芯片基础原理与物理微缩极限
在接下来的大约30分钟里，我想介绍一下芯片是如何从原材料制造出来的，以及超越硅之后、后硅时代的硅芯片未来会是怎样。之后，我还会介绍我们实验室如何尝试利用AI来推动硅材料或半导体材料的发现与制造。首先再稍微做一下自我介绍：我是Harry Wong，目前是杜克大学电子与计算机工程系的助理教授。我在上海交通大学完成本科，在麻省理工学院获得博士学位，在加州理工学院做了几年博士后，大约三年前在杜克大学建立了我的实验室。今天我想从“硅谷101”开始讲起。好的，我想先讲一点类似于“硅基101”的基础知识，给大家普及一些关于芯片究竟是如何工作的基本常识。
英文原文
[music]So um in the next maybe 30 minutes I want to introduce like uh maybe how chips are built from raw materials and what's the future of the of the silicon chip after or beyond silicon and then after that I will introduce our so so our lab we try to uh use AI to boost the silicon or or semiconductor discovery and manufacturing. So again introduce myself a little bit. I'm Harry Wong. I'm currently assistant professor at Duke ECE department and I finished my uh undergrad in Shanghai Jaong University and then a PhD at MIT. I spent a few years at Caltech postoc and I started my lab in Duke about three years ago. So I want to start with today is valley 101. Okay. I want to start with something like like I will call it a silicon 101 and and basically give you some basic knowledge about how how the chips actually working. Okay.
首先我想说，我非常喜欢这组幻灯片，它实际上与上一场圆桌讨论相呼应。就像以前我会说，AI绝对需要芯片来运行、训练、推理等等；但现在在过去的两三年里，大家注意到硅或者芯片相关的行业至少在股市上迎来了爆发，对吧？所以在这种情况下，我们也可以看到一些初创公司正试图利用AI来促进芯片的设计等等。因此我认为这可以形成一个闭环：AI需要芯片，芯片也需要AI。但是究竟什么是芯片？芯片的基本元件又是什么？基本上，一切都是二进制的。即便在现代，所有的大语言模型和计算都变得非常复杂，但在最底层，一切依然归结为0和1。
英文原文
So firstly I want to say I really like this slides actually corresponds to the previous panel like like uh previously I would say like AI definitely needs chips to run it and train inference and everything but now in last two three years you notice like silicon uh or the chips uh related industry booms in the stocks at least right so in that case like I will say like we also see some startups are trying to use AI to to to to boost the design of chips and everything like that. So I will say this can become a loop like AI needs chips and chips needs AI but what what are the chips and what are the basic elements of chips. So basically everything is uh binary. So so even though like all the uh large language models and the computing become very complex uh in the modern uh era but still everything goes to zero and one at the very bottom.如果你看芯片是如何工作的，最基本的元件被称为晶体管，它更像是一个水龙头。好的，如果我们关掉栅极，电子就无法从源极流向漏极；如果我们把它打开，强行开启它，你可以看到电子就可以穿过源极流向漏极。这就是最基本的元件，通过从0切换到1、从1切换到0，从而实现了一切。现在回到芯片和AI，我们再次谈到了扩展定律（Scaling Law），但扩展定律在硅或芯片、在半导体领域其实早就存在了。我们确实相信，如果你在单个芯片中堆叠更多的晶体管，它就会更强大，具有更高的性能。这里展示了从单个晶体管积累到数十亿个晶体管，它就能成为一块芯片并具备强大的功能。如果看这个例子，英伟达的Blackwell 300 AI处理器在单个芯片上实际上集成了2000亿个开关或晶体管。所以在这种情况下，基本上如果你想拥有更好的芯片，就需要在单个器件中容纳更多的晶体管。
英文原文
So if you see like how does chips work the basic element is called transistor which is more like a like like a faucet. Okay. So if we turn off the gate so the electrons cannot flow from the source to the ring. So if we turn it on and force it on you can see here the electrons can flow through the source to the ring. And this is the very basic element and turn from zero to one and one to zero and realize everything. Now based on the chips and based on the AI all right so so again like we we talk about the scaling law but the scaling law is already there in uh in in silicon or in the chips okay so in semiconductor so we do believe like like if you stack more transistor in a single chip it's just be more powerful and has higher performance okay here shows like from a single transistor if we accumulate to billions of transistors It's can become a chip and become very functional. And showing you here if we see this example like if we we see the Nvidia Blackwell 300 AI processor is actually has 200 billion switches or transistors in a single chip. Okay. So, so in that case like basically if you have you want to have better chips you need to have more transistors in a single device.摩尔定律的演进与短沟道效应瓶颈
于是这就引出了一个非常自然的问题：我们如何在单个芯片中塞进这么多的晶体管？这基本上就是硅谷成功的起点，对吧，我们这里是硅谷，名字就是这么来的。基本上高登·摩尔总结出了摩尔定律，它的核心含义是，如果你的晶体管更小，每个元件尽可能小，你就能在一定的芯片面积内塞入更多的晶体管。这里展示了过去50年晶体管的发展历程：最初是相当大的晶体管，然后把它缩小，从微米尺度缩小到纳米尺度；然后从平面结构发展到FinFET，接着又发展到纳米片（nano sheet）结构，基本上就是把所有东西堆叠在一起。然而最近出现了一些问题。目前我们拥有FinFET并将其堆叠，但堆叠之后会出现其他问题，比如热问题，在这些界面处产生的热量引发了一些麻烦。
