Is this the dawn of the Tokenpocalypse?
Is this the dawn of the Tokenpocalypse?
这会是“Token末日”的黎明吗？
Microsoft recently announced major pricing changes for GitHub Copilot — changes that were drastic enough that a Reddit user said their company has started calling it the Tokenpocalypse. On the latest episode of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I discussed what those changes might mean for the larger AI ecosystem.
微软最近宣布了 GitHub Copilot 的重大定价调整——这些调整非常剧烈，以至于一位 Reddit 用户表示，他们公司已经开始将其称为“Token末日”（Tokenpocalypse）。在 TechCrunch 的《Equity》播客最新一期节目中，Kirsten Korosec、Sean O’Kane 和我讨论了这些变化对更广泛的 AI 生态系统可能意味着什么。
After all, as Anthropic and other big AI companies plan to go public, leading to awkward questions about profitability, we’re likely to see similar price increases for other AI products, and more usage restrictions as businesses try to keep costs under control. “Can these AI labs collapse that cost [and] progress the tech enough in a way that it eventually meets in the middle with customers’ appetite for spending?” Sean wondered.
毕竟，随着 Anthropic 和其他大型 AI 公司计划上市，引发了关于盈利能力的尴尬问题，我们很可能会看到其他 AI 产品出现类似的涨价，以及随着企业试图控制成本而实施更多的使用限制。Sean 质疑道：“这些 AI 实验室能否降低成本，并推动技术进步，使其最终与客户的消费意愿达成平衡？”
Kirsten, meanwhile, suggested that this also reflects “how quickly things are moving.” In just a few months, companies became obsessed with “tokenmaxxxing,” then turned against it due to the high costs. So as AI companies write their IPO filings, she asked, “How do you even write these risks in, because they are evolving before our eyes?”
与此同时，Kirsten 指出，这也反映了“事物发展的速度有多快”。在短短几个月内，企业从沉迷于“Token最大化”（tokenmaxxxing）转变为因高昂成本而对其避之不及。因此，当 AI 公司撰写 IPO 申请文件时，她问道：“你该如何将这些风险写入其中？因为它们正在我们眼前不断演变。”
Keep reading for a preview of our conversation, edited for length and clarity.
请继续阅读我们对话的预览，内容经过删减和润色。
Anthony Ha: When we were planning for this, Sean, you called this the Tokenpocalypse. And I want to hear more about what you think about it, but there was an example of Microsoft deciding with GitHub Copilot that they’re going to start charging more per token [instead of a flat rate]. This whole ecosystem is heavily, heavily subsidized by investor money. And so stuff that seems like it has no cost is, in fact, incredibly expensive. And now we’re going to get to a point where more of that cost is going to get passed on to the end consumer, to the customer. How is that going to change behavior? I don’t think we know, but there’s going to be a lot of pain.
Anthony Ha： Sean，我们在策划这期节目时，你称之为“Token末日”。我想听听你对此的更多看法。微软决定 GitHub Copilot 开始按 Token 收费（而不是固定费率），就是一个例子。整个生态系统在很大程度上依赖于投资者的资金补贴。因此，那些看起来没有成本的东西，实际上极其昂贵。现在我们正处于一个临界点，更多的成本将转嫁给终端消费者和客户。这会如何改变人们的行为？我认为我们还不知道，但这将会带来很多阵痛。
Sean O’Kane: I mean, how many token-related risk factors do we think are going to be in the Anthropic’s S-1? This is a big question. It’s something that I’ve mentioned a lot on this show and we seem to just keep running into it, where Uber has done like the full arc in the span of a month and a half of saying, “Boy, we kind of blew through our budget on this stuff way quicker than we thought this year.” And then, “Ooh, maybe this is going to be a little too expensive, we need to put caps on this, and we need to limit people’s usage inside the company.” That’s just a little worrying.
Sean O’Kane： 我的意思是，你认为 Anthropic 的 S-1（上市申请）文件中会有多少与 Token 相关的风险因素？这是一个大问题。我在节目中多次提到这一点，而且我们似乎一直在遇到这种情况：Uber 在一个半月的时间里走完了整个过程，先是说：“天哪，我们今年在这方面的预算超支比预想的快得多。”然后又说：“噢，也许这太贵了，我们需要设置上限，限制公司内部的使用。”这确实有点令人担忧。
Imagine if you see that happen so quickly at a company like Uber, that is using this stuff a lot, and it’s just a question of: Can these AI labs collapse that cost [and] progress the tech enough in a way that it eventually meets in the middle with customers’ appetite for spending? A funny thing to think back on is, I don’t think there was really any strategy involved in charging $20 a month [for ChatGPT Plus] when ChatGPT originally came out. It was just sort of like, “Let’s spit out a number.” And we’ve all been reckoning with that ever since. Clearly, people pay more for the more advanced models, but even that still isn’t enough to close that gap to the true cost. So that’s clearly the biggest question here.
