# (77) Demis Hassabis On What AI Will Do Next - YouTube
- Transcript ID: `huawz_br8wm`
- Source tier: `youtube_auto`
- Original language: `en_or_source_language_unlabeled`
- Translation status: `agent_translated_reviewed` (49/49 chunks)
- Original source: https://www.youtube.com/watch?v=huAwz_BR8WM
- Raw evidence: `data/transcripts/youtube/demis-hassabis-two-minute-papers-2026-huAwz_BR8WM.json`
> 中文是派生翻译，英文是留存的源文本。中文未复核时，以英文和原始音视频为准。
## 00:00:00
**中文**
我们来几个简单问题。请尽量用一句话回答，有时甚至一个词。费曼还是牛顿？哦，这更难了。德米斯，我不知道有没有告诉你：我母亲最近拿到一些健康扫描结果。那是个巨大的视频文件，我们等了几周才得到评估，非常焦虑。
**English / Original**
We can easy questions. [music] Please try to answer in one sentence. Sometimes one word. >> Oh, wow. That's hard. Feynman >> or Newton? >> Oh, wow. That's even harder. >> [laughter] >> Dennis, I don't know if I told you about my mother recently got some health scan results. It was a massive video file and we had to wait for weeks for the evaluation and we were really anxious.
## 00:00:22
**中文**
然后我想，等等，Gemini 有长上下文，为什么不让它看看？它分析了扫描结果，然后说别担心，没事。后来医生验证，它确实判断对了。我真的很感激。我正要说谢谢，但与此同时你们还发布了 Gemma 4。
**English / Original**
And then I thought, wait. Gemini, long context, you know, why not give it a look? And it analyzed the scan and then said, don't worry. It's it's fine. And and later the doctor verified that that it's it it nailed it. And I am really grateful for that. And I was about to say thank you. Uh, but in the meantime you also released Gemma 4.
## 00:00:45
**中文**
这是一种免费的本地 AI，经过一点压缩可能也能做同样的事。我只想说，这些是送给人类的礼物，谢谢你们为普通人做这些。太好了。我很高兴听说你母亲没事。我们听到很多人因健康原因使用 Gemini 的类似故事，甚至有些情况是挽救生命的建议；这是一个不可思议的应用场景，很高兴它帮到了你。
**English / Original**
Free local AI that can likely do the same with a little compression. Just wanted to say these are a gift to humanity and thank you so much for doing this for the little man. >> Fantastic. Well, I'm I'm glad to hear that. It's it's it's I'm glad to hear your mom's fine, but it's we've had a lot of anecdotes like that of people using Gemini's for health reasons and actually in some cases life-saving so advice. So I think it's an incredible use case. So I'm glad it helped.
## 00:01:11
**中文**
英伟达的黄仁勋说，他把大语言模型当作决策的知己。你呢？研究以外会用 Gemini 吗？会。我还没有把它当作知己，不过将来也许会。我经常用它头脑风暴：项目想法、项目名称、创意。我很喜欢把它作为这种用途的切磋伙伴。
**English / Original**
>> Yeah. Yeah. Thank you very much. Now, Jensen Huang of Nvidia says that he uses LLMs as a confidant for decision making. What about you? Do you use Gemini for things beyond research? >> Yes. Um, I don't use it quite as a confidant yet although maybe at some point I will. I use it a lot for brainstorming. So, you know, project ideas, project names, uh, you know, think creative ideas. I quite like it as a kind of sparring partner for that.
## 00:01:40
**中文**
这大概是主要用途。我也用它浏览、总结某个我不太擅长但想快速了解核心要点的新研究领域。你会让它批评你的想法吗？比如启动 DeepThink、让数学奥林匹克模型批评我的想法？
**English / Original**
Uh, and that's probably the main use. And then I also use uh, it for for enter, you know, look at summarize some new area of research, some new body of research I'm not so expert in but I want to get a quick take on, you know, the main key points. >> Mhm. >> Do you use it to get your ideas criticized? Like like fire up deep think, you know, mathematical olympiad there and I criticize my idea.
## 00:02:03
**中文**
会，我主要用它帮助我梳理自己已经在想的一些步骤。你可以称之为批评，但我更试着以协作的框架使用它；也许我该要求它更严厉些，找出这个想法的缺陷。
**English / Original**
>> I have a Yes, I I mostly use it for helping me kind of think through some of the steps I've been thinking of as well. I mean, I guess you could call it a critiquing it, but it's I try it in a more collaborative frame rather than or maybe I should try it with more like be harsher, come up with the the the the flaws in this.
