AI 解过的最难问题：DeepMind CEO 访谈

Cleo Abram

非常感谢你做这件事。

Thanks so much for doing this.

Demis Hassabis

太好了，谢谢。

It’s great. Appreciate it.

Cleo Abram

你已经知道 Huge Conversations 是另一种访谈。我不会问财务，也不会问管理风格，别处都覆盖过了。我希望这场对话更像我们现场一起做一期 explainer。我带了道具。这本来不是要玩 Jenga 的——每块积木代表一个项目或模型，我想谈它们怎么拼在一起。布置的时候我们开始拿它们玩 Jenga，比我计划的好玩太多。我也知道你喜欢游戏。

You already know that Huge Conversations is a different kind of interview. I’m not going to ask you about financials, I’m not going to ask you about your management style. All well covered elsewhere. What I’m hoping to do in this conversation is think about it more like an explainer that we’re making live together. And I have some props. This was not actually meant to be a Jenga game. Each block represents a project or a model and I want to talk about them and how they fit together. They were meant to be visual aids, but as we were setting up, we started playing Jenga with them and it turned out to be way more fun than anything I had planned. Also, I know that you like games.

Demis Hassabis

是，我爱游戏。所以这很好。访谈里还是第一次。

Yes, I love games. So this is great. First in an interview anyway.

Cleo Abram

我希望我们一起做这期 explainer，帮人看清 AI 现在到底在发生什么，以及你看到的未来。你这场对话想做什么？

My hope in this conversation is to make this explainer together and to help people see what’s happening right now in AI really, and what is the future that you see coming. What are you hoping to do in this conversation?

Demis Hassabis

我三十多年前进 AI，很大原因就是推进科学和医学。我一直觉得 AI 可能是做这件事的终极工具。希望今天能聊这个——这是我把 AI 用在什么上的热情所在。当然它也能用在很多别的地方。

A lot of the reasons that I got into AI 30-plus years ago now is to advance science and medicine, and I’ve always thought of AI as potentially the ultimate tool to do that. So I’m hoping we’re going to talk about that today, and really that’s been my passion for what to apply AI to. But of course it can be applied to many things.

Cleo Abram

这盘 Jenga 里很多积木是大家听说过的，这块是 Gemini。但我会争辩：AI 正在有意义地塑造人们生活的方式，大多是他们看不见的。所以我想从你拿诺贝尔奖的那个项目开始。AlphaFold。

In this Jenga game that we have, a lot of these are blocks that people will have heard of. This one is Gemini. But I would argue that the ways in which AI is meaningfully shaping people’s lives most are the things that are invisible to them most of the time. So I want to start by talking about the project that you won the Nobel Prize for. AlphaFold.

Cleo Abram

我想把 AlphaFold 连同所有戏剧性一起讲，因为有人可能没听过，但我会很快进到这类科学的最前沿。这么多问题里，你为什么决定啃这个？

I want to tell the story of AlphaFold with all of its drama because some people might not have heard it. But then I want to get really quickly to the cutting edge of this sort of category of science. Why did you decide to tackle this problem out of all of the many?

Demis Hassabis

我其实是剑桥本科时碰到的。很多生物学朋友，其中一个特别迷蛋白质折叠问题。身体里的一切都靠蛋白质，三维结构部分决定功能。蛋白质折叠问题就是：能不能只从一维氨基酸序列预测这个三维结构。这是五十年的重大挑战。

I came across it actually as an undergrad in Cambridge. I had a lot of biologist friends, and one of them specifically was obsessed with what’s called the protein folding problem. Proteins are what everything in your body relies on. What’s important about them is their 3D structure. They fold up into 3D structures and those structures determine what function they have, or partially determine it. The protein folding problem is: can you predict this 3D structure just from the one-dimensional amino acid sequence? That’s the 50-year grand challenge of protein folding.

我爱挑战、爱谜题。有人把它说成生物学的费马大定理，谁能不感兴趣？而且我第一次听到时就觉得，这种问题有一天适合 AI——尽管那是九十年代末，我们还没有任何能做这个的 AI。最后是冲击：若破解了，会打开下游研究，尤其是药物发现和理解疾病。我觉得把 AI 用在改善人类健康上，是最重要的事。

I love challenges, I love puzzles. So I couldn’t resist it from a scientific point of view as this probably is described to me as the equivalent of Fermat’s Last Theorem but for biology. When I first heard it, I thought the kind of problem it was would be suitable for AI one day, even though this is in the late ’90s, we didn’t have any kind of AI that would be possible to work on this. And then the final thing was just the impact it would make if you cracked it, because it would open up all these downstream possibilities for research and especially in things like drug discovery and understanding disease. I think the most important thing to apply AI to is improving human health.

Cleo Abram

原因是：直到那时，要开发新药，得花几十万美元、数年人力，用 X 射线才能搞清一个蛋白质的结构。我们搞清楚了一些，但又慢又贵。

The reason that this would be huge for human health is that up until now, in order to develop new medicines, we’d have to spend hundreds of thousands of dollars and years of human effort to find out the structure of a single protein by shooting X-rays at it. We had figured out some protein structures, but it was slow and expensive.

Cleo Abram

我跳过你和团队海量的辛苦。用我问问题的方式，大家会很明显知道你们解了。有一个时刻你意识到它真的有用，解了现代医学里最重要的未解问题之一。那是 2021 年。你在开会。我特别高兴那次会有摄像机，那是我见过最不可思议的瞬间之一。你们在谈给科学家搭一个提交蛋白质、拿回折叠结果的服务器，然后有人提出完全不同的想法。你能走一遍那次会吗？你的反应太精彩了，我想知道你当时在想什么。

I’m skipping over an enormous amount of hard work here by you and your team. But I think by the way that I’m asking the questions, it is very obvious to people that you solved it. There’s this moment where you realize that it is genuinely useful and you have solved what had been called one of the most important unsolved problems in modern medicine. And it’s 2021. You’re in a meeting. I am so glad that there was a camera in this meeting because it is one of the most incredible moments I have ever seen. I think you’re talking with your team about setting up a system where scientists could send in a request for a specific protein, like a website, and then get the protein folded. And then someone else has a very different idea. Can you walk me through what happens in that meeting? And then your reaction is incredible and I really want to know what you were thinking.

Demis Hassabis

摄像机碰巧在那次会上，很疯，他们很少跟我们，偏偏赶上那天。这类预测模型的传统做法是搭服务器，别的科学家把序列发过来：「我对这个蛋白质感兴趣，把预测结构寄回来。」整个领域四十多年都这么干，因为多数算法很慢，也许要几天，再邮件把结构寄回去。

It was funny that the cameras happened to be in that particular meeting. They very rarely followed us, but it was fit for that meeting. Normally for these sorts of prediction models, the traditional thing is you set up a server and then other scientists send you their protein sequences and say, I’m interested in this protein, can you send me back the predicted structure. That’s how it’s been done in the whole field for the last 40-plus years. The reason is most of the prediction algorithms are quite slow. Maybe it would take a few days and then you’d email back the structure.

但那次会上我意识到：我们不但准，而且快，几秒就能折好。我在心里做信封背面计算——科学已知、自然界已知的蛋白质有多少？2 亿。我们有多少计算机？如果我们每 10 秒折一个，会中那次会、我还在摆弄手机的时候意识到：一年内能做完。那为什么还要搭服务器、数据库、邮件客户端？我们可以自己把任何人可能要的全部折完，放到某个数据库上，免费给全世界科学家用。突然就想到：我们应该直接这么干。

Once I realized in that meeting how quickly — not only how accurately we could fold the proteins but how quickly, in a matter of seconds — I was just doing the back-of-the-envelope calculation. How many proteins are there known to science, known in nature? 200 million. And then how many computers do we have? If we folded one every 10 seconds, I realized in the middle of that meeting while I was fiddling on my phone that it would be possible in a year. So why go to all the effort of building the servers and the databases and the email client when we could just actually fold everything ourselves, everything anyone could ever request and ever want, and then put it on a database somewhere for free for all the scientists in the world to use. It just suddenly hit me. We should just do that.

