行业动态
garymarcus.substack.com ·
@Gary Marcus
Gary Marcus批评黄仁勋无证据宣称AGI已到来
Gary Marcus发文批评黄仁勋在无定义、无证据的情况下宣称AGI已到来，认为这是企业试图以权威取代科学讨论，并指出Astra在基准测试中并未显著超越前代。
AI推荐理由 Gary Marcus对黄仁勋的AGI声明提出批评，强调缺乏定义和证据。来源为个人博客，但作者是知名AI学者，观点有深度。对理解AGI争议有参考价值。
65
正文
几小时前，黄仁勋宣称通往AGI的竞赛已经结束。不幸的是，黄仁勋没有给出任何证据和定义，在我看来，这像是试图通过企业命令来接管一个科学问题。我敦促他阅读由Hendrycks、Yoshua Bengio及许多其他人（包括我自己）撰写的agidefinition.AI，并考虑Astra在我与Brundage的10项赌注中列出的那类例子上的表现如何。自动形式化可能终于触手可及，也许（？）可靠的编程也能实现；我怀疑Astra能否命中其他八项中的任何一项。按照传统定义，Astra仍然有所欠缺。在没有定义的情况下宣布胜利只会混淆视听。§ 除了缺乏定义之外，我预计如果Astra真的是AGI，它应该会遥遥领先于竞争对手。相反，许多人认为它在实际应用中并不比Fable 5.1强多少：而且一套备受推崇的基准测试表明，Astra确实有真正的改进，但并未显著偏离趋势。当真正的AGI到来时，我们不需要眯着眼睛看。我们也不需要黄仁勋的认可。到那时，结果会不言自明。立即订阅 更新：一位读者指出ARC-AGI测试是一个标准。以下是ARC发明者对此的说法：更新二：尽管黄仁勋今天祝贺了OpenAI，但大约六个月前，在Astra之前，他已经就一个早期模型宣布了胜利。
查看原文（中文正文为 AI 翻译）
A couple hours ago Jensen Huang declared that the race to AGI is over. Unfortunately, Huang gave no evidence and no definitions, which feels to me like an effort at a takeover of a scientific question by corporate fiat. I would urge him to read agidefinition.AI by Hendrycks, Yoshua Bengio & many others (including myself), and to consider how Astra is doing on the kinds of examples I laid out in my 10 -item bet with Brundage. Autoformalization may finally be in reach, and maybe (?) reliable coding; I doubt that Astra will have hit any of the other eight. By conventional definitions, Astra still falls short. Declaring victory without a definition simply muddies the waters. § Aside from the lack of definitions, I would expect that if Astra really were AGI, it would be a quantum leap ahead of its competitors. Instead, many see it as not much more than on a par with Fable 5.1 in real-world applications: And one well-respected set of benchmarks suggests that Astra is a genuine improvement but not significantly off-trend. When real AGI arrives, we won’t need to squint our eyes. And we won’t need Jensen’s approval, either. The results, at that point, will speak for themselves. Subscribe now Update: One reader pointed to the ARC-AGI test as a criterion. Here’s what the inventor of ARC said about this: Update two: Although Jensen congratulated OpenAI today, about six months he already declared victory, pre-Astra, with respect to an earlier model.