## 00:02:22
**中文**
但我肯定把它当作切磋伙伴。太棒了。听说你获得诺贝尔奖时，我想：终于对《主题公园》AI 有了真正的认可。没错，等了很久。后来我听约翰·江珀说了一句让我印象深刻的话：他期待看到有人用你的 AlphaFold 技术发明某种能获奖的东西。
**English / Original**
Um but I definitely use it as a kind of sparring partner. >> Amazing. Now, when I heard you got the Nobel Prize, I thought, "Finally, some proper recognition for the theme park AI." >> Yes, exactly. [laughter] Exactly. Took a while. >> Yes. And then I heard John Jumper say something that really stuck with me. He said that he's looking forward to seeing someone use your AlphaFold technique to invent something to win the prize.
## 00:02:48
**中文**
我称它为“二阶诺贝尔奖”。你预期会发生吗？这是非常有趣的想法。我认为有可能。现在已有超过 300 万研究人员在使用 AlphaFold，他们都在开展极其重要、有影响力的工作。所以约翰可能是对的，某个时候这也许会发生；那将是惊人的时刻。
**English / Original**
I thought let's call it the second-order Nobel. So, do you expect that to happen? Second-order Nobel. >> That's a really interesting idea. I I I am I think it's it's possible. Given the number of researchers who are using AlphaFold, you know, over 3 million at this point, and they're all doing incredibly important work and impactful work. So, yeah, I guess that that that John's right, that may happen at some point. That would be an amazing moment.
## 00:03:13
**中文**
你们有一个叫 Co-Scientist 的系统，可以发明新东西。能介绍吗？可以把它看成经过微调的 Gemini，上面专门加了额外工具和框架，用于生成假设、分析数据、总结文献。它是某种开端，几乎像有一位出色的研究助理帮助你完成日常工作。
**English / Original**
>> Amazing. Now, you have a system called Core Scientist that can invent new things. Can you tell me about it? >> Yeah, you can think of Core Scientist as a sort of fine-tuned version of Gemini that's specifically with extra tools and extra harnesses on top that is specific for helping with hypothesis generation, helping you analyze data, helping you summarize literature as well. So, it's the beginnings of a kind of almost like having a great research assistant that's helping you in your daily work.
## 00:03:47
**中文**
顺便说，新假设生成器很不可思议，我也试了，它给了我想法。它很有意思，因为它会让你收窄想法。你只说点什么，等大约 8 小时，结果绝对惊人。
**English / Original**
>> It's incredible, by the way, the new hypothesis generator. I also tried it. It gave me ideas. >> Yes. >> And then it was >> Interesting. >> Interesting because it it it wants you to narrow down the idea. Because you just you just say something. >> Yes. >> And then just the result was absolutely amazing. Like wait wait the 8 hours and then I was like this this is incredible.
## 00:04:06
**中文**
我在自己的领域试过。什么领域？光线追踪、全局光照。训练数据很少，做这个的人也很少，所以看到它在这方面表现良好很酷。它给出了一些有趣的科学想法并能帮助你。我希望自己有更多时间。
**English / Original**
And I tried it on on on my >> What area? >> area. >> Yeah. >> Ray tracing. Global illumination. Very little training data. Very few people do that. So that's I I think it's pretty cool to see it perform on that. >> On that. Okay. And it came back with some interesting scientific ideas. Helps you. Okay. I wish >> I wish I had more time.
## 00:04:23
**中文**
我也是。问题恰恰是现在要找到时间做这些事。我们几乎需要优秀的 AI 助理来处理行政工作，好让我们有更多时间使用 Co-Scientist；这也是我的梦想。2025 年 4 月 20 日，你说：“我想也许有一天我们能在 AI 帮助下治愈所有疾病。”
**English / Original**
>> [laughter] >> Me too. Me too. This is the problem is actually finding the time now to to do that. We almost need I feel like we almost need really good AI assistants to help deal with our admin work we have to do all of that stuff so we have more time for using co-scientists, right? This is my dream, too. >> On April 20th, 2025, you said, "I think one day maybe we can cure all disease with the help of AI.