脑子里那些事突然接到一起，这才是显然该做的，而且可能比搭服务器还省事，反而省时间。

Suddenly all these things must have been going on in the back of my mind and I suddenly realized that that would be the obvious thing to do and it would be actually probably less effort than standing up the server. So actually it would save us time.

Cleo Abram

能不能说，我们现在已经预测了几乎所有科学已知蛋白质的结构？

Is it correct to say that we have now predicted the structure of almost all proteins known to science?

Demis Hassabis

是。而且我们还在更新。每次有人从海里舀一桶水，里面有大量不同生物，他们全部测序。测序技术比人类基因组测序时已经好了好几个数量级，卡住的是结构生物学。有了 AlphaFold 2 这种计算资源，我们能跟上：「这里有一百万条新奇生物的遗传序列，结构在这儿。」欧洲生物信息研究所的一个小团队每年更新当年新发现的序列，我们始终在前沿，大体知道这些蛋白质结构长什么样。

Yes. And we keep updating it. Every time somebody scoops a pail out of the ocean somewhere and there’s loads of different types of organisms in that bucket of seawater and then they sequence them all. Sequencing technology has improved many orders of magnitude since the human genome was sequenced. The problem was structural biology was far lagging behind genetic sequencing. Now with computational resources like AlphaFold 2, we can actually keep up: here’s a new million genetic sequences from some new strange organisms — oh, here are the structures. We have a small team at the European Bioinformatics Institute that keeps updating every year all the new sequences that have been found that year. We’re now always at the cutting edge. We know what all of these different protein structures mostly look like.

对研究比较冷门生物的人尤其惊人，比如小麦。我发现很多植物的基因组数据比哺乳动物和人类还多，好像有多份基因组拷贝。植物科学朋友没有人类基因组那种资源，但作物对人类仍很重要。现在他们能直接跳到自己真正关心的科学问题上，比如让作物更抗气候变化，而不被结晶蛋白质拖住。

It’s especially amazing for researchers that work on slightly more obscure organisms. For example, wheat. I found out that a lot of plants have way more genomic data than mammals and humans, which is very strange. They seem to have multiple copies of their genome. My plant scientist friends don’t have the resources like with the human genome. But some of these more obscure organisms that are still really important for humanity, like crops, now we’re able to immediately jump to the science around what they want to do with the proteins — maybe help them be more resilient to climate change — and they can jump straight to the problem they’re actually interested in rather than getting bogged down with trying to crystallize the proteins.

另一块是被忽视的疾病：疟疾、查加斯病、利什曼病，影响全世界数亿人，但因为发生在更穷的地方，大药厂没多少钱可赚，研究被忽视。了不起的非营利组织在做，但没钱没资源。把疟原虫相关蛋白质结构给他们，他们就能直接进药物发现阶段。

Another boon is for researchers who work on neglected diseases that affect primarily the more developing parts of the world — malaria or Chagas disease or leishmaniasis. They affect hundreds of millions of people around the world, but there’s not a lot of money in that if big pharma tried to research that, because they’re in the poorer parts of the world. So they tend to be neglected. There are these amazing non-profit organizations that do the research, but they don’t have a lot of money or resources, so giving them the structures of the proteins involved in, say, the malaria virus is a huge boon, because they can go straight to the drug discovery phase.

Cleo Abram

全世界科学家都用上 AlphaFold 了，地图会亮起来，但我不太容易找到一个例子：科学家用了它，药的流程加快了，变成我现在能吃的药。你最喜欢、观众能懂的例子是什么？

There’s this moment where scientists all around the world have access to AlphaFold. You can see the map lights up. But I wasn’t easily able to figure out a great example of a scientist using AlphaFold and then that speeding up a drug process that results in a drug that I could now take. What is your favorite example of a scientist using AlphaFold for something the audience might understand or have seen?

Demis Hassabis

现在超过 300 万科学家在用 AlphaFold。我们觉得差不多每个生物学家都在用。一位药企科学家跟我说：从今往后几乎每一种新药，大概都会在流程里用到 AlphaFold。这很让人震惊。药物发现仍要时间，我们多数还在基础生物学阶段：理解疾病、打哪个蛋白质、是不是对的机制。据我理解，有些药已经进了临床试验，希望几年后能看到几十种至少部分被 AlphaFold 帮过忙的药。

Over 3 million scientists are now using AlphaFold. We think it’s pretty much every biologist in the world at this point. One scientist at a pharma company said to me that almost every drug developed from now on will have probably used AlphaFold in its process, which is sort of mind-blowing. It still takes time with drug discovery. We’re still mostly in the fundamental biology stage of understanding the disease, what is the protein we’re targeting, is that the right biological mechanism. As I understand it, some of these drugs are now in the clinical trials phase, and hopefully we’ll see in a few years’ time a whole dozens of drugs that were partially helped by at least AlphaFold.

我目前最喜欢的突破是核孔复合体：身体里最大的蛋白质之一，像甜甜圈环，开合控制营养进出细胞核。它又大又复杂，很难结晶看到。AlphaFold 放出大约六个月到一年后，有团队把它和实验数据合在一起，终于搞清了这个门户蛋白质的美丽形状。那对我来说太惊人了。

In terms of my favorite breakthrough so far with the help of AlphaFold: there’s this protein called the nuclear pore complex, and it’s one of the biggest proteins in the body. What it does is a very important job. It’s basically the gateway that opens up and closes to let nutrients come in and out of the cell nucleus. It’s like a big donut ring that opens and closes. We didn’t know until very recently what the structure of this was because it’s so big and complicated, pretty hard to crystallize and actually see. Pretty much six months or a year after we put AlphaFold out, some teams used it along with experimental data to finally work out what this beautiful shape was of this gateway protein. That was amazing to me.

我们自己还分拆了 Isomorphic Labs，把 AlphaFold 当作拼图的一块，大幅加速药物发现。平均一种药大概要 10 年，失败率极高，只有大约 10% 能走完所有临床阶段。要改善人类健康，必须大幅改进这个，我认为要用 in silico 方法，AlphaFold 2 是其中一块。知道结构只是药物发现的一小部分，还要大量化学：该设计什么化合物去结合。Isomorphic 在做相邻系统，跟更先进的 AlphaFold——你可以叫它 AlphaFold 3、4——端到端做副作用极小、非常有效的药。我们现在大概有 18、19 个药物项目，从心血管到癌症到免疫，几乎每个治疗领域最终都该能帮上。

We ourselves spun out a new company, Isomorphic Labs, that tries to build on AlphaFold and uses it as one of the pieces of the puzzle to massively speed up drug discovery. On average it takes like 10 years to develop a drug. Huge failure rates. Only about 10% of drugs actually get through all the clinical stages. We need to vastly improve that if we want to improve human health, and I think the way to do that is by using in silico methods, AlphaFold 2 being one of those components. Knowing the structure of a protein is only one small part of the drug discovery process. You need a lot of chemistry, like what compound should you design to bind to it. We’re trying at Isomorphic to build adjacent systems that work with AlphaFold — more advanced AlphaFold, AlphaFold 3, AlphaFold 4, you could call it — and then end-to-end create these drugs that have very minimal side effects and are incredibly effective. We’re working on I think at this point like 18, 19 different drug programs across the gamut, from cardiovascular heart disease to cancer to immunology. Eventually these types of technologies should be able to help across almost every therapeutic area.

