mainframes or supercomputers—aimed at really hard problems, which are not actually the thing that gets diffused throughout society? People are building things like Pi, an open-source [agentic] harness. One of the things we’re working on at my nonprofit, the AI Disclosures Project, is the idea of an open-memory consortium. Mark Zuckerberg’s thesis is, he’s going to lock you in because Meta will give you the AI that knows you best. The open-source vision needs to say no to this. Open source can give you the ability to switch models, switch providers, and maintain all the context that it needs.
我不会预测未来。但我可以说，世界正朝着我希望的方向发展。如果大型前沿模型最终变得像大型机或超级计算机一样——只针对极难的问题，而实际上并不是在社会中普及的东西，那会怎样？人们正在构建像 Pi 这样的开源（代理式）工具框架。我的非营利组织“AI 信息披露项目”（AI Disclosures Project）正在研究的一个方向是“开放记忆联盟”。马克·扎克伯格的论点是，他要锁定你，因为 Meta 会提供最了解你的 AI。开源愿景必须对这种做法说不。开源可以让你有能力切换模型、切换服务商，并保持其所需的所有上下文。
You recently wrote a piece in the Economist about how Elon Musk—as well as other tech leaders—have so much power in their companies that they are anti-capitalist, in the sense that their desires, rather than the marketplace, determine strategy. 你最近在《经济学人》上写了一篇文章，谈到埃隆·马斯克以及其他科技领袖在公司中拥有过大权力，以至于他们是“反资本主义”的——因为是他们的个人欲望而非市场在决定战略。
In general, Silicon Valley has become anti-capitalist. They use the language of markets. But around 2010, the whole model came into play with Uber and Lyft. Funded by VCs, they literally spent billions of dollars to subsidize rides. The venture capitalists picked the winner rather than the market. That became the template. All the capital piles on and chokes out other possible paths.
总的来说，硅谷已经变得反资本主义了。他们使用市场的语言，但在 2010 年左右，随着 Uber 和 Lyft 的出现，整个模式发生了变化。在风投的资助下，他们实际上花费了数十亿美元来补贴乘车费用。是风险投资人选择了赢家，而不是市场。这成为了模板。所有的资本都堆积在一起，扼杀了其他可能的路径。
In AI, all the money is funneling into a few companies, but none seem to have a lock. 在 AI 领域，所有的资金都涌向了少数几家公司，但似乎没有哪家公司拥有绝对的锁定优势。
Because it’s a failed model. The real action is open source. Just like in the early 1990s, when everybody thought it was all a battle for who ruled the PC, and then the web came out of left field while nobody was paying attention. I think that the real AI future will be a ferment of innovation that’s not being funded by venture capital—in the same way that the web wasn’t funded by VCs.
因为这是一个失败的模式。真正的行动在于开源。就像 90 年代初，当每个人都认为这是一场关于谁统治 PC 的战争时，互联网却在无人注意的情况下横空出世。我认为真正的 AI 未来将是一场创新的发酵，它不会由风险资本资助——就像互联网当初并非由风投资助一样。
One of those disruptions might be your own company, built on books and writing. 其中一个颠覆者可能就是你自己的公司，它建立在书籍和写作之上。
The book business has been in decline for 25 years. At our peak, our book business was maybe $70 million, and it’s now $30 million. As books generated less revenue, we had to find a way to compensate people for sharing their knowledge, because that’s our fundamental business. So now AI comes along and hoovers everything up. Can we build a set of tools that help people with that? Being able to invoke the knowledge of experts is a superpower, and we have to figure out what that looks like now.
图书业务已经衰退了 25 年。在巅峰时期，我们的图书业务收入约为 7000 万美元，现在只有 3000 万美元。随着书籍产生的收入减少，我们必须找到一种方法来补偿分享知识的人，因为这是我们的核心业务。现在 AI 出现了，把一切都吸走了。我们能构建一套工具来帮助人们应对这种情况吗？能够调用专家的知识是一种超能力，我们现在必须弄清楚这在当下意味着什么。
When I edit this interview I will boil down our long conversation to a shorter one, but I won’t use AI to do it. And of course, the writing in the introduction will be all mine. It’s actually our policy. 当我编辑这次采访时，我会把我们漫长的谈话浓缩成简短的版本，但我不会使用 AI 来完成。当然，引言部分的文字将全部由我撰写。这实际上是我们的政策。
I think that’s a legacy thing that’s going to… 我认为这是一种终将……的遗留习惯。