AI 干得更多，决定仍归人类
论证要点
- 研究基于复旦团队参与的 Atria Dawn Preview 项目，分析 56 名参与者、700 多条任务日志；该模型为 7440 亿参数 MoE 架构，在 16 项基准中的 5 项领先，但整体未超越竞品。
- AI 参与了 96.5% 的任务，四周内 agent 动作与人类输入的比值中位数从 11 升到 28.5；作者提醒这不能解读为自主性提升，因为每个决策仍由人类做出，只是带出更多 agent 步骤。
- 455 个已完成任务中有 151 个（约三分之一）被参与者判定为没有 AI 就无法以同等范围和质量完成，且分布在 56 人中的 27 人，并非少数重度用户。
- 方法与参数上最常见的是「AI 提议、人类选择」（55.4%）；人类做出 85.5% 的方法与参数决定，AI 仅 9.2%；目标与范围上人类最终拍板 93.4% 的情况。
- 出错时 76% 的任务靠人类介入推进，其中人类主要提供信息：补充上下文或澄清需求（35.2%）、诊断问题并切换方法（34.7%）；亲自改代码仅 3.2%，完全接管仅 0.7%。
- 作者提出 AI 角色经历三阶段：研究对象、个人任务工具、项目伙伴；第四阶段递归自我改进仍是推测，模型能在训练任务上变强，不等于能更好地开发继任者。
对实践的启示
对做 AI 应用的人来说，当前可行的分工是让 AI 大量产出方案与执行，把目标设定、方法选择和异常诊断留给人。真正的瓶颈不是 AI 的执行力，而是人类判断的带宽；当 agent 链路长到无人能审时，审批会退化成盖章，这需要提前设计。
作者承认的反方与局限
作者承认存在「橡皮图章」风险：当每个决定都建立在人类无法完整审查的长链条 agent 工作之上，监督会变得困难，人类可能只能对所见内容盖章通过；不少参与者为免打断长任务而让 agent 自主运行，这条边界是出于便利而非对 AI 权限的深思。
原文摘引
AI proposes, humans choose
Human judgment, rather than execution, was the bottleneck.
In the worst case, humans become reviewers who can only rubber-stamp what they see.
▸ 展开原文全文（32 段，抓取自 The Decoder）
AI agents do more of the work in model development, but humans still make the decisions
A research team documented how humans and AI agents worked together to build a new AI model. The findings challenge some expectations about how independently agents can work.
When AI agents help build new AI models, who makes the decisions? A team involving researchers from China's Fudan University studied its own project to find out. It analyzed more than 700 task logs from 56 participants, along with logs from the agents they used.
The project centered on developing an agentic language model called Atria Dawn Preview, built on a mixture-of-experts architecture with 744 billion parameters and designed for research and engineering tasks.
The model was trained through a pipeline that ties each task to a real execution environment. It calls tools, generates intermediate results, and gets checked against external signals like tests, metrics, or source evidence. The team says it leads on five of 16 benchmarks, including web search and cybersecurity, though it doesn't hold an overall edge over competitors.
A third of completed AI-assisted tasks wouldn't have been attempted without AI
AI was used in 96.5 percent of the tasks reviewed. Over the course of the project, participants handed off more and more to agents. The median ratio of agent actions to human inputs rose from 11 to 28.5 over four weeks. The team cautions against reading this as growing autonomy. Each human decision led to more agent steps, which didn't mean the agents were making more decisions themselves.
Participants were also asked whether they could have completed their share of a task without AI, at the same scope and quality. Of 455 completed AI-assisted tasks, 151 were rated infeasible without AI, roughly a third. These tasks were spread across 27 of the 56 participants, so they didn't come from just a handful of power users. AI didn't speed up existing work in these cases. It made work possible that would never have been started otherwise.
AI proposes, humans choose
For methods and parameters, the most common pattern was "AI proposes, human selects" at 55.4 percent. Overall, humans made 85.5 percent of decisions about methods and parameters, while AI made just 9.2 percent. Humans made the final decision on goals and scope in 93.4 percent of cases.
AI's share of proposals ranged from 17 to 55 percent depending on the decision type. Its share of final decisions stayed in the single digits. Who proposed the options varied widely, but humans consistently made most of the final choices. Even among the 151 tasks rated infeasible without AI, humans chose the goal 95.4 percent of the time.
Humans supply context, not manual labor
The same pattern shows up when things go wrong. Of 588 tasks with a recorded difficulty, 76 percent moved forward through human intervention, and in 23 percent the agent solved the problem on its own. Human help almost always came in the form of information, either by adding context or clarifying requirements (35.2 percent) or by diagnosing issues and switching methods (34.7 percent). Humans rarely did the work themselves. Partial edits accounted for 3.2 percent of cases, and full takeovers just 0.7 percent.
When AI outputs needed revision, the AI handled the changes itself 75.4 percent of the time after receiving human feedback. Human judgment, rather than execution, was the bottleneck.
The team describes three phases in AI's role, from a subject of research to a tool for individual tasks and now a project partner. In that current role, AI drafts and adjusts plans within goals set by humans. A speculative fourth phase would involve recursive self-improvement, with stronger models producing stronger successors.
The authors say a model can improve at its training tasks without getting better at developing its successor. How AI could propose varied research directions and assess their value before results are available remains an open question.
The rubber-stamp risk
When every decision rests on a longer chain of agent work than any human can review, oversight gets hard. In the worst case, humans become reviewers who can only rubber-stamp what they see, the team writes. Many participants also ran agents in autonomous modes to avoid interrupting long runs with constant approvals. That boundary was drawn out of convenience, not from any deliberate choice about how much authority AI should have.
The paper lands in the middle of a debate about recursive self-improvement. Anthropic considers an AI that develops its own successor possible sooner than expected, and CEO Dario Amodei is calling for a speed limit for the industry as a result. According to Anthropic, humans now make only a single-digit percentage of decisions about research direction at the company. OpenAI uses GPT-5.6 Sol across its entire development cycle, while Google and DeepMind let AI agents explore alternative strategies through recorded search trajectories with Dream-RSI, though they only improve the search strategy, not the model itself.
Over a thousand employees at leading AI companies recently warned that their organizations may be on the verge of automating AI research. A separate study from Princeton and the UK AI Security Institute reached a conclusion more in line with the Atria team's findings, showing that frontier models can handle research engineering but fail at the judgment calls that actually matter.
AI News Without the Hype – Curated by Humans
Subscribe now
Read on for the full picture.
Subscribe for hype-free coverage.
- Full access to every article on THE DECODER
- No ads
- Join the comments and community discussions
- A weekly AI news recap via mail
- 6x/year: "AI Radar" — deep dives on the AI topics that matter most
- Daily AI news, always up to date
- Our full ten-year archive
- Covered by a team with 10+ years in AI