Cleo Abram

准备这场访谈时，我背景采访了同获诺贝尔奖的 John Jumper。他强调这只是药物发现更大问题的一部分。所以我们来到今天的前沿。现在的前沿是什么？

In prep for this, I did a background interview with your fellow Nobel Prize winner, John Jumper. He really stressed that it’s one part of a larger problem of drug discovery. And so that brings us to the cutting edge today. What is the cutting edge now?

Demis Hassabis

我们在建很多能拼在一起的组件。AlphaFold 是枢纽之一，是蛋白质结构。你知道形状、知道行使功能的那一块，若要阻断或增强它，就知道该结合表面的哪一部分。然后要发现能接到正确位置的化合物，还要知道结合有多强。更重要的是：别接到别的东西上，那就是毒性、副作用。

We’re building many different components that can kind of go together. AlphaFold is one of the linchpins, so that’s the structure of the protein. Let’s say you understand what the shape of the protein is, and which bit does its function. If you want to block the effect of that protein or enhance it, you know which part of the protein surface you have to bind to. Now you have to discover a chemical compound that will attach to the right place, and how strong will it attach. Even more important is not just will it attach to the thing you’re interested in — make sure it doesn’t attach to other things, because if it does, that would be toxicity, side effects.

有了这些算法工具，我们可以做虚拟筛选：AI 设计的化合物，预测它和蛋白质表面结合有多强；然后几小时内，检查这个化合物怎么接到人体另外 2 万种蛋白质上。几分钟就能做完，再不断改化合物，让副作用越来越少、对目标越来越强。这是一套自改进、自修改过程，in silico 极快。最后才进湿实验室验证预测。这样可以搜索几千倍、将来也许百万倍的化合物，只在终点检查。比今天在湿实验室里搜索高效得多。

Because we have all of these amazing algorithmic tools, we can do a virtual screen. Here’s a compound one of our AI systems has designed. This is our prediction of how strong it binds to the protein surface. Then we can check that very quickly, like in a matter of hours, how that particular compound attaches to any of the other 20,000 proteins in the human body. We can do it within a few minutes, and then keep modifying the compound so that it has less and less side effects, ideally none on any of the other proteins, but increasingly strong effect on the one that you want. I’ve just outlined a self-improvement process or self-modification process. This is extremely fast and efficient if you can do it in silico. Only at the final stage do you check it in the wet lab. You still have to validate it. You can search thousands of times more compounds, or maybe even millions at some point, more quickly and efficiently that way, and then just at the end check that they’re correct. That’s so much more efficient than doing the search in the wet lab, which is what effectively is done today.

Cleo Abram

我特别喜欢的还有 AlphaGenome。我又联系了一位诺贝尔奖得主——Jennifer Doudna，她上过我的节目，带了一个问题。她说：CRISPR 现在几乎能打任何 DNA 序列，但对大多数遗传病，我们仍不完全知道哪些 DNA 变化真正在驱动问题，尤其是不编码蛋白质的那 98% 基因组。有了 AlphaGenome 开始解码那 98%，我们离 AI 能可靠指出导致病人疾病的精确遗传改变、好让 CRISPR 去修，还有多近？

One of my favorites also is AlphaGenome. I reached out to yet another Nobel Prize winner, Dr. Jennifer Doudna, who I’ve had on the show. She sent a question. CRISPR, the gene-editing technology that she pioneered, can now target nearly any DNA sequence. But for most genetic diseases, we still don’t fully understand which changes in the DNA are actually driving the problem, especially in the 98% of the genome that doesn’t code for proteins. With tools like AlphaGenome starting to decode that 98%, how close do you think we are to the moment where AI can reliably point to the exact genetic change causing a patient’s disease so that technologies like CRISPR can fix it?

Demis Hassabis

问得太好了。我以前跟她讨论过，非常令人兴奋。AlphaGenome 就是这种技术：吃进很长的遗传序列，预测如果你在某个单碱基位置做突变，会是致病的有害突变，还是良性。我们刚发布的 AlphaGenome 是全世界预测这件事最好的系统。若未来版本准到真能知道「这个突变加上那个」——多基因疾病、突变级联更难，但正适合 AI——然后有一天可以用 CRISPR 进去修那个突变。AlphaGenome 和 CRISPR 的组合会极强，希望有一天能跟 Jennifer 他们合作。

What an awesome question. I’ve discussed this with her actually in the past and it is really exciting. AlphaGenome takes the big long genetic sequences and then it tries to predict, if you had made a mutation to this particular single letter, single position in the genetic sequence, will that be a harmful mutation that might cause disease or is it benign. AlphaGenome which we just released is the best system in the world for predicting that. That’s exactly what you then want. It’s still not probably good enough yet, but you can imagine a future version of AlphaGenomes accurate enough to really know: that particular mutation in combination with this other one. The hard part is multigenic diseases where there are cascades of mutations. Those are even harder to detect, but actually perfect for AI to try and help with. Then you could go in with something like CRISPR maybe one day and fix that mutation. A combination of things like AlphaGenome and CRISPR could be incredibly powerful and hopefully one day will be collaborating with the likes of Jennifer on that.

Cleo Abram

去年你对《卫报》说：如果由我决定，我会把 AI 在实验室里多留一阵，多做像 AlphaFold 这样的事，也许治好癌症之类。外面看起来的故事是：你创办 DeepMind，使命是解决智力再用它解决其他一切；卖给 Google 是因为他们让你们自由探索科学；很长时间那是你的唯一焦点。然后 ChatGPT 出来，Google 进入 code red，你成为所有 Google AI 的负责人，包括你以前没花那么多时间的消费产品。从远处看，这镜像了过去两年 AI 的巨变。那次转变得到了什么、失去了什么？

Last year you said something to the Guardian that I found really interesting. You said that if I’d had my way, I would have left AI in the lab for longer, and the quote is done more things like AlphaFold, maybe cured cancer or something like that. From the outside, it looks like you found DeepMind with the mission to solve intelligence and use it to solve everything else. Then you sell to Google specifically because they will allow the freedom to explore science. For a long time that’s your exclusive focus. Then ChatGPT comes out, Google goes code red, and you become the head of all Google AI including the consumer products. Watching that from afar, it mirrors somewhat the larger experience of AI, this incredible change in the last couple of years. What was gained and what was lost in that change?

Demis Hassabis

你描述的正是里面的感觉。对我来说，AI 最好的用例是改善人类健康、加速科学发现。我进 AI，是因为对现实的本质、意识的本质这些大问题感兴趣，觉得我们需要一个工具，帮哪怕最好的科学家消化数据和信息、找到洞察。这件事正在发生，AlphaFold 是我们第一次、目前最好的表达。还有很多类似问题。考虑到 AGI 有多重要、可能是人类历史上最变革的技术，我觉得通向终点的后半程最好用科学方法，非常小心、精确、严谨，最好的科学家用 CERN 那样的方式合作，确保每一步我们都理解。也许要多十年甚至二十年，但以这件事的体量，那样才说得通。

That’s exactly right what you described is sort of how it felt from the inside too. For me the best use case of AI was to improve human health and accelerate scientific discovery. I got into AI in the first place because I was interested in all the big questions in the world, the nature of reality, nature of consciousness, and I felt we needed a tool to help even the best scientists make sense of the amount of data and information out there and find insights in that. That’s happening, and obviously AlphaFold was our first and so far best expression of that. Given how important AGI is and how transformative a technology it is, maybe the most transformative one in human history, I thought it would be best to approach the latter stages of building it which we’re in now using the scientific method very carefully, very precisely, very thoughtfully and rigorously, with all the best scientists collaborating in kind of a CERN-like effort on making sure we understood each step as we got to the final goal of building AGI. That might take a lot longer, maybe a decade even two decades longer, but I think that would make sense given the enormity of what we’re dealing with.