## 00:04:48
**中文**
“也许在未来十年内，我看不出为什么不行。”看到这话我很兴奋，为此做了一个网站：curealldisease.com。我在里面放了一些数据，写着“常回来看看，我会用新数据更新它”，但从未更新。我们正在努力。不会是渐进的过程，我更像按 AlphaFold 的方式来想。我们在 Isomorphic Labs、我们的衍生公司以及 DeepMind 科学组都在建设。
**English / Original**
Maybe within the next decade. I don't see why not." When I saw that, I got so excited. I made a website for that. >> Oh, cool. >> curealldisease.com >> Okay. >> And I have some data in there and I wrote, "Check back often because I will update it with new data." >> Fantastic. >> And never updated. >> Yeah. We're [laughter] working hard. Uh no, yeah, but it won't be it won't be uh a gradual thing. It'll be more like AlphaFold is the way I'm thinking about it. We're building at Isomorphic Labs and and also our spin out and also uh at DeepMind as well in our science group.
## 00:05:25
**中文**
越来越多的工具。你可以把它看作在构建一个平台。AlphaFold 及其先进版本只是组件之一；蛋白质结构很重要，却只是药物发现流程中的一步。我们正在为药物发现流程的不同部分构建大约另外半打到一打 AlphaFold 级别的模型。
**English / Original**
Um more and more tools. You can think of it as building a platform. You know, AlphaFold's one of the components of that the the you know, um advanced versions of AlphaFold. But as you know, protein structure is only one component. It's an important component, but it's only one step in the drug discovery process. So we're building you can think another half dozen to a dozen AlphaFold level models that are on different parts of the of the drug discovery process.
## 00:05:52
**中文**
然后必须将它们整合，并在一些疾病图谱上测试，这正是我们目前在临床前阶段所做的。证明这一点还要几年；随后我们可能拥有可应用于几乎任何疾病领域的平台或引擎。这是希望。像 AlphaFold 2 一样，准确度足够后，突然就能在一年内折叠全部 2 亿种蛋白质。因此可能几年内你的网站没有更新，随后希望出现重大突破，让你不得不每天更新。
**English / Original**
And then we've got to put that all together and of course test it on some some disease profiles, which is what we're doing now in preclinical stage. And then once we've proven that out, which I think will take a few more years, then we may have a an engine that can be applied to, you know, a platform that can be applied to almost any disease area. That's the hope. So a bit like with AlphaFold 2, you know, you get it accurate enough and then suddenly you can fold all 200 million proteins in 1 year. So it's going to be more like that. So you may not have any updates on your website for a few years and then suddenly, I hope there'll be some big breakthroughs where then you'll have,
## 00:06:29
**中文**
我想起沃森的演讲：这个过程会呈指数式发展。第五年人们说，你才到 8%，在做什么？他说：“不，这太棒了。如果每年翻倍，我们就遥遥领先。”
**English / Original**
you know, you won't be able to you'll have to update it every day. >> Okay, [clears throat] I think I've heard something similar from from Watson's talk. He said that the process is going to be exponential. >> Mhm. >> And then at year year five, they said that well, you're only at I don't know 8%. What are you doing? And he said, "No, this is fantastic. I mean, if it's if it's doubling every year, we are way ahead."
## 00:06:50
**中文**
他可能指的是基因组计划或人类基因组计划。正是如此。我们已在 AlphaFold 上见过一次，所以需要复制这种成功，尽管它复杂得多。我也指平台可能提出那些潜在疗法；你仍需要在临床中测试并走完流程，这仍可能需要更多时间。
**English / Original**
>> Yes, he may he may he may be referring to the genome project probably or or the human genome project. That's exactly right. And and we've already seen this once with AlphaFold. So we need to rep, you know, replicate that success. Obviously, it's much much more complex. And I also meant by that as well that we would potentially have a platform that could come up with those potential cures. You still would need to test them in the clinic and and and go through that process. That could still take more time.
## 00:07:17
**中文**
能加快吗？我认为 AI 也能帮助：分层患者、更好预测剂量等。药物发现有两部分：药物发现过程和临床试验。我们目前聚焦第一部分，但 AI 也能帮助第二部分。届时人类健康将发生跃迁。
**English / Original**
>> Can you speed that up? >> I think actually AI could also help with that, too. Like stratify patients, maybe predict dosages better, all of those things. So I actually think AI could probably speed that up, too. So there's two parts, you know, to drug discovery. You've got to do the drug discovery process and then you've got to do the clinical trials. And right now we're focusing on the first part, but I think AI could also help with the second part. That's when we would then have a step change in in human health.
