I expect rapid progress but not towards general superintelligence
打开互动全文版（逐段中英对照 + 针对本篇提问）→本文认为，AI 将在工程与基础设施能力上快速进步，使模型在分布式 GPU 工程以及优化训练和推理栈方面超越人类，但这不会通向通用超级智能。核心论点是：近期收益来自扩展推理时计算和自动化可验证的工程任务，而非模型研究能力的戏剧性跃升。作者预测，架构与数据选择方面的预训练研究可能在未来两到三年内实现自动化，推动模型智能的有效成本近乎指数级下降，并引发智能体模型的杰文斯悖论。结论是：即便在数学和编程之外没有具备经济价值的超人特质，这种工程加速也将大幅推动 AI 向经济各领域的扩散，并促成跨学科科学发现，而改进强化学习环境仍是一大唾手可得的成果。
This article argues that AI will progress rapidly in engineering and infrastructure capabilities, making models superhuman at distributed GPU engineering and optimizing the training and inference stack, but this will not lead to general superintelligence. The core claim is that near-term gains come from scaling inference-time compute and automating verifiable engineering tasks, not from dramatic step-changes in model research capabilities. The author predicts that pretraining research in architecture and data selection could be automated within two to three years, driving near-exponential declines in the effective cost of model intelligence and triggering a Jevons paradox for agentic models. The conclusion is that this engineering acceleration will massively boost diffusion of AI into the economy and enable cross-disciplinary scientific findings, even without economically valuable superhuman traits outside math and coding, while improving RL environments remains a major low-hanging fruit.
我常常感到惊讶，当听到业界顶尖研究人员说他们认为 AI 将在几年内比他们更擅长自己的工作，而我并不真正明白自己为何怀疑这一点。我得到的解释是，研究人员看到了模型在基础设施和工程能力上的巨大加速。我完全同意，模型将在几年内成为超人的分布式 GPU 工程师。这将使实验和调整模型公式变得容易得多，但不会使我们的模型在本质上发生巨大变化。
I've often been surprised when I hear from top researchers in industry that they think AI will be better than them at their job in a few years, and I didn't really know why I doubted it. The explanation I arrived at is that researchers are seeing this massive acceleration in the infra and engineering capabilities of the models. I fully agree that the models will be superhuman distributed GPU engineers in a few years. This will make experimentation and tinkering with the formulation of the models far easier, but it will not make our models dramatically different in nature.
在某些方面，Ilya 早早宣布 AI 进入研究时代是正确的。如今有了编码智能体，研究变得容易多了。这种便利是通过一种灵活且引人入胜的不同工作形式实现的，这标志着一个时代的开始，在这个时代，好的想法在软件中可能比好的执行更有价值。
In some ways, Ilya was early in his declaration of AI being in the era of research. Research is much easier today now that we have coding agents. This ease is through a different form of work that is flexible and engaging, which signals the start of an era where good ideas can be much more valuable than good execution in software.
我们在这条路上还处于非常早期的阶段。这也不是该领域第一次在 AI 更偏向“研究”还是“工程”之间平衡的转变。在深度学习起飞之前，AI 也更偏向研究。如今，众所周知，最优秀的研究人员是根据他们在复杂基础设施中实现和扩展自己想法的能力来评判的。我们即将经历几年的巨大加速，届时瓶颈将再次越来越少地来自工程。有人称之为 RSI，但更接地气的形式只是“并行化的、AI 辅助的语言建模”。你不需要相信起飞场景，就能看到即将到来的基础设施改进有多大，以及这将如何改变 AI 工作的性质。
We're very early on this path. It's also not the first transition the field has seen across the balance in if AI is more "research" or "engineering." Before deep learning took off, AI was also far more of a research endeavor. Today, the best researchers are well-known to be judged by their ability to implement and scale their ideas in complex infrastructure. We're about to go through a massive acceleration over a few years where the bottleneck is less and less engineering again. Some call this RSI, but a more grounded form is simply "parallelized, AI-assisted language modeling." You don't need to buy into takeoff scenarios to see just how much infrastructure improvements are coming, and how this'll change the nature of working in AI.
本文的目的是概述为什么这种工程加速会发生，并触及它不会完全解决的问题。即使我们在数学和编码之外没有获得_具有经济价值的_超人特质，这也将极大地促进技术向经济的扩散。许多近期进展来自于用当前工具扩展推理时算力，而不是模型在研究能力上的戏剧性阶跃变化。
The goal of this piece is to give an overview on why this engineering acceleration will happen, and touch on what it won't quite fix. This'll be a massive boon for diffusion of the technology into the economy, even if we don't get _economically valuable_ superhuman traits outside of math and coding. A lot of this near-term progress comes from scaling the inference-time compute we have with current tools, rather than dramatic step-changes in how good the models are at research.