另一个想法是：不必等 AGI 到来才开始受益。可以用更专门的系统，借用为 AGI 开发的通用算法，但它们本身不是通用智力，是窄 AI，像 AlphaFold 只做一件事。我们可以一边用科学方法建 AGI，一边做很多种 AlphaFold 和 Isomorphic，让人类先从治癌症、新能源、新材料里受益。二三十年前我出发时，觉得那会是理想展开方式。

My other idea was we don’t have to wait till AGI arrives to start getting the benefits of AI. We could use more specialized systems that maybe make use of the general technologies, the general algorithms we’re developing for AGI, but are not in themselves general intelligences. They’re narrow AIs if you want to call them, like AlphaFold which does a specific purpose and only that purpose. We could create many types of AlphaFolds and Isomorphics while we’re building AGI in this careful scientific way, and humanity could benefit from the proceeds of that like cures for cancer or maybe new energy sources or new materials. Looking at this from 20–30 years ago when I started out, that would have been the ideal way for it to play out in my opinion.

Demis Hassabis

没有那样发生，因为技术不可预测。事实证明语言比我们所有人——包括乐观的人——预期的容易得多。现在想起来很滑稽：语言、概念、抽象，这些基础模型如 Gemini 现在做得极好，我们以为还要一两个或三个突破。结果是我 Google 同事发明的 transformer，再加上一些强化学习，就足以破解语言。

Now it didn’t happen like that because technology is unpredictable. It turns out that things like language were a lot easier than we were all expecting, even those of us who were obviously optimists. It seems funny to think of it now, but language and concepts and abstractions, things that the current foundation models like Gemini do incredibly well, we thought that maybe there would be one or two or three more breakthroughs needed before we could get there. But it turned out transformers which my Google colleagues invented and some reinforcement learning as well on top was enough to crack things like language.

我们和其他领先实验室也在玩这些。当然 ChatGPT，公平地说 OpenAI，他们把它规模化并放出去。我想连他们都说那有点像研究实验，没意识到会这么病毒式传播，我们谁也没想到。当时我们有相当接近的系统。你离技术太近，非常清楚它做不到的事、幻觉那些我们现在仍在改进、仍没完全修好的缺陷，意识不到外面的人即使用幻觉的系统也会觉得有用：总结、头脑风暴，今天人人都用聊天机器人做这些。

We were sort of playing around with that with the other leading labs, but of course with ChatGPT, and fair play to OpenAI, they scaled it and then they put it out there. I think even they say it was kind of a research experiment. They didn’t realize it would go so viral and I think none of us did. We had sort of fairly equivalent systems at the time. When you’re building that technology, you are so close to it, you’re very aware of the things it can’t do, the flaws it has, and you don’t realize that actually people out there would find use even though it was hallucinating and doing other things that we’re obviously all still trying to improve on now, still not completely fixed. There’s still interesting use cases like summarizing things or brainstorming that people use, everyone uses chatbots for today.

坏处是我们锁进了凶猛的商业压力竞赛，再加上美中地缘竞赛，多层压力要快。好处当然是进步极快，对好的用例是好事。第二，观众用到的最前沿 AI，也许只落后实验室三到六个月，这很疯狂，也在民主化 AI，让社会用增量步骤去习惯，而不是某天突然 AGI。第三，系统不被几百万人压测，你无法完全理解它。室内测试再好，也比不上几百万人去试、看什么浮上来。所以有正有负。不是我多年前梦想的那种哲学沉思。我们必须面对找到的世界，尽量做得最好：推进前沿，同时尽量负责任地部署 Gemini 和 AlphaFold 这类非常强的技术。

The downside is we’re in this sort of ferocious commercial pressure race that everyone’s locked into currently. And then on top of that there’s geopolitical issues like the US-China race. Multiple levels of pressure to move fast. The benefit is of course you get faster progress. The progress is just like at lightning speed these days, so that’s good for all the good use cases. The second benefit is that everybody, all of the viewers out there, you’re all getting to use the most cutting-edge AI technology perhaps only three to six months behind what is actually in the labs. That’s kind of mind-blowing. It’s also great because it’s democratizing AI. It’s giving everyone a feeling for what it’s like to interact with cutting-edge AI. That’s good for society to start normalizing itself to what is going to be an enormous change, better that we get to sample that in incremental steps rather than it’s just a shock to the system. The final thing on the benefit side is you can’t really fully understand your systems until they’re stress tested by millions of people. It doesn’t matter how good your in-house testing is. Millions of smart people trying out things and then seeing what bubbles to the top is really important for building more robust systems. So I think there’s positives and negatives about the way it’s gone. It’s not the way I dreamed about years ago where we would be contemplating this philosophically. We have to deal with the world as we find it and make the best of that. We try to do that by advancing the frontier, but also trying to be as responsible as we can as we deploy these very powerful technologies like Gemini and AlphaFold.

Cleo Abram

同时还有另一个故事。我想讲 AI 出人意料地有创造力。2016 年 3 月 10 日，一位非常著名的围棋选手坐下，对阵你们设计的系统。那时计算机已经在各种游戏里赢过人类，但围棋有意思：可能的走法比宇宙中的原子还多。然后你们的系统走了一步极不可能由人类想出的棋——第 37 手。李世石坐在那儿，头埋进手里。对你们这种人来说，那是看到一种非常不同的创造力的时刻。一类是给大量数据做新预测；另一类是不给数据、给规则，像数学、物理、围棋，有巨大的创造空间。那一步发生时你在哪？你看见了怎样的未来？

There’s another story happening at the same time. I want to tell a story about AI being very creative, unexpectedly creative. Let’s go back to March 10, 2016. There’s a very famous Go player that sits down to play against a system that you designed. At this point computers have beat humans at all kinds of games, but Go is really interesting because there are more potential moves in Go than atoms in the universe. Then your system makes a move that is so surprising because it is incredibly unlikely that a human would figure out a move like that, move 37. You see Lee Sedol sitting there. He’s just got this shock on his face. He’s got his head in his hands. It really was this moment where I think people like yourself saw ahead to the creativity that we would find in AI systems that are very different than the systems that we’ve talked about so far. There’s a category where you’re giving a huge amount of data and you’re asking to make new predictions. Then there’s a category where you’re not giving data, you’re giving rules. Like with math or physics or games like Go. And it has this incredible opportunity for creativity. Where were you when that moment happened? And what future did you see ahead?

Demis Hassabis

那是不可思议的时刻，差不多正好十年前，感觉像一个世纪。在很多方面那是现代 AI 时代的黎明。此前有很多能当世界冠军的 AI，比如象棋，但那是专家系统：聪明程序员加象棋特级大师，把知识蒸馏成规则，再用 IBM Deep Blue 那种超算蛮力赢卡斯帕罗夫。系统只是笨拙地执行那些启发式。

It was an incredible moment, almost exactly 10 years ago now, which feels like a century ago. I think in many ways it was the dawn of the modern AI era. Until that point there were many AI programs that could be world champions at games, things like chess, but they were done with what’s called expert systems. A team of smart programmers with a team of smart chess grandmasters came together, tried to distill the knowledge into a set of rules, kind of a brute force system that would use a lot of compute like IBM did with Deep Blue to beat Garry Kasparov. The system would sort of dumbly execute those rules and heuristics.