## 00:07:41
**中文**
所以我们正按轨道前进：9 年治愈所有疾病？我不会具体说那个时间，但未来 10 到 20 年内，我看不到物理定律阻止这一点。对，想想熵。读药物发现论文时我很兴奋，但专家说：没错，那只是一步，还有一千步。现在 AI 能让我们在药物研发中做到以前做不到的哪些事？
**English / Original**
>> So with that, I think we are on track, right? So, cure all diseases in 9 years. >> You know, I wouldn't say specifically roughly that kind of time frame, but I think in the next 10 to 20 years, there's I don't see any laws of physics that prevent that, right? >> Yeah. Think entropy. >> Yeah. Exactly. >> And when I read drug discovery papers, it's it's not my area, but I get really excited. And then I ask an expert, and they say, "Yes, uh exactly what you said. That's just one step, and there's a thousand other steps." So, I'm wondering, what can we now do with AI that we couldn't before in drug development?
## 00:08:17
**中文**
我们在做的包括 AlphaFold 的先进版本。如今可预测许多相互作用：不仅是蛋白质静态图像，还有蛋白质之间、蛋白质与分子之间的作用。接下来的难点是预测它在体内会做什么：结合到哪里、ADME 性质、吸收和毒性——这些最终都会成为副作用。我们必须建立预测所有这些不同因素的模型。
**English / Original**
>> Well, I think what the kind of things we're working on is um so, there's obviously there's advanced versions of AlphaFold, you can imagine. Um we we can predict lots of interactions now, so not just static picture of a protein, proteins interacting with other proteins, proteins interacting with molecules. Um and then the next the hard parts are then predicting what is it going to do in the body. So, when you know, where does it bind to? Are there, you know, what's the ADME properties, absorption, and and toxicity, all of these things which end up being side effects. So, we've got to build models that predict all of these different things.
## 00:08:51
**中文**
然后把这些整合，并用生物化学模型确定要设计哪种化合物、如何制造，以及它将如何、在哪里结合蛋白质的特定口袋或目标。我的印象是，药物发现中现在做不到的许多事，将能借助 AlphaFold 及其后续 AI 做到。
**English / Original**
Um and then put that all kind of all together, uh as well as like kind of biochemistry models of like what exact compound would you design, and how is it going to be made, and how and where how will it bind to the particular pocket or whatever it is that you're targeting in the in the protein. >> Mhm. >> My impression is that there's a lot of things in drug discovery that we cannot do at the moment, and we will be able to do with AI like AlphaFold and and and and whatever follows.
## 00:09:18
**中文**
有哪些事现在做不到、你觉得也许不会变？我不知道，例如某些逃避物理规律的监管问题？监管显然有些超出 AI 可直接帮助的范围，但 FDA 等监管是流程中极其重要的一部分。它们可因更多证据而加快：一旦几种 AI 设计药物完整通过监管流程，就可能有足够数据；比如 10 种 AI 设计药物中有 9 种——
**English / Original**
What are the things that we cannot do at the moment, and you you see that that's maybe not changing? I don't know, maybe something about regulations that, you know, that that evades physics >> Yes. >> Regulations, obviously, that's a bit outside the scope of AI to help with, but what I hope it with there, you know, if you think about regulations, FDA, these types of things, uh obviously incredible pop and important part of the process, but they could be sped up, like we were saying earlier, with um, more evidence that once a few AI-designed drugs, I think, get through the whole uh, regulatory process, then you might have enough data, say there was 10 AI-designed drugs and nine of them
## 00:09:55
**中文**
成功了。那就可以回测模型预测到底多准确、哪些可信、哪些不可信。这会成为向监管者证明可加快某些流程、甚至跳过或用更高效东西替代某些步骤的证据。但首先需要证据证明，几种 AI 辅助或 AI 设计药物能以大致传统的方式走完全程，真正帮助患者；然后我们才可以考虑改变做法。
**English / Original**
worked, right? Then you would be able to back test and how accurate really were your models in their predictions? Which ones can you trust? Which ones can't you? And then that would be evidence to the regulator, I think, that that some of those process could be sped up, uh, and maybe some of the steps could be could be skipped or or or replaced with something more efficient. But I think it's going to require evidence, first of all, that a few of these AI-assisted or AI-designed drugs get all the way through to actually help the patient in the probably traditional way, and then we can change, you know, maybe think about changing the way that that's done.