AI 训练和推理栈的许多部分是非常可验证的。训练指标如每 GPU 每秒令牌数，这只是训练速度。推理指标看起来像是每个提示的令牌数、每个令牌的 FLOPs，或者实际上只是每个答案的成本。这两者都非常可优化，尤其是因为存在已建立的子问题和架构权衡可以改进它们。我预计 AI 智能体将在几年内帮助端到端地优化这个过程，届时我们的推理能力将非常接近 GPU 等加速器上可能的最大底层算力。在过去几年中，公司在推理效率上已经能取得巨大收益，例如在宣布某个价格点后，为模型提供服务的成本节省了约 10-30%。整个栈将复合增长，我预计未来几年模型智能的有效成本将接近指数级下降（可能比近期趋势更快）。
Many pieces of the AI training and inference stack are very verifiable. Training metrics like tokens per second per GPU, which is just training speed. Inference metrics look like tokens per prompt, FLOPs per token, or really just cost per answer. Both of these are very optimizable, especially as there are established sub-problems and architecture trade-offs which can improve them. I expect AI agents to help optimize this process end-to-end in a few years, where our inference capabilities get very close to the underlying maximum compute possible on our accelerators like GPUs. Over the previous few years, the gains companies can make on inference efficiency were already giant, e.g. saving like 10-30% on cost to serve a model after announcing it at a certain price point. This entire stack will compound, and I expect the effective cost of model intelligence to decline near-exponentially over the coming years (potentially faster than recent trends).
这些唾手可得的效率提升应该只需几年就能实现。更长期来看，将是加速器与模型在更长时间线上的协同设计，这能在灵活的 GPU 平台之上再带来数个数量级的效率提升。在效率快速提升的这段时间里，GPU 的灵活性应当是有益的，因为架构上的探索空间似乎很重要，并且像是自动化研究的一个主要杠杆。我认为，预训练研究——至少是服务于我们当前这类模型的架构和数据选择方面——在 2-3 年内实现自动化，这一预测是合理的。
This low-hanging fruit of efficiency should take only a few years to capture. The longer term will be co-design of accelerators and models on longer timelines, which give extra orders of magnitude in efficiency on top of the flexible GPU platform. During this time of rapid efficiency gains, the flexibility of the GPU should be kind, as exploration space in architecture feels important and like a major lever of automated research. A prediction of pretraining research, at least in architecture and data selection to serve our current class of models, being automated in 2-3 years feels reasonable to me.
所有这些都将对智能体式模型引发巨大的杰文斯悖论。我预计需求只会增加，因为该行业在很大程度上受限于如何找到更好的方式来定位和交付智能体。Meta 的 Muse 智能体是一个早期迹象，我们应该期待面向不同受众和目标用例的更多类似 Muse 的体验。其价值来自于理解智能体如何工作，而非推动性能的前沿。
All of this will be a massive trigger of Jevons paradox for agentic models. I expect demand to only increase, as the industry is largely bottlenecked on figuring out better ways to orient and deliver the agents. Meta's Muse agent is an early indicator of this and we should expect more Muse-like experiences for different audiences and target use-cases. It's value created by understanding how agents work, rather than pushing the frontier of performance.
在生物学和化学等更广泛的科学文献中，相应的阶段将是一个由模型发现大量跨子领域发现所定义的时代。AI 模型在爬取文献、在稀疏网络之间建立联系方面是超人级的，而这些网络目前由很少互动的小型科学家社区管理。这将预示着一个科学发现的新时代（例如治愈大多数癌症），还是加速科学原本就有的发展轨迹，两者之间只有一线之隔。
The equivalent phase in the broader scientific literature like biology and chemistry will be an era defined by models finding numerous cross-sub-field findings. The AI models are superhuman at crawling literature and making connections across sparse networks which are currently stewarded by small communities of scientists who rarely interacted. There will be a fine line between this heralding a new era of scientific discovery, e.g. with cures of most cancers, or an acceleration of the arc science was already on.
另一个工业规模的唾手可得的成果是提高强化学习环境的总体质量。不用费力就能看到又一家新的强化学习数据公司收入突破 1 亿美元或 10 亿美元。数量远超我的预期。该行业的平均产出质量极低。无数研究人员都同意，他们购买的很多东西 frankly 就是垃圾。与此同时，领先的实验室从购买这些数据中看到了明确的投资回报。
Another industrial-scale piece of low-hanging fruit is improving the general quality of RL environments. You don't have to look far to see another new RL data company that's crossed $100M or $1B in revenue. There are way more than I expected. The average outputs from this sector is remarkably low-quality. Countless researchers agree that lots of what they buy is frankly crap. At the same time, the leading labs see clear return on investment from buying the data.
许多强化学习环境糟糕的方式显然是可修复的。
The way that many RL environments are cruddy is clearly fixable.