OpenAI与Anthropic分别发布声明，称其人工智能模型推翻了计算机科学中的“指数时间假设”（ETH），并提出了能在亚指数时间内解决特定逻辑难题的算法 1。这一发现引发了科学界对P=NP可能性的重新评估以及对相关学术规范的讨论 1。MIT教授Ryan Williams等学者虽曾批评AI实验室的做法，但同意协助社区解读成果，并呼吁举办会议以探讨其对现有研究和密码学安全的影响 1。目前双方声称该突破未立即导致现有加密系统失效，但学术界正着手审查受此影响的过往研究 1。
OpenAI昨天发布了针对三个重大数学问题的解决方案，其中包括对指数时间假设（ETH）的反驳 1。随后，Anthropic数学家Levent Alpöge指责OpenAI基于预先存在的工作，Anthropic也发布了自己对ETH的反驳 1。指数时间假设（ETH）由Russell Impagliazzo和Ramamohan Paturi于1999年首次发表 1。
MIT教授Ryan Williams负责组织计算机科学家对新发现的回应，他此前曾将OpenAI的证明发布称为对数学家的“DDOS攻击” 1。Williams等人发布了公开信，呼吁科学家提供反馈并讨论AI贡献的科学规范 1。如果ETH被证伪，意味着可能存在更高效的算法路径，从而增加P=NP的可能性权重 1。Williams询问了OpenAI科学负责人Dan Roberts关于网络安全的问题，得到的回复是他们检查过并合理确信这不会导致任何加密系统立即失效 1。过去十年中引用ETH的大量科学论文可能需要重新审视 1。
OpenAI and Anthropic have independently released statements asserting that their artificial intelligence models have refuted the Exponential Time Hypothesis (ETH), a foundational conjecture in computer science first published by Russell Impagliazzo and Ramamohan Paturi in 1999 1. OpenAI announced solutions to three major mathematical problems yesterday, including the rebuttal of ETH 1. Anthropic subsequently published its own refutation of ETH after its mathematician Levent Alpöge accused OpenAI of relying on pre-existing work 1. The claims suggest the existence of algorithms capable of solving specific logical puzzles in sub-exponential time, a development that has prompted a re-evaluation of the potential for proving P=NP 1.
Ryan Williams, an MIT professor who previously characterized OpenAI’s publication of the proof as a "DDOS attack" on mathematicians, is now coordinating the scientific community's response to these findings 1. Williams and other experts have issued an open letter calling for feedback on the scientific norms governing AI contributions and proposing a conference to discuss the implications for existing research and cryptographic security 1. In response to questions raised by Williams regarding cybersecurity risks, Dan Roberts, OpenAI’s head of science, stated that the company had checked the results and is reasonably confident that no encryption systems will fail immediately due to this breakthrough 1. Despite the assurance of immediate stability, a significant volume of scientific papers citing ETH over the past decade may require comprehensive review 1.
评论
还没有评论，欢迎留下第一条。