## 00:10:33
**中文**
也许类似 mRNA 疫苗？有可能。那时有新技术，也有 COVID 带来的紧急需求，因而加快了测试和试验。这是可能路径之一。但这关乎未来十个世纪的人类健康，不必在未来 5 到 10 年匆忙推进；只是我看到的未来技术非常令人兴奋，我觉得任何事都可能。
**English / Original**
>> Maybe similarly as what we had with the mRNA vaccines, right? >> Yes, potentially. Right. So, you know, there was a new technique there, there was this emergency need, obviously, with COVID, and then that accelerated the tests and the trials on those things. You know, that's that's one potential way it could go, but I don't think there's I mean, this is this is something about human health for the next, you know, next 10 centuries, right? We we don't need to rush it in the next 5 to 10 years, but I just think what I see is such exciting technology coming down the road. Um, I just think anything may be possible.
## 00:11:07
**中文**
作为前玩家，能谈谈与《EVE Online》的新合作吗？当然。你以前玩过？太好了。这是很棒的游戏，社区非常特别；很少有社区通过建立联盟和派系，某种意义上也参与构建游戏。
**English / Original**
>> That's incredible. Now, as a former player, can you tell me a bit about your new partnership with the EVE Online game? >> Oh, yes. Oh, you were former player? >> Oh, yeah. >> Okay, fantastic. Yeah. Well, it's an awesome game, and uh, I think its community is very special, actually. What, you know, very rare that community sort of builds the game, too, really, with the way they build their alliances and factions.
## 00:11:31
**中文**
它极其深邃，是大型多人游戏，整个宇宙不仅由游戏设计者、也由玩家创造。它有运作中的经济和很有意思的动态剧情。我们长期欣赏他们的作品；我很了解公司 CEO Hilmar，我们从我还做游戏时就是老朋友。它是测试有趣 AI 想法的绝佳沙盒，可帮助游戏玩法，并测试 AI 如何与经济互动。
**English / Original**
Um, it's incredibly deep game. You know, massive multiplayer. It's a whole universe really that's created not just by the game designers but also by the players. And it has a functioning economy and and really interesting kind of dynamic storylines. So we think that we've admired their work for a long time. I know Hilmar, the CEO of the company very well. We're good friends going way back to when I was still doing games and we think it's a fantastic it's sort of sandbox if you like for testing out some interesting AI ideas that will help the gameplay and you know, be interesting kind of to test out how it interacts with economies and
## 00:12:12
**中文**
以及与剧情、玩家联盟等互动。因此它将是有趣的试验场，对游戏和社区都很迷人；他们一直处于前沿，这延续了传统。对我们而言，这也延续 DeepMind 使用游戏作为安全试验场的传统。
**English / Original**
and and the storylines and and player alliances and things like that. So it's going to be an interesting proving [clears throat] ground. I think fascinating for the game and the and the community and and also like they're always been at the cutting edge really. So this is continuing that tradition. And then for us it's also continues in the tradition of DeepMind of using games as a as a as a safe kind of proving ground.
## 00:12:34
**中文**
这种概念是否可让某些 AI 智能体嵌入其中与玩家一起玩？有可能，或协助玩家。我们还处于早期探索阶段，也想同社区头脑风暴。也可以想象它像游戏主持人一样推动剧情。
**English / Original**
>> Would the concept work in a way that that some AI agents would be embedded play with the players? >> potentially or maybe assisting the players. I think we're we're still early days of exploring that. We want to brainstorm that with the community too. But you can imagine also maybe it's almost like a a games master that drives the storyline along as well.
## 00:12:56
**中文**
太酷了。我迫不及待想看到 AI 机器人在吉他（Jita）被封禁。你得重新开始玩。对了，如果需要有人把你的钱翻倍，我认识一个人。其他都是骗局，这个保证成功。
**English / Original**
>> That is so cool. Look, I cannot wait to see when AI bot gets gets banned in Jita. >> You have to start playing again. You have to start playing again. >> [laughter] >> Yeah, and and by the way, if you need someone to double your money. >> Yeah. >> I know a guy. Okay. >> Look, all the others those are scams. This one guaranteed.
## 00:13:14
**中文**
我们会确保 AI 智能体去找你做这个。现在，Co-Scientist 已经发明了哪些东西？我的理解是它不直接写论文，科学家在写论文。没错。如今它更像助理，而非自主发现东西；也许那是下一步。目前它在协助科学家和数学家开展工作。
**English / Original**
>> we'll make sure our AI agent comes to you for that. [laughter] >> Now, can you tell me a bit about the things that the cool scientist has invented already? My understanding is that it's not directly writing papers. Scientists are writing up the papers. Yes. >> That's right. So, it's it's more like an assistant today rather than autonomously discovering things. You know, maybe that's the next step. Um but right now it's assisting uh scientists in in you know, in mathematicians in their work.
