Google DeepMind CEO Demis Hassabis: AI's Next Brea

章节 1：AI发展的现状与质疑

📝 本节摘要：

本节作为访谈的开篇，主持人Alex回顾了过去一年中业界关于“大语言模型（LLMs）是否已触及天花板”的质疑。Google DeepMind CEO Demis Hassabis对此作出回应，表示DeepMind内部从未有过此类担忧。他反驳了“数据枯竭”将导致停滞的观点，指出通过合成数据以及优化现有架构，AI技术仍有巨大的提升空间（Headroom），目前的进展远未停止。

[原文] [Alex Kantrowitz]: google DeepMind CEO Dennis Asabis joins us to talk about the path from here to AGI when Google's AI glasses are coming and whether the pace of AI progress can keep up at this rate that's coming up right after this welcome to a special edition of Big Technology Podcast from Davos i'm Alex Caneritz and I'm joined today by a special guest Dennis Estabis the CEO of Google Deep Mind demis welcome back to the show

[译文] [Alex Kantrowitz]: Google DeepMind CEO Demis Hassabis（德米斯·哈萨比斯）做客本期节目，我们将探讨从现状通往AGI（通用人工智能）的路径、Google智能眼镜的发布时间，以及AI的发展速度能否保持当前的频率。精彩内容稍后即呈。欢迎来到在大达沃斯（Davos）录制的《Big Technology Podcast》特别版，我是Alex Kantrowitz。今天我邀请到的特别嘉宾是Google DeepMind的CEO Demis Hassabis。Demis，欢迎回到节目。

[原文] [Demis Hassabis]: it's great to be here

[译文] [Demis Hassabis]: 很高兴来到这里。

[原文] [Alex Kantrowitz]: a year ago there were real questions about whether AI progress was tailing off it was in fashion to ask whether LLMs were going to hit a wall and those questions seem like they've been settled there's been a tremendous amount of progress over the past year uh could you tell us what specifically has happened that's gotten the AI industry from that moment of question last year to the point that it is today

[译文] [Alex Kantrowitz]: 一年前，人们曾真正质疑AI的进步是否正在放缓；当时很流行讨论大语言模型（LLMs）是否会遭遇瓶颈（hit a wall）。而如今这些问题似乎已经有了定论，过去一年里取得了巨大的进步。你能否具体告诉我们，发生了什么让AI行业从去年的质疑时刻发展到了今天的地步？

[原文] [Demis Hassabis]: well I'm for us internally we were never questioning that just to be clear I think we we've always been seeing um great improvements um so we were a bit puzzled by by why there was this question in in the air i mean some of it was to do we people were worried about data running out um and there is you know some some truth in that is all the data had been used can we create synthetic data that's going to be useful to learn from but actually it turns out you can ring more uh more juice out of the existing architectures and data so there was plenty of room I think and we're still seeing that in both the pre-training the post-training and the thinking paradigms and also the way that they all kind of uh uh fit together so I think there's still plenty of headroom there just uh with the techniques we already know about and and tweaking and and and kind of innovating on top of that,

[译文] [Demis Hassabis]: 嗯，对于我们内部来说，为了澄清一下，我们从未怀疑过这一点。我认为我们一直都在目睹巨大的改进，所以我们对当时空气中弥漫的这种质疑感到有些困惑。我的意思是，这部分原因在于人们担心数据会耗尽。这确实有一定道理，比如是否所有数据都已被使用？我们要不要创造有用的合成数据（synthetic data）来供模型学习？但实际上，结果证明你可以从现有的架构和数据中榨取更多的“汁水”（价值），所以其实还有很大的空间。我认为无论是在预训练（pre-training）、后训练（post-training）还是思维范式（thinking paradigms）上，以及它们如何相互结合的方式上，我们仍然看到了这种进步。所以我认为，仅凭我们已经掌握的技术，通过微调和在此基础上的创新，仍然有着充足的上升空间（headroom）。