辛顿：大模型已展现推理能力，AI或成更优智能形态

导语

在本次专访中，图灵奖得主杰弗里·辛顿再次就人工智能的潜在风险发出严峻警告。他明确指出，当前的大语言模型已展现出通过“链式思考”进行推理的能力，这颠覆了此前认为AI无法真正推理的论断。辛顿将当前AI的发展类比为人类养了一只注定会长成猛兽的“可爱虎崽”，强调超级智能一旦出现，人类可能因智力差距而完全失去控制。他强烈呼吁社会正视AI可能带来的生存威胁，并对当前主要AI公司在安全研究上的投入不足表示严重关切。

金句

• "It's very hard to get your head round the fact that we're at this very, very special point in history. We're in for a relatively short time, everything might totally change." ——很难让你真正理解，我们正处在一个历史上非常、非常特殊的时刻。在相对较短的时间里，一切都可能彻底改变。
• "We like somebody who has this really cute tiger cub. It's just such a cute tiger cub. Now, unless you can be very sure that it's not going to want to kill you when it's grown up, you should worry." ——我们就像某人养了一只非常可爱的虎崽。它真是只可爱的虎崽。现在，除非你非常确定它长大后不会想杀死你，否则你应该感到担忧。
• "I think if the public realized what was happening, they will put a lot of pressure on governments to insist that the AI companies develop this more safely." ——我认为，如果公众意识到正在发生的事情，他们将向政府施加巨大压力，要求AI公司更安全地开发技术。
• "Previously what the large language models had to do is they spit out one word at a time and that would be it. Now, they spit out words and they're looking at the words they spit out... that's called chain of thought reasoning." ——以前大语言模型要做的就是一次吐出一个词，仅此而已。现在，它们吐出词语并观察自己吐出的词语……这被称为链式思考。
• "If you can get that gradient information, we know that we can then train a big system that starts with random weights to do wonderful things. The brain needs to get information like that, and it probably gets it in a different way." ——如果你能得到那个梯度信息，我们知道我们就能训练一个从随机权重开始的大系统去做奇妙的事情。大脑需要得到那样的信息，而且它获取信息的方式可能不同。

原文链接

查看完整访谈