英文原文
So in that case that's come a very natural question like how we can fill in so many transistor in a single chip. So so basically this is basically the the the the beginning of the success of the Silicon Valley right we are a Silicon Valley and that's the name come from. So uh basically so so golden moore basically realized the moors law actually saying that if you have a smaller transistor each element as small as possible you can fill in more transistor in a single in certain footprint. Okay so in that case here shows a just a look back of how transistor uh develop in last 50 years. So originally from pretty large transistor and then shrink it down to from a micron scale to nano scale and then from uh kind of like plary to the the we have a finfat and then we have like nano sheet structure basically we stack everything together. So however recently there's some problems. So if you so so currently we have fin fat and then we stack it but after stacking it there'll be other problems such as the the thermal problem the heat generated at this interfaces uh generate some problems okay so in that case like um so所以我想稍微介绍一下晶体管的一种常见失效模式。当晶体管缩小到很小的尺度时，有时会出现一个问题，这就像你家里的水龙头一样。如果它非常厚，电子有时不仅在你关掉它时无法直接流过，还可能从底部流过。我们称之为漏电。正是这种漏电消耗了大量的电能，浪费了大量的能量，并产生了热量。在这种情况下，如果你使用的是非常厚的材料，电子就可能从其下方穿过。那样的话，你就必须把材料做得尽可能薄。如果它足够薄，我们在关断状态下就能关紧水龙头，没有电子能够流过；而在导通状态下，电子可以顺畅地流动。我想向大家介绍一下“硅基101”或“半导体101”中的一些常见概念：对每个元件来说尺寸更小更好，而且对每个元件或半导体材料来说厚度更薄更好。
英文原文
I want to introduce a little bit a common common failure mode of the transistor so transistor when is shrink it into small scale so sometimes there's a problem as it also happens in your in the faucet in your home Okay. So if if it's very thick, so basically the the electron sometimes not just flowing if you turn it off it's not flowing directly from here but also can flow from the underneath. So we call it a leakage. So this leakage actually that's why like consume a lot of power and waste a lot of power and generate the heat. Okay. So in that case like if you use a thick material like very thick like this sometimes like like like like like like the the electrons can go underneath it. In that case you have to make the material as thin as possible. So in that case if it's thin then we can close the faucet in the off state no electrons can flow here. In the on state the electrons can flow well. Okay. So in that case like it's actually I try to give you uh some common uh maybe concept in in silicon 101 or semiconductor 101. So smaller is better for each element and thinner is better for each element or semiconductor materials.我们已经看到了这种趋势，但我想让大家认识到，目前硅材料存在很大的问题。上一场圆桌讨论谈到了金融方面的问题或者发展势头，但我希望让大家明白，硅目前确实存在一些瓶颈。原因是，我们在物理上无法无限地微缩硅材料。如果硅材料从相当大的尺度缩小到5纳米以下，电子的迁移率和流动速度就会变慢，最终你会看到它无法正常流动；在这种情况下，如果硅的尺度极小，在导通状态下也不会有电子流动。作为科学家，我们认为有可能出现下一代半导体，我们称之为以石墨烯、二硫化钼为代表的二维半导体——这些材料获得了2010年的诺贝尔物理学奖。与硅相比，二维半导体即使在5纳米以下，也能够保持其电子迁移率或电子流动能力。
英文原文
So we already see this trend but I want to convince you that currently the silicon has big problem. Okay, I I really value like uh in the previous panel we we talked about the finance maybe some problem or it's still going right but I want to convince you like the silicon currently has some problem. The reason is if we we we basically cannot shrink silicon infinitely. So if silicon goes from uh uh uh from like pretty large scale to shrink it to below five nanometer. So electrons become the mobility and the speed of electron become slow or finally like like you can see it's not flowing well in that case in the on state so no electrons can flow if you basically have a really small scale of silicon. So in that case like uh as a scientist so we think it's possible to have a next generation semiconductor we call it two-dimensional semiconductor represented by graphine malleabon dulfide uh like which material won the Nobel physics pri uh Nobel prize in in two 2010 right so so basically so so but comparing with the two-dimensional semiconductor so two dimensional semiconductor can keep its electron mobility or the capability of flowing electrons even it's below five nanometers.二维半导体材料与台积电技术路线
