arXiv:2505.22954 · 中英对照阅读
Darwin Gödel 机器:自我改进智能体的开放式演化
Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents
中文速览
论文想让 AI 不再只按人类预先设计的架构工作,而能自己改进自己的代码、越做越会改。作者提出 Darwin Gödel Machine(达尔文哥德尔机),让多个版本的编程智能体轮流修改代码、在编程基准上接受测试,并把表现各异的版本存进档案,供后续探索和改进。实验中,它在 SWE-bench 上的成绩从 20.0% 提升到 50.0%,在 Polyglot 上从 14.2% 提升到 30.7%,也超过了缺少自我改进或版本档案的对照方法。它的重要性在于展示了一种让 AI 累积已有成果、尝试不同改进路线并持续进步的可行方向,不过目前只验证了编程智能体,实验也依赖基准测试和安全管控。
摘要
当今大多数人工智能系统受限于由人类设计的固定架构,无法自主且持续地自我改进。另一方面,科学方法是一种累积且开放式的体系,每一项创新都建立在先前成果之上,从而促成未来的发现。人们日益期待,当前推进人工智能发展的手动过程本身能够实现自动化。若能安全地实现自动化,这将加速人工智能的发展,使我们能更早地从中获益。这一前景引出了一个问题:人工智能系统如何才能不断自我改进,同时提升解决相关问题的能力?元学习(meta-learning)能够自动发现新算法,但受限于一阶改进,以及由人类设计合适的搜索空间。Gödel 机器(Schmidhuber,2007)提出了一种理论上的替代方案:一种自我改进人工智能,能够反复以可证明有益的方式修改自身。遗憾的是,在实践中,证明大多数改动总体上有益是不可能的。我们提出 Darwin Gödel 机器(DGM),这是一种新颖的自我改进系统,它反复修改自身代码(从而也提升其修改自身代码库的能力),并使用代码基准测试对每项改动进行实证验证。受达尔文式进化和开放式演化研究的启发,DGM 不断扩充生成智能体档案库。它从该档案库中抽取智能体,由这些智能体通过自我修改,创造出新颖且有趣的自身版本。这种开放式探索形成了一棵不断生长、包含多样化高质量智能体的树,并能沿着搜索空间中的许多不同路径并行探索。在实证上,DGM 自动提升了自身的编码能力(例如,改进代码编辑工具、长上下文窗口管理和同行评审机制),使其在 SWE-bench 上的表现从 20.0%提升至 50.0%,在 Polyglot 上的表现从 14.2%提升至 30.7%。此外,DGM 的表现显著优于不具备自我改进或开放式探索能力的基线方法。所有实验均采取了安全防护措施(例如,沙箱隔离和人工监督)。总体而言,DGM 是迈向自我改进人工智能的重要一步,它能够沿着不断展开、通向无尽创新的道路,收集自身的演化基石。所有代码均已开源,网址为 https://github.com/jennyzzt/dgm。
Most of today’s AI systems are constrained by human-designed, fixed architectures and cannot autonomously and continuously improve themselves. The scientific method, on the other hand, is a cumulative and open-ended system, where each innovation builds upon previous artifacts, enabling future discoveries. There is growing hope that the current manual process of advancing AI could itself be automated. If done safely, such automation would accelerate AI development and allow us to reap its benefits much sooner. This prospect raises the question of how AI systems can endlessly improve themselves while getting better at solving relevant problems. Meta-learning can automate the discovery of novel algorithms, but is limited by first-order improvements and the human design of a suitable search space. The Gödel machine (Schmidhuber, 2007) proposed a theoretical alternative: a self-improving AI that repeatedly modifies itself in a provably beneficial manner. Unfortunately, proving that most changes are net beneficial is impossible in practice. We introduce the Darwin Gödel Machine (DGM), a novel self-improving system that iteratively modifies its own code (thereby also improving its ability to modify its own codebase) and empirically validates each change using coding benchmarks. Inspired by Darwinian evolution and open-endedness research, the DGM grows an archive of generated coding agents. It samples agents from this archive, which self-modify to create new, interesting versions of themselves. This open-ended exploration forms a growing tree of diverse, high-quality agents and allows the parallel exploration of many different paths through the search space. Empirically, the DGM automatically improves its coding capabilities (e.g., better code editing tools, long-context window management, peer-review mechanisms), increasing performance on SWE-bench from 20.0% to 50.0%, and on Polyglot from 14.2% to 30.7%. Furthermore, the DGM significantly outperforms baselines without self-improvement or open-ended exploration. All experiments were done with safety precautions (e.g., sandboxing, human oversight). Overall, the DGM represents a significant step toward self-improving AI, capable of gathering its own stepping stones along a path that unfolds into endless innovation. All code is open-sourced at https://github.com/jennyzzt/dgm.
术语表
- Darwin Gödel Machine (DGM)
- Darwin Gödel 机器(DGM)
- self-improving AI
- 自我改进人工智能
- self-modification
- 自我修改
- self-referential system
- 自指系统
- meta-learning
- 元学习
- Gödel machine
- Gödel 机器
- coding agent
- 代码智能体
- coding benchmark
- 代码基准测试
- SWE-bench
- SWE-bench
- Polyglot
- Polyglot
- open-endedness
- 开放式演化
- open-ended exploration
- 开放式探索
- archive of generated agents
- 生成智能体档案库
- stepping stones
- 演化基石
- search space
- 搜索空间
- empirical validation
- 实证验证
- formal proof
- 形式化证明
- Darwinian evolution
- 达尔文式进化
- natural selection
- 自然选择
- large language model (LLM)
- 大型语言模型(LLM)
- transformer
- Transformer
- recurrent neural network
- 循环神经网络
- attention mechanism
- 注意力机制
- foundation model (FM)
- 基础模型(FM)
- frozen pretrained model
- 冻结的预训练模型
- quality-diversity algorithm
- 质量—多样性算法
- sandboxing
- 沙箱隔离
- human oversight
- 人工监督