## 00:13:41
**中文**
希望我们很快会公布一些事。早期版本的这类系统已经能寻找更高效的矩阵乘法，也用于 AlphaEvolve 和 AlphaTools，基本上是在改进计算机科学算法。某种意义上，让发明作用于自身，使自身更高效，这很酷。
**English / Original**
Um I hope we'll be announcing some things pretty soon, but earlier versions of these these types of systems were doing things like finding more efficient matrix multiplication or like we used it for uh an alpha volve and alpha tools as well for basically in in improving like computer science algorithms. So, kind of turning uh invention on itself to make itself more efficient, which is pretty cool.
## 00:14:05
**中文**
我们显然刚把它推出，还只是触及其潜力的表面。现在，这会不会是 AI 的终极图灵测试？例如爱因斯坦测试：把知识截断在 1901 年左右，看它能否提出狭义相对论。那爱因斯坦测试呢？
**English / Original**
And and I think that we obviously we're just putting it out there now and we're just scratching the surface of what I hope will be [clears throat] we'll see the potential of it's going to be. >> Now, wouldn't this be the ultimate Turing test of of AI? Like you have the Einstein test where you where you stop it at 1901 or somewhere in come up with special relativity. Now, what what about the Einstein test?
## 00:14:24
**中文**
平方，是的——假设大于一。那么，今天的门槛是发明一个新的、值得诺贝尔奖的东西。那当然是我们想要的。为了证明这一点，我提出的爱因斯坦测试是：若把系统带回 1901 年，限制其知识，它能否做出爱因斯坦 1905 年的“奇迹年”成果，提出四篇在不同领域都有突破的惊人论文，包括狭义相对论？如果能，那么就可以把一个基于现代物理训练的模型拿来，要求它给出——
**English / Original**
>> squared Yeah. >> Assuming assuming bigger than one. Uh so, so the cutoff is today like invent something new that's like a Nobel Prize worthy thing. >> That's obviously what we would want. And but in order to prove that that's the case, we would this this Einstein test that I've I've been saying about, you know, if you took it back to 1901 with the knowledge cutoff, could it have invented a you know, Einstein's annus mirabilis of you know, 1905 where he came up with you know, four amazing papers all breakthroughs in their different areas, right? Including special relativity. So, if it was able to do that, then presumably you could turn that model obviously trained on all modern-day physics and then ask it for
## 00:15:03
**中文**
比如，比弦理论更好的东西。然后你或许应当认真对待它返回的结果。这种回测可表明那类系统和技术已准备好真正发明新的科学。
**English / Original**
you know, something better than string theory, right? And then maybe you should take quite seriously what it came back what it came what it came back with, right? So, that's that So, that would be the kind of back test to show that that kind of system, those types of techniques would be ready to actually invent new science.
## 00:15:19
**中文**
我真想把这个问题放进去让你听。在光线追踪研究中，我们大量做蒙特卡洛积分，向约翰·冯·诺伊曼祈祷。本质上，称为采样器的模块会产生一张有噪声的图像。
**English / Original**
>> I'm I'm I'm going to try to put put in this one because I I I really want you to to hear about this. Now, I'm I'm thinking that in in ray tracing research, we do a lot of Monte Carlo integration and we we pray to John von Neumann. >> Yes. >> And essentially, you have uh you get a noisy image from a module that we call a sampler.
## 00:15:37
**中文**
然后交给去噪器，于是得到更干净的图像。那样能行，但最佳系统不止于此：去噪器和采样器之间有连接。去噪器会告诉采样器：“我希望这张图在高频区域有更多样本。”
**English / Original**
>> Mhm. >> And then you pass it over to a denoiser. >> Mhm. >> So, you get you get a a cleaner image. Now, that kind of works, but but the best systems do more than that. They have a connection between the denoiser and the sampler. So, the denoiser actually tells the sampler that, "Hey, I want this image with more samples in the high-frequency regions."
## 00:15:57
**中文**
这样去噪器能做得更好。它们不再是两个独立单元，而是融合成一个单元。我在想，假设生成器和验证器之间是否也有类似机制？那会像闭环自动化发现，对吧？
**English / Original**
>> Mhm. >> So, that the denoiser can do a better job. >> Mhm. >> So, it's not two separate units anymore. It's it's fused into one unit. >> Yeah. >> Now, what I'm thinking is that do you have something like this between the hypothesis generator >> Mhm. >> and the verifier? >> Mhm. >> Well, I that that would be kind of like closed-loop automated discovery, right?