九十年代我读本科时看到这个就不满意。我不觉得那是真正的 AI。Deep Blue 是象棋世界冠军水平，但别的什么都不会：不会语言、不会机器人，连严格更简单的井字棋都不会。没有人类特级大师会学不会井字棋。智力定义显然不对：它没有泛化，也没有学习，答案是被给进去的。智力不在系统里，在特级大师和程序员脑子里。他们解了象棋，程序只是执行解。

For me that was not satisfactory when I saw that in the ’90s. I was doing my undergrad at the time. I didn’t feel like that was proper AI because that system, let’s take Deep Blue, it’s world champion level at chess, but it can’t do anything else. Not only can’t it do language and robotics, it can’t even play a strictly simpler game like tic-tac-toe. Something’s obviously not quite right about the definition of intelligence. You could imagine a human grandmaster not being able to learn how to play tic-tac-toe. It would make no sense because it’s strictly simpler. There’s something wrong about its generalization capability and the fact that it didn’t learn. It was just given the answer. If you could ask where the intelligence resided — it wasn’t in the system, it was in the minds of the chess grandmasters and the programmers. They solved the problem of chess and then implemented the solution. The program just dumbly executed the solution.

围棋是游戏的最后边疆，人类发明过最复杂的游戏，也是最古老的，非常美。在中国、日本、韩国，它占据象棋那种智识位置，但更直觉、更艺术：漂亮的图案往往就是强。顶级棋手会说它封装了宇宙的奥秘。原始复杂度是 10 的 170 次方种盘面，比宇宙原子还多，没法用象棋那种蛮力。而且因为太直觉，也没有容易封装给机器的规则。你问围棋大师为什么走那儿，他们会说「感觉对」；象棋手永远不会这么说，他们会讲计算。那种直觉没法直接编程。所以它是我们 DeepMind 早期深强化学习的完美试验场：系统能不能直接从经验里学习。

Go is the final frontier for games. It’s the most complex game humans have ever invented. It’s also the oldest game. It’s also very beautiful. In China and Japan, Korea, they play it instead of chess. It’s a much more intuitive game, sort of artistic. You play patterns that look beautiful and they turn out to be really strong, which is why the game has a little bit of a mystical element. The top Go players would say it encapsulates the mysteries of the universe. Just its raw complexity has more possible board positions, 10 to the power 170, than there are atoms in the universe. There’s no way you can brute force it in the way that we did with chess. Furthermore, because the game’s so intuitive, there aren’t really these rules that you can encapsulate easily. When you talk to a Go master, unlike a chess master, they’ll tell you things like, why did you play there? They’ll say, it felt right. A chess player will never say that. They would say I did it because I’m calculating this. That intuitive feeling is obviously very hard to encapsulate. You can’t really program that directly. So it’s the perfect proving ground for these new techniques that we were pioneering in the early days of DeepMind of deep reinforcement learning. Can you build systems that learn from themselves directly from experience?

AlphaGo 先看人类在网上下过的棋，学习人类会走的类型，再叠上蒙特卡洛树搜索，让它从人类已知出发，发现知识树的新枝。我们希望会发生的就是这个。那场比赛全世界约 2 亿人看，我们 4-1 赢了，这是主目标。但第二局走出了著名的第 37 手：棋盘第五线，开局很早，围棋里那是大忌，师傅会打你手腕。它不仅是好棋，一百、两百手之后它在正确的位置上，像预知地把子放在那儿。那是决定胜负的关键一手。它改变了所有围棋手的下法。对我来说，那是我等了六年的信号：我们造的学习系统能做到别的系统做不到的事——游戏 AI 的珠穆朗玛。而且赢的方式带有第 37 手这种新想法。那就是我们可以把它转向 AlphaFold 这类科学问题的信号。

AlphaGo started by looking at all the games on the internet that humans have played and learning the types of moves humans would do, but then we overlaid it with a Monte Carlo tree search that allowed it to discover new branches of the tree of knowledge in Go, starting with what humans knew and then going beyond that. That’s what we hoped was going to happen. That match ended up being watched by 200 million people around the world. Not only did we win the match 4-1, that was the main objective, but in game two specifically, it played this famous move 37. It was on the fifth line of the board and early in the game, and it’s sort of a big no-no to do that in Go. If you were being taught by a Go master, they would slap your wrist. Not only was that a great move, it ended up winning the game for AlphaGo. 100 moves, 200 moves later, it was in the right place, as if it sort of presciently put the stone there. It was the critical move. Obviously it’s changed the way all Go players play Go, but for me it was the moment I’d been waiting for. We’d already spent six years building these types of learning systems that could achieve something no other system could, this Mount Everest of games AI. Not only did it win the match, but it was how it won and with these creative new ideas like move 37. That for me was the signal that we were ready to turn it to scientific problems like AlphaFold.

Cleo Abram

复述一下：为什么要让想理解未来的观众懂第 37 手，是因为如果 DeepMind 能造出那种系统，也许也能造能下任何游戏的系统，也许也能在量子计算、核聚变、矩阵乘法、芯片设计里找出最好的解。这些系统里，第 37 手那种令人惊讶的创造正在哪发生？

To say this back to you, the reason why it’s important that this audience understand what happened with move 37 is because the implication is if DeepMind can build a system that can do that, it can also perhaps build a system that can play any game. It can also perhaps build systems that can figure out in real world problems what is the best solution in quantum computing or in nuclear fusion or in matrix multiplication or chip design. Could you tell me about the cutting edge here? What is the move 37 of the surprising creative element going on?

Demis Hassabis

AlphaZero 很值得讲，它是 AlphaGo 的演化。我们到了围棋巅峰、证明它能在围棋里想出新点子之后，进一步泛化成 AlphaZero。AlphaGo 从网上人类对局起步，还内置了棋盘对称之类围棋特有的东西。我们想去掉所有这些假设，从零开始，程序对它要做的事一无所知。Zero 指的就是：去掉数据和启发式里所有人类知识。

AlphaZero is very interesting to talk about, which was the evolution of AlphaGo. After we got to the pinnacle of Go and showed that it could come up with new ideas, at least in Go, move 37 and many other ideas, we then generalized it further to a system called AlphaZero. With AlphaGo, we started with all the human games we could find on the internet. There were a few other things that were specific about Go built into the AlphaGo system, like the symmetry of the board. We wanted to get rid of all of those assumptions completely and actually start from scratch as if the program and the algorithm didn’t know anything about what it was trying to do. That’s what the zero refers to in AlphaZero: AlphaGo, but now removing any knowledge, human-crafted knowledge, both in the data and in any of the heuristics.

我们测的是：它能不能从零学围棋，然后打败 AlphaGo。做到了。大约 17 代：AlphaZero 一开始随机，只有规则，自己下 10 万局，看到哪些着法赢了输了，哪怕近乎随机也会有稍好一点的着法。用那些数据训出版本二，比版本一稍好；再训出版本三、四。每一代和新旧对打，看是否显著更好。围棋、象棋这类，大约 16、17 代就够从随机到超过世界冠军。

What we tested AlphaZero on was first of all, could it learn Go from scratch and then beat AlphaGo. We managed to do that. It takes 17 evolutions of the program. AlphaZero starts off random. It only has the rules of the game, plays randomly. It creates its own data set by playing 100,000 games against itself, and then it can see which moves won or lost. Even though it’s playing more or less randomly to begin with, there’ll be some moves that are slightly better. We train a new version of itself, version two, with that new data. Version two is slightly better than version one. Then version two gets trained into version three, version four. Each time that new system gets played against the old system and sees is it significantly better or not. In Go and chess and things like that, around 16, 17 generations of that is enough to go from random to better than world champion.