想象一下，如果像 Uber 这样大量使用 AI 的公司都如此迅速地发生这种情况，问题就在于：这些 AI 实验室能否降低成本，并推动技术进步，使其最终与客户的消费意愿达成平衡？回想起来很有趣的一点是，我认为 ChatGPT 最初推出时，每月 20 美元（ChatGPT Plus）的定价并没有什么策略。当时就像是“随便报个数字”。从那以后，我们一直在为此买单。显然，人们愿意为更先进的模型支付更多费用，但这仍然不足以弥补与真实成本之间的差距。所以这显然是目前最大的问题。
Kirsten: All of this, to me, illustrates how quickly things are moving. I mean, when you really think about it, the whole tokenmaxxxing thing has become a thing, peaked, and now is seen disfavorably, within six months. The scale of this, the whole pricing mechanism, to your point, was put in place before business models were really shaped and solidified around AI labs. And then, at the same time, you have the government trying to catch up. Also this week, President Trump signed an executive order — it is a narrow version, but this is designed to give the government a chance to review powerful AI models. So you have all this happening at a pace that I don’t think I’ve ever experienced. That’s why I’m really looking forward to some of these S-1 IPO registration statements, because of the risk [factors]. How do you even write these risks in, because they are evolving before our eyes, and day by day?
Kirsten： 在我看来，这一切都说明了事物发展的速度有多快。当你仔细想想，整个“Token最大化”现象在六个月内就经历了兴起、达到顶峰，现在又被视为负面。正如你所说，这种规模和整个定价机制是在 AI 实验室的商业模式真正形成和稳固之前就建立起来的。与此同时，政府也在试图跟上步伐。本周，特朗普总统签署了一项行政命令——虽然版本较窄，但旨在让政府有机会审查强大的 AI 模型。这一切发生的节奏是我从未经历过的。这就是为什么我非常期待看到一些 S-1 IPO 注册文件，因为其中的风险因素。你该如何将这些风险写入其中？因为它们正在我们眼前、每一天都在演变。
Anthony: Uber is an interesting example, Sean, because you mentioned their AI spend, but they’ve also come up in the AI discourse because sometimes, people who think there’s this bubble, they’ll bring up just how wildly unprofitable these tools are, these companies are, and then people will bring up Uber as a response. People talked about how unprofitable Uber was, but eventually you get to scale and then you close that gap. And I think that’s true. But also, for Uber to do that, it had to really transform itself as a company in a lot of ways. What Uber was at the beginning and what it is now, all the different areas of business that it’s had to expand into, the different ways that customers and drivers have gotten squeezed, those are things that had to happen to get to the point where it could be a profitable company. And I think you’re going to have to see similar transformations for a lot of these AI companies if they’re going to survive.
Anthony： Sean，Uber 是一个有趣的例子。你提到了他们的 AI 支出，但他们在 AI 的讨论中也经常被提及，因为有时那些认为存在泡沫的人会指出这些工具和公司是多么的无利可图，而其他人则会以 Uber 作为回应。人们曾谈论 Uber 是多么不赚钱，但最终你达到了规模效应，从而弥补了差距。我认为这是事实。但同时，为了做到这一点，Uber 必须在许多方面真正实现公司转型。Uber 起初的样子与现在的样子，它必须扩展的所有不同业务领域，以及客户和司机被挤压的不同方式，这些都是为了成为一家盈利公司所必须经历的过程。我认为，如果这些 AI 公司想要生存下去，我们也必须看到类似的转型。
Sean: Is there any way that these labs can squeeze pennies like Uber has squeezed the drivers over the years? Is there something squishy enough there for them to do that? I don’t know. This seems like harder, more straightforward costs in a lot of ways, so it’ll be interesting.
Sean： 这些实验室有没有办法像 Uber 多年来压榨司机那样去抠成本？它们有足够的空间来做这件事吗？我不知道。在很多方面，这似乎是更硬性、更直接的成本，所以这会很有趣。