## 00:16:14
**中文**
如果能融合它，大多数前沿实验室都在思考递归自我改进。我认为它可在编程和数学等领域奏效，因为验证器能非常容易、快速地验证答案是否正确或有进展，也可借此生成合成数据。但在物理、化学、生物等自然科学中更难，因为验证器可能还需要自动化实验室或原子世界的东西，这会让循环长得多。我们在思考——
**English / Original**
If one could fuse that. Um most of the the the frontier labs are working on trying to think about recursive self-improvement. I think it could work in some domains like coding and math because the verifier that you can verify very easily and very quickly whether the answer is correct or making progress. Uh and you can also generate synthetic data that way as well. So, um I think it's a bit harder if it's a physical science or kind of natural science like physics, chemistry, or biology where the verifier will probably also require an automated lab or something in the world of atoms. And that will obviously make the loop a lot longer. Um we are thinking about things
## 00:16:55
**中文**
例如在 Isomorphic 建自动化实验室，我认为某个时候会做。我还在等，看看哪些数据是 CRO 等拿不到、我们真正需要的，也在等机器人技术再先进一点，DeepMind 正在做且进展不错。或许 18 到 24 个月后，我能想象建立某种这样的实验室。我们正在为材料科学做一个，在伦敦建设，因为我们有 20 万种新材料设计；其中可能有惊人的材料，但我们没有足够快的测试方法，里面有超导体和各种有趣东西。
**English / Original**
like automated labs uh in isomorphic. Um and I think at some point we will do that. I'm I'm waiting a little bit to see what what type of data do we really need that we can't get from a CRO or something like that. And I'm also waiting for robotics to get a little bit more advanced, too, which we're doing at DeepMind. So, that's going well. So, maybe in 18 months, 24 months, I could imagine setting up some kind of lab like that. We're actually doing one for material science. We're putting a automated lab together. We're We're building one out in London because we have uh material designs. Actually, we're sitting on 200,000 designs of new materials, which we don't know but there
## 00:17:32
**中文**
也许有很棒的材料，但我们不能足够快地测试。对大多数科学领域，验证步骤将需要物理验证。问题是发现循环的瓶颈在哪里：是假设生成，还是验证步骤？
**English / Original**
could be an amazing material in there but we haven't got a way of testing that fast enough. There's some superconductors in there, all sorts of really interesting things. >> that work? >> Yeah, so you need So, I think for those types of domains, most of the science domains, the verify step is going to require some physical verification. And then the question is is like what's the bottleneck in the in the discovery loop, right? Is Is it the hypothesis generation or is it the validation step?
## 00:17:58
**中文**
所以做那种递归自我改进未必容易。另一个问题是，即使在数学和编程里，所有前沿实验室也都在思考这个：必须考虑这种过程的安全性，因为如果没有人类在环中——
**English / Original**
So, it may not be so easy to do that kind of recursive self-improvement. The other issue is even in maths and coding, uh all of the frontier labs are kind of thinking about this is you've got to think about the safety of that type of uh process cuz it's just sort of with, you know, if if no human is in the loop. >> Mhm.
## 00:18:15
**中文**
最后来一轮快速问答。简单问题，请尽量一句话、偶尔一个词回答。离散数学还是连续数学？提示：只有一个正确答案。真的？好吧，我偏好离散，但不知道这是否是正确答案。
**English / Original**
>> Now, at the end, this is going to be a quick lightning round. >> Okay. >> We can easy questions. Please try to answer in one sentence, sometimes one word. >> Okay. >> Okay. Discrete or continuous mathematics? Hint, there's only one good answer. >> Really? [laughter] Okay. Well, I prefer discrete but I don't know, you know, is that is that the right answer?
## 00:18:35
**中文**
对你来说。对我来说，是的，我认为这一切都是离散的。费曼还是爱因斯坦？很难。我个人大概选费曼。费曼还是牛顿？这更难。也许我得选牛顿，为了我与剑桥的联系。
**English / Original**
>> um >> [laughter] >> for you. >> For me? For me, yes, exactly. I think all of this is discrete. >> Okay. Feynman or Einstein? >> Oh, wow, that's hard. Probably Feynman for me. Put my personal favorite. >> Okay. Feynman or Newton? >> Oh, wow, that's even harder. >> [laughter] >> Uh there, maybe I'll have to go with Newton. Maybe I'll have to go with Newton for my Cambridge connections.