象棋我甚至现场看过：早上随机，午饭时我还能勉强跟它下，下午茶时超过所有特级大师，晚饭时超过世界冠军。你看着整段演化从零发生。它下的新象棋，连 Stockfish 那种专家系统蛮力也没发现。AlphaZero 是 AlphaGo 思想的完全泛化。

At least in the case of chess, which I actually once watched live happen because I was fascinated — I was even playing chess myself — it starts in the morning random, then by lunchtime I could still just about compete with it myself, and then by tea time it’s better than all grandmasters, and then by dinner time it’s better than the world champion. You’ve just seen the entire evolution of that from scratch. Also it’s playing interesting new chess that even chess computers like Stockfish, the more expert-system brute-force ones, haven’t discovered. AlphaZero was the full generalization of the AlphaGo ideas.

有意思的是，这些想法现在要回到 Gemini 这类基础模型上。它们是对一切的广义模型，不只是围棋。我们仍需要在模型之上搜索、思考、推理，有时叫世界模型。怎么做还没完全破解。把 AlphaGo 思想带回来，但用到整个世界，也许还有材料设计、芯片设计、量子计算机。梦想就是我能沉浸在所有科学分支里，因为 AI 是通用工具。

Interestingly, I think we need these types of ideas back here now with our foundation models, Gemini and these kinds of things, which you can think of as generalized models of everything, language, the world around us, not just a game like Go. We still need this ability to search and think and reason on top of those models. Sometimes we call those world models. That still hasn’t fully been cracked yet, how to do that. Bringing back some of these AlphaGo ideas, but now instead of just a narrow game, applying it to the whole world, and maybe parts of science too, like material design and chip design and quantum computers. This is sort of the dream. I love every branch of science, and I get to indulge myself in all these different areas because AI is such a general tool.

一个例子是设计新材料：你要某种特殊性质，我们能不能超出材料科学已知。AlphaGo 式过程会很有用，第 37 手的等价物会像 AlphaTensor 找到让矩阵乘法更好、更快的新算法。矩阵乘法是所有神经网络的基础，快 5% 就是百亿级训练成本的巨大节省。芯片上的布线是 NP-hard，像旅行商，AlphaChip 这类程序在有些情况下比人类芯片设计师更好。我们才刚刮到表面。今天更通用的系统，叠上 AlphaGo 和 AlphaZero 的想法，未来几年会非常可观。

One example is just designing new materials. If you want a material with a special type of property, can we go beyond what is currently known in material science? AlphaGo-like processes could be very useful there. The equivalent of a move 37 would be like AlphaTensor finding a new algorithm that makes matrix multiplication better, faster. You can apply it in algorithmic space, which is quite exciting because then the algorithm itself gets faster, so there’s some circular improvement. Matrix multiplication is the basis of all neural networks. If you just make that 5% faster, that’s a huge cost saving on the tens of billions being spent on training. The design of chips on a die, making the routing as efficient as possible, it’s a kind of NP-hard problem, like the traveling salesman. AlphaChip programs like that are really good, better in some cases than human chip designers. I think we’re just scratching the surface of what’s going to be possible in the next few years with today’s more general systems combined with these types of ideas from AlphaGo and AlphaZero.

Cleo Abram

AlphaFold 开头的故事、AlphaGo 开头的故事，是让我真正乐观的那类 AI。真正乐观——你公开也常这么做，我很感激——是把可能出错的路径想透，以及我们能做什么来防止。所以我要插进另一块。这是实时战争游戏。视频里这套系统把人类打得落花流水，工程师在为胜利欢呼。但我没造这系统，我会想：如果那是真的呢？我们正处在军队和政府用 AI 的大辩论里。我希望这场对话十年后还有用，所以不想谈具体公司或服务条款。更大的图景是：政府会用 AI。作为造这些系统的人，如果有一根魔杖，你希望他们拿它做什么？

These two categories, the story that starts with AlphaFold, the story that starts with AlphaGo, these are the kinds of AI that make me feel really optimistic. I also think that being really optimistic, and you do this a lot in public, which I appreciate, is fully thinking through the ways in which something can go wrong and what we can do to prevent that. I want to insert one other in here. This is a real-time war game. In the videos where this system is absolutely crushing humans, you can see the engineers cheering for the victory of their system. But of course, as someone who didn’t build the system, I’m thinking to myself, what if that’s real? We’re speaking right now during a time when the debate about militaries and governments using AI is a huge topic of conversation. I want this conversation to last for 10 years, so I don’t want to talk about specific companies, specific terms of service. Bigger picture, governments are going to use AI. If you could wave your magic wand, what would you hope that they use it for?

Demis Hassabis

政府应该用 AI。我们想支持所有民选政府。我希望看到、也在把系统设计成擅长的，是改善公共卫生、教育。这些都需要重想。效率增益和能做的好事对公民会极可观。新加坡、阿联酋在往这些用例靠。我还希望用于能源，比如优化电网。我们在自己的数据中心做过，冷却系统省了 30% 能源。把 AI 大规模用在这些领域，社会收益巨大。我们想支持这些。

I think governments should be using AI, and we want to support all sort of democratically elected governments. The things I would love to see them use it for, and what we’re trying to build our systems to be good for, is things like improving public health, education. All of these things need to be rethought. The efficiency gains and the amount of good governments could do with it for their citizens could be incredible. Some countries are doing it, like Singapore and UAE, I think are leaning into these types of use cases. I would love to see it being used for things like energy, like optimizing energy grids. We did that with our data centers and save 30% of the energy used for the cooling systems. I think there’s enormous societal gain from applying AI at scale to these types of areas. That’s what I’ve always thought about and hope that governments will pick up and use, and we want to support all of that.

Demis Hassabis

当然，现在的地缘政治很复杂，这些是两用技术。大图景里我担心两件事。一是坏人——从个人到民族国家——把我们想用来治病、推进材料科学和能源的技术，挪用到有害目的，无论无意还是有意。二是 AI 本身变强之后出轨、脱轨。不是今天的系统，也许是未来两、三、四年，尤其当我们进入正在进入的智能体时代。智能体是能独自完成整项任务的系统。我们当然想要，当助理会非常有用，但那也意味着它们越来越有能力、越来越自主。

Of course the geopolitics of the world is very complicated right now, and these are dual-purpose technologies. I worry about a couple of use-case things that can go wrong with AI. In the bigger picture, there’s two things to worry about. One is bad actors, whether that’s individuals or all the way up to nation states, using, repurposing these technologies that we’re trying to build for good, like curing diseases and advancing material science and energy, for harmful ends, whether that’s inadvertently or intentionally. Then the second branch of things I worry about is the AI itself going rogue, or going off the rails, as they get more powerful. That’s not today’s systems, but maybe in the next two, three, four years, especially as we go towards more of the agentic era, which we’re entering now. By agents I mean systems that are capable of completing entire tasks on their own. Of course we want those because they’ll be very useful, like as an assistant. But also that means they’ll be increasingly capable and autonomous.

作为前沿实验室，我们都必须想护栏：如何确保它们精确做被要求做的事、目标规定得足够清楚，没有办法绕开或意外突破护栏。如果你想到这些系统最终会有多强、多聪明、多有能力，这是极难的技术挑战。这些现在几乎已经是中期问题——三四年其实也不算中期。我觉得人们此刻对它们关注得还不够，会是我们要安全走过 AGI 时刻必须面对的最大议题。

How do we make sure as one of the frontier labs — and the frontier labs all have to think about this — that the guardrails are put in place, that we can ensure that they do exactly what they’ve been told to do or the goals they’ve been given, and they’ve been specified clearly enough, and there’s no way of them circumventing that or accidentally breaching those guardrails. That’s an incredibly hard technical challenge if you think about how powerful and how smart and capable these systems eventually are going to get. I tend to worry about those. You could call them medium term now, even though three, four years is not really medium term. Those are the things I think people are perhaps not paying enough attention to at the moment, and I think will be the biggest issues that we’re going to have to contend with if we’re going to get through the AGI moment in a way that’s beneficial for humanity.

