Anthropic 表示,人工智能模型的编码能力已经达到了一定水平,在发现和利用软件漏洞方面,它们可以超越除了最熟练的人类之外的所有人。
AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities, Anthropic said.