## 00:18:58
**中文**
最喜欢的《Two-Minute Papers》一期？哇。我喜欢你们团队的很多内容，经常用它来快速跟进，但我特别喜欢你们的 AlphaFold 视频，全部都很优秀，是对 AlphaFold 如何运作最好的解释之一。太感谢了。最好的阿西莫夫小说？
**English / Original**
>> All right. Favorite Two-Minute Papers episode? >> Oh, wow. Okay. Oh, I love loads of your team. I use it a lot to to get up to speed with things but I I tell you what, I loved your AlphaFold uh videos. I I thought they were excellent, all of them. Some of the best explanations out there of of how AlphaFold works. >> That is very kind. Thank you. Best Asimov novel?
## 00:19:19
**中文**
《基地》系列。它对我影响很大。你读过机器人系列、《钢穴》吗？有趣的是没有。我只读过《基地》系列，但我大概也该读机器人系列。现在，德米斯·哈萨比斯爵士，爵士头衔是一回事，但为了你今天对人类的贡献，我将授予你 Two Minute Papers 荣誉徽章。
**English / Original**
>> Uh Foundation series. Yes. That's what was really influential for me. >> Have you read the Robot series? Caves of Steel? >> No, I haven't, interestingly. Yeah, I I just read the Foundation series, so that's but I I should probably read the Robot ones, too. >> And now, Sir Demis Hassabis, a knighthood is one thing, but for your contributions to humanity today, I shall bestow upon you the Two Minute Papers Badge of Honor.
## 00:19:45
**中文**
哇。你是首位获得它的科学家。太棒了。谢谢。给你。谢谢，我会自豪地佩戴它。我也在看你们所有视频。太好了。谢谢你，德米斯；我今天学到了很多。
**English / Original**
>> Oh, wow. >> You're the first scientist to ever receive it. >> Amazing. >> So, with that said, >> Thank you very much. >> There you go. >> Thank you very much. I will wear it with pride. >> [laughter] >> Thank you and I'm watching all your videos. Awesome. >> Thank you, Demis. Thank you so much. I learned a lot today.
## 00:20:01
**中文**
太有趣了。非常感谢，也谢谢这个徽章，太棒了。现在切回我在家录的幕后片段，字面上，因为你看不清我们。现在好些了。这是巨大荣誉、梦想成真，我不敢相信自己能在那里。但后来发生什么？我把这个枕头放在桌上，告诉德米斯我拿不到正规的那个。
**English / Original**
>> So much fun. Yeah, thank you so much. And thanks for the badge. Brilliant. >> Now, this is me from back home with some behind the scenes footage. Literally, because you can't see us properly. Okay, better now. So, this was a huge honor and a dream come true. I can't believe that I got to be there. But, what happened afterwards? Well, I put this pillow on the table and told Demis that I couldn't get a proper one.
## 00:20:28
**中文**
我问：“想看另一面吗？”这就是它的样子。我给他看时，现场爆发大笑，可惜没拍到。不过露西·柯利当时在房间里，拍了照片，去看看吧。非常感谢。如果一切顺利，下一次采访是杰夫·迪恩；如果喜欢并想看，请订阅并点铃铛。这里可见我通过 Lambda GPU Cloud 运行完整 DeepMind AI 模型，6710 亿参数，速度极快——
**English / Original**
So, I asked, "Want to see the other side?" And this is what it looks like. And when I showed him, there was huge laughter, which unfortunately we couldn't get on camera. However, Lucy Curley was in the room and shot a photo of it. So, check it out. Thank you so much. If everything goes well, next interview is Jeff Dean. So, subscribe and hit the bell if you enjoyed this and would like to see that, too. Here you see me running the full DeepMind AI model through Lambda GPU Cloud, 671 billion parameters running super fast
## 00:21:07
**中文**
而且极其可靠。这太疯狂了，我很喜欢并经常使用。Lambda 提供强大的英伟达 GPU，用于运行自己的聊天机器人和实验。请现在访问 lambda.ai/papers 或点击说明中的链接试试。
**English / Original**
and super reliably. This is insane. I love it and I use it on a regular basis. Lambda provides you with powerful Nvidia GPUs to run your own chatbots and experiments. Seriously, try it out now at lambda.ai/papers or click the link in the description.