Cleo Abram

我进来最想问你的问题之一：如果这辈子能跟你聊一小时，下次我读到标题，该怎么权衡未来三十年我们都会有的那些担心？哪些事人们担心过多，哪些不够？

One of the biggest questions I came in for you with: if I get an hour with you in my life, next time I read a headline, how do I weight the concerns that we’re all going to have over the next 30 years? What are the things that people are worrying too much about, and what are the things that they are not worrying enough about?

Demis Hassabis

我刚说的两件，也许是普通人担心得不够的，甚至连一些专家和科学家也是。那些更影响社会。还有近期该担心的，比如深度伪造、虚假信息。我们做了 SynthID，一种 AI 水印，所有 Google 技术，Veo、Nano Banana，都带这种水印，可以检测并标记给用户或政府：这些是假的。我会主张所有做生成式 AI 的公司都内置这类技术，至少能检测哪些东西是用他们的技术做的。这件事会越来越重要。但跟 AGI 本身变得非常有能力、如何确保护栏、我们是否理解那些系统比起来，它仍然相形见绌。需要更多研究、更多努力，我也希望看到国际合作，前沿实验室之间、AI 安全研究所、学术界一起想怎么走下一步，因为造这种技术是没有先例的。

The two things I just mentioned are the things that maybe the average person is not worrying enough about, but even I think some of the experts and the scientists in the field. I feel like those are the key things that are more societal affecting. There are other things that we need to worry about too, like deepfakes. Those are immediate-term worries: misinformation, deepfakes. We work on this system called SynthID, which is a watermarking system, an AI watermark. All the Google technologies, Veo and everything else and Nano Banana, they all have this watermarking technology, so we can detect and flag to the user or government or whoever that these are fake. I would advocate all companies working on generative AI should build in some kind of technology like that, so at least it can be detected which things have been built with their technologies. That’s going to be increasingly important. But I think that still pales as a small issue compared to some of these bigger issues around AGI itself becoming very capable and how do we make sure that guardrails are put in place, that we understand what those types of systems are capable of as we get towards AGI. A lot more research, a lot more effort needs to go into that from everyone. I would love to see international cooperation amongst the leading labs around the safety issues, including places like the AI safety institutes and also academia, to help work out how we navigate that next step, because it’s unprecedented to create technology like that.

Cleo Abram

如果把这演下去，极限在哪？哪些事你觉得 AI 做不到、人能做？你把这称作你一生的中心问题。

If we play this out, what’s the limit here? What are the things that you think AI cannot do that humans can do? You’ve called this the central question of your life.

Demis Hassabis

它跟我一直以来的英雄之一艾伦·图灵的科学思考很相关。他描述了图灵机，理论上能计算任何可计算的东西，现代计算机基本上都是图灵机。我们在建的系统是近似图灵机；包括我在内的很多神经科学家觉得，大脑的一个好模型也是近似图灵机。问题是：也有朋友像罗杰·彭罗斯，相信脑子里可能有量子效应。我们有过很好脾气的辩论。到目前为止神经科学没在脑子里找到量子效应，不代表永远找不到，但人们找得相当仔细，还没找到。看起来脑子里大部分是经典计算。因此，AI 系统最终能做、能模仿的极限并不清楚。这是经验问题。意识也定义得不好，但我们都有直觉。造一个智能人工物的旅程，几乎会像对人类心智的对照实验，我们会看到差别、心智独特在哪。我对那一点心态很开放：可能有独特的东西，人与人之间那些独特连接也永远不会被这些 AI 复制。但很多我们现在够不着的，像长期规划、推理、某些形式的创造，我觉得 AI 最终能做。

It is and it’s very related to some scientific thinking of some of my all-time heroes like Alan Turing. He described Turing machines, theoretical constructs that actually all modern computers are basically Turing machines that are able to compute anything that’s computable. The systems we’re building are approximate Turing machines, and potentially a lot of neuroscientists including me think that maybe the brain — a good model for the brain is an approximate Turing machine. The question is, there are others like friends of mine like Roger Penrose who believes there might be some quantum effect in the brain. We’ve had some very good-natured debates about this. But so far neuroscience hasn’t found any quantum effects in the brain. Doesn’t mean they won’t be found, but so far people have looked quite carefully and we haven’t found any. So it looks like most of what’s going on in the brain is kind of classical computation. Therefore it’s not clear what the limit would be in terms of eventually what an AI system could do and could mimic. I think that’s an empirical question. The questions around consciousness — I don’t think it’s very well defined what it is, but we all intuit what it is. This journey we’re on of building an intelligent artifact will have almost like a controlled study comparison to the human mind, and then we’ll see what are the differences and what’s unique about the mind. I’m very open-minded about that. I think there could be unique things and certainly unique connections between humans that will never be replicated by these AI systems. But I think a lot of things that we currently are not in reach of, like long-term planning and reasoning and maybe some forms of creativity, I think eventually AI systems will be able to do.

Cleo Abram

我想诚实说出我脑子里正在发生的事：我正在做人类历史上一直在做的事，在找我们为什么特别的理由。我们必须在宇宙中心——哦我们不是。我们必须是情感上最敏锐的——哦大象也有葬礼。我们必须能创造艺术——哦 Gemini 也能。你也会这样吗？这就是你描述 AI 未来时我的反应。

I want to be honest about what’s happening in my mind right now and it is that I am doing exactly the thing that humans have done throughout history. I am trying to find the reason why we are special. We have to be at the center of the universe. Oh wait, we’re not. We have to be the ones that are emotionally attuned. Oh wait, elephants have funerals. We must be the ones that can be creative and create art. Oh wait, Gemini can do that. Do you find yourself doing that as well? That’s my reaction as you’re describing the future of AI.

Demis Hassabis

我觉得我们是特别的，宇宙怎么运转有很多深奥谜团，包括脑子里的，也包括物理学里的。我很小就决定做 AI，因为上学时就迷大问题。物理是我最爱的科目，因为那是你该学的、用来对付大问题的学科。但十几岁读科学书和传记时——费曼是我一直以来的英雄之一——我意识到：我们虽然发现了很多、知道很多，不知道的仍多得不可思议。我们不知道时间是什么。这对我来说太疯狂了。我们游在里面，它到底是什么？当然有熵之类，但对它真正是什么没有任何令人满意的描述。我们也不真正理解很多量子效应和引力，还有意识——其实是我们最在乎的大多数东西。我觉得多数人整天用电视和游戏分心，不太为此发愁。我从来不是那样，这些深谜一直压在心上。我对答案会是什么心态很开放。我最终想要的是现实的本质，我想用 AI 当工具帮我们理解周围现实的本质。无论答案是什么我都坦然。在这个意义上我是真正的科学家：对答案该是什么没有预设，我只想知道答案。

I think we are special and I think there is a lot of deep mysteries about how the universe works, including a lot of things that are in our minds, but also things out there in physics. I decided from a very young age to do AI because I was obsessed when I was a kid at school with the big questions. Physics was my favorite subject at school because that is the subject you’re supposed to study when you’re interested in all the big questions. But I just realized as a young teenager reading all these science books and biographies on the best scientists — Richard Feynman is one of my all-time heroes — that although we discovered a lot and we know a lot about the world, there’s so much we don’t know. We don’t know what time is. This is insane to me. We’re swimming in it, but what is it? Of course it’s entropy and things like that, but it’s nothing satisfactory about what it really is. We don’t understand a lot of quantum effects and gravity properly, and consciousness — actually most of the things we care about. I feel like most people we just distract ourselves all day with TV shows and games and don’t worry too much about it. But I’ve never been like that. These deep mysteries kind of play on my mind all the time. I’m quite open-minded about what the answers might be eventually about what’s going on here, the nature of reality. I think that’s ultimately what I’m after and I want to use AI as a tool to help us understand the nature of reality around it. And I’m quite sanguine about whatever the answer might be. I guess I’m a true scientist in that sense. I don’t really have any prescribed notion of what the answer should be. I just want to know the answer.