大家可能会感到困惑：看新闻时，半导体行业总是宣传现在已经有2纳米节点，未来还会有1纳米节点，为什么不像我刚才说的5纳米就是极限呢？我想说明的是，这里展示的所有制程节点数字目前都只是标称节点（nominal nodes）。如果我们回顾半导体和晶体管的结构与历史，目前行业采用的是纳米片结构，也就是把硅层叠堆叠起来。在这种情况下，哪怕是7纳米，如果有3层，7除以3大约就是2纳米。类似的情况在英特尔早期就已经出现过：我们最初有奔腾1、奔腾2、奔腾3，而现在拥有12核或14核。因为缩小单个元件非常困难，所以工业界决定堆叠更多层。因此在根本上，硅仍然存在物理局限：你无法像其他二维材料那样把它做得极小。
英文原文
Okay. So, so you you might be feel confused. you you you look at the news actually the the semiconductor industry always tell you oh we now now have a two nanometer node in the future we have one nanom meter node why it's not like what I just said is 5 nanometer is a limit right so I want to to say like every like the node numbers showing here is is currently is all nominal nodes that basically means like if we check uh like like the the the the the structure and the history of semiconductor and transistors. Uh so currently we are using the nano sheet structure. Nano sheet structure you stack the silicons. So in that case even though you have seven nanometer you have three layers then you have seven over three that's around two nanometer. Okay. So similar thing we actually already seen in the Intel's original we have pantom pantom one panum 2 panum 3 and now we have like 12 cores or like 14 cores. So because it's it's hard to shrink a single element. So so basically industry decide to stack more. Okay. So in that case like um I want to convince you like like silicon still fundamentally has a problem. You can make it you cannot make it extremely small like like other two dimensional materials.正如我所提到的，这类材料只有一个分子层厚度，这里展示的是两层。包含一个钼原子和两个硫原子，整个材料的厚度在1纳米以下，大约是0.65纳米。作为半导体，学术界在过去十年中已经证明，基于这种材料构建的单器件性能远优于采用硅制造的单器件。那么这种材料何时能进入你的手机、电脑，真正成为电子产品呢？这里展示的是台积电（TSMC）的技术路线图。它基本上显示出，目前我们正处于从纳米片结构向CFET过渡的阶段；但在不久的将来，这将会走向二维过渡金属硫族化合物，也就是我刚才展示的二硫化钼（MoS2）。
英文原文
So but this materials as I mentioned it's only has one molecular layer. Here shows two layers. So you have one molecular uh uh uh mabulum uh uh uh atom and two sulfur atoms on and in total this material has only the thickness below one nanometer is about 65 65 nanometer. So so so but still as a semiconductor it has been showed in academia in last 10 years a single device built upon this material is much better than a single device of using the silicon to build it. Okay. So if if we see like uh what when this material will enter your phone into your computer become really electronics in product. So this is shows a road map by TSMC. It's basically shows like so so now we are at between the nano sheet structure moving to the CFAT as I just showed to you. But in the very near future this this will go to the two-dimensional uh transition metal that charges or I just showed you as MOS2. Okay.制造瓶颈：从化学气相沉积到良率管理
大家可能会想，如果这种材料真如我所说的那样非常薄，而且能作为晶体管的沟道材料，那么核心问题就是如何高效地制造它。虽然我们能获得微小的薄片并证明单个器件性能优异，但单个芯片始终需要成百上千亿个晶体管。为了制备这种材料，目前在大学实验室以及一些正在从事类似研发的工业界合作伙伴中，我们采用一种称为化学气相沉积（CVD）的方法：利用两种前驱体，前驱体随气体流动，在衬底表面生长出这种极薄的半导体。问题在于，这个单层虽然在晶体管中能表现出优异的沟道性能，但它不同于硅。硅是通过物理方法制备的，纯度极高，可以达到10个9或12个9；这意味着在12英寸晶圆上检查每个原子，可能只有一两个堆垛层错。然而这种新材料目前仍难以达到硅那样的完美程度。如果不像硅那样完美，现有的工业制造流程就必须改变。
英文原文
So in that case so so you may think like like if this material just the material like I I tell you it's very thin and can perform as a channel material of transistor. So basically uh so the problem is like how to build it efficiently. So even though we have some small flakes and we demonstrate single device is great but we always need a lot of like as mentioned like hundreds of billions of transistor in a single chip. So by building this material uh basically currently in the university lab and I I I I know there's some uh industry partners are working on the similar things. We use a method called chemical vapor deposition have some two precursors. the precursor can carry with a gas and grow this extremely thin semiconductor on top of a substrate. Okay. So the problem is like this one layer can perform really well as a channel in the in in the transistor but it's it's not like silicon. Silicon is we use a physical way to build it. So have really high purity like 10 eyes or 12 9. That means it's on a 12 inch wafer. If you check every item so only one or two stacking for on 12 inch wafer however this material still struggling about like get is perfect as a silicon. So if it's not as perfect as silicon basically the current industrial flow should be changed.