Cleo Abram

描述你在做的事的一种方式，就是造一个不是特别擅长这一件或那一件、而是像你一直说的 AGI、什么都擅长的系统。我知道你爱科幻。能不能把你脑子里那部科幻电影的情节演给我听：你真的做成了之后的未来？

One way to describe what you’re trying to do is to create a system that wouldn’t be especially good at one thing or another thing, but rather to create, as you’ve been saying, AGI, artificial general intelligence that would be good at it all. I know you’re a fan of sci-fi. Could you play out for me the plot of the sci-fi movie in your head that is the future where you actually do this?

Demis Hassabis

我也爱科幻，小时候读得可能太多，能解释一些事。我最喜欢的系列之一是 Iain Banks 的文化系列。它画了一个很有意思的后 AGI 世界——他没叫 AGI，但那就是我们在描述的，设定在一千年后。但我觉得五十年内有些就能发生：我们安全越过 AGI 时刻，它建好了，对社会有帮助，也许还在口袋里。我们用它破解了我所说的科学里的根节点问题。AlphaFold 就是一个：如果你把知识之树的根节点破解了，会解锁一整枝新研究或新应用。还有聚变，或者常温常压超导体，再配上最优电池。能源问题会有解：以某种方式几乎免费的可再生清洁能源，聚变或更好的太阳能。那会解锁我们去旅行星际，因为 SpaceX 那些了不起的工作里，主要成本仍是火箭燃料、能源成本。如果因为我们破解了聚变、能用海水造无限火箭燃料，成本几乎为零，那就真正解锁太空，可以挖小行星。科幻里的那些事，我觉得五十年内会非常可信。太阳周围的戴森球，水星位置和材料都刚好合适。然后最大程度的人类繁荣，治好这些可怕疾病，活得更长更健康，把意识带给银河其余部分。那会是惊人的结局，我觉得五十年内可以发生。

I love sci-fi too and I probably read too much of it when I was a kid. One of my favorite series was the Culture series by Iain Banks. I think it just paints a really interesting actually post-AGI world. He didn’t call it AGI, but that’s what we’re describing, like a thousand years in the future. But I think even 50 years some of this could happen, where we’ve got through the AGI moment safely. It’s built. It’s helpful for society and maybe we will have it in our pockets even. We’ve used it to crack some of these what I call root node problems in science. AlphaFold was one of those. These are problems if you think of the tree of all knowledge, these are kind of root node problems which if you cracked it, it would unlock a whole branch of new research or new applications. I think there are other things like fusion, or better maybe room temperature superconductors at atmospheric pressure that you could then combine with optimal batteries. There will be a solution to the energy problem. So pretty much free renewable clean energy one way or another, fusion or better solar. And then that will unlock us to really travel the stars, because the main cost of Elon does amazing work with SpaceX, but the main cost is still the rocket fuel, the energy cost. If that’s sort of zero somehow because we can just make infinite rocket fuel out of seawater because we cracked fusion, then that really unlocks space. And then we’ll be able to get a lot more resources because we can mine asteroids. All of these things that are the purview of science fiction become I think very plausible in the next 50 years. Dyson spheres around the sun. Mercury’s sort of conveniently in the right place actually made of the right material. And then that should hopefully lead to maximum human flourishing and we help cure all these terrible diseases. So we live much longer healthier lives and traveling to the stars, bringing consciousness to the rest of the galaxy. That would be I think an amazing outcome and I think could happen within the next 50 years.

Cleo Abram

如果我是自己葬礼墙上的一只苍蝇，他们说完她爱丈夫、家人和朋友之后，我希望他们会说：她把一生用来帮人看见乐观的未来，好让他们能参与把它实现，更快、对更多人更好，或无论人们决定拿看见的愿景做什么。所以最后一个问题：你希望他们怎么说你？

If I were a fly on the wall at my own funeral, after they said she loved her husband and her family and her friends, I would hope that they would say that she spent her life trying to help people see optimistic futures so that they can be part of making them happen. That they can make them happen more quickly or better for more people or whatever it is that people decide to do with the vision that they see. So my last question for you is what do you hope that they say about you?

Demis Hassabis

我希望他们会说，我的一生对人类有益、是服务。我想那就是我在做的。也许那会是最好的话。

I would hope that they would say that my life was of benefit and service to humanity. That’s I think what I’m trying to do. So that’s maybe would be the best thing.

Cleo Abram

非常感谢你的时间。

Thank you so much for your time.

Demis Hassabis

谢谢，非常感谢。

Thank you. Really appreciate it.

Demis Hassabis

我觉得我们覆盖了很多。GenCast，这是天气预报。对，我们没讲这个。Navier-Stokes，我完全忘了解决那一整枝。有意思的一块是模拟。我们没怎么谈模拟，也没谈 Genie。模拟帮你理解某个科学领域，甚至社会科学比如经济学：实验要么很贵，要么没法做对照。我一直爱模拟。DQN 当然是起点，Atari 那些。

I think we covered a lot actually. GenCast, this is weather prediction. Oh yeah, we didn’t cover that. Navier-Stokes, I completely forgot about solving that whole branch of things. One interesting thing is simulations. We didn’t talk much about that or Genie, which is the role of simulations to help you understand some area of science that or even social science like economics that you can’t — very hard to run either expensive to run experiments or you can’t run controlled experiments in. I’ve always loved simulation. DQN of course started it all off, the Atari stuff.

Cleo Abram

如果有人看到这里，对你描述的未来很乐观，也有你同样的担心，想参与。你会怎么建议他们参与？

One of the questions I think people will have: if they’re watching this and they are very optimistic about the futures that you’ve described, they have all of your same concerns, they’ve gotten to the end of this conversation and they’re thinking I believe in this future and I want to be part of it. How would you advise them to participate?

Demis Hassabis

我在大学和中学演讲时会说：你得顺着方向走。我会让自己沉浸在每一种能用的工具里，几乎变成带着那些工具和能力的超能力者。即便在前沿实验室，我们也有太多工作要去做下一代前沿模型以及相邻模型——对我们来说是 Veo、Nano Banana、Gemini——连我们都只能探索这些应用的一小部分。能力的悬垂越来越大，发布节奏越来越快。机会空间对真正擅长用这些工具、再应用到新领域的人会巨大。我觉得现在的孩子也许能用这些工具、用没人想过的新方式，创办数十亿美元的公司。OpenClaw 就是一个好例子。

When I do talks at universities and schools, I would say you’ve got to just go with the flow of the direction. I would immerse myself in every tool available and just become almost like superpowered with those tools and those capabilities. My impression is even at the frontier labs, so much work has to go into just making the next versions of these frontier models and then all the adjacent models. For us like Veo and Nano Banana and Gemini, even we can only explore a fraction of the applied things you could do with it, the applications you could make with it. That gap’s getting bigger and bigger in terms of the overhang of the capabilities, all the cool stuff on the latest models, and the release schedules are getting faster and faster. I think the opportunity space is getting huge for people who are really expert at using those tools and then apply it to some new domain. I think a kid these days could probably start a multi-billion dollar business using these tools in some new way that no one had thought about. I think things like OpenClaw is a good example of that.