此外我还想提一下，这种材料不仅可以用来做半导体。在我们实验室，我们还发现如果把这种材料打印在某种柔软的基底上，它能够对光产生响应。你可以看到它就像尺蠖一样只能朝一个方向蠕动。我常和朋友开玩笑说，这就像在读博士：当导师和学生谈话时，他们就稍微弯曲一下；如果我不找他们说话，他们就直接躺平，对吧？不过我们一直在前进，你不可能一直弯着身子。或许在场的管理者在笑，大家懂我的意思。那么大家会产生疑问：既然这种材料看起来如此神奇，为什么它还没有进入手机、没有进入日常生活？我认为问题可以用一个词来概括：良率（Yield）。
英文原文
So in that case like uh so so I want to also to mention that this material not just can be bu semiconductor. In my lab, we also developed something. We found if we print this material on top of some soft substrate and uh this material can respondse to to the light. You can see it's like inch warm can just move on one direction. I'm joking with my friends. This is doing PhD. When the advisor talking with student, they just bend a little bit. If I don't talk with them just laying down, right? So, [laughter] but but we are moving. We are always moving. So, on you cannot always bend, right? So uh so so so so maybe the manager is laughing so you know what's going on. So uh so so so then you are questioning like this material seems very magical but why is still not it's not in your phone not in the daily life. The problem is like uh I will say that just one word it's Yel period.台积电之所以强大，是因为他们拥有世界上最顶尖的良率管理能力；英特尔和三星在良率管理方面也非常出色。这意味着在半导体行业中，任何企业想要立足并赢得份额，良率管理都至关重要：你必须在单个芯片上制造数百亿个晶体管，但你如何确保每一个晶体管都能正常工作？台积电做得非常好，但对于这种新型二维半导体，良率管理需要全新的逻辑，依然存在需要解决的新问题。在整个半导体行业中，目前仍然依赖人工来进行良率管理，至少需要靠人去观察这些材料、寻找缺陷、判断质量好坏。
英文原文
So TSMC is great because you can manage Yel the best in the world. Okay. And the Intel and Samsung also very good on yield management. That means like like any semiconductor industry if you want to uh uh like like uh like get a share in the industry. So the yield management very important that means you have to build a chip about hundreds of billion transistor on one chip. But but but you how how do you make sure everyone is working everything is working or every transistor is working. So TSM is doing great but for this new 2D semiconductor Y management needs some new logic but still something like new unit to solve. Okay. So basically I will say like in in the whole go back a little bit like to in the whole semiconductor industry so we still rely on human to do the yel management and at least to see these materials see the defects see the uh see see it's good or see it's not good. Okay.利用AI实现显微镜自动化与纳米材料检测
于是在大约两年前，我们实验室开始研究能否利用AI来观察这些材料、检测材料中的缺陷。昨天有一场关于望远镜的精彩演讲，那是通过宏观尺度去观测恒星等事物；而在我们实验室，如果要在纳米尺度上进行工程制造，就必须尽可能细致地观察这些微小的事物。因此我们在去年发表了一篇论文，利用AI——具体来说是大语言模型——来控制显微镜，让显微镜去观测这些纳米材料，对其进行表征并理解这些材料。这项工作是两年前完成的，当时甚至“智能体（Agents）”的概念还处于早期阶段，所以我们称之为工作流。我们利用大语言模型来控制显微镜，对采集到的图像进行分割，这里采用了Meta的Segment Anything模型，然后结合有监督聚类这种非常基础的机器学习算法。别忘了机器学习是AI的开端。通过这种方式，我们可以训练AI去识别存在什么材料、表面是否干净、是否有缺陷，或者是否空无一物。
英文原文
So in that case then like about two years ago uh so my lab we started to to look at like whether we can use AI to to see these materials or to see the defects of these materials. So so so by by doing this yesterday there was a really good talk about telescope like to see the stars and everything that goes through macro scale of of seeing things. So in our lab if we want to engineer in nanocale we have to to see these small things as as as small as possible. So therefore like we published paper last year we call it atomic we use AI basically large language model to control a microscope and let the microscope to see these nanomaterials and characterize these mu uh nanomaterials and understand these materials. So it's it's it's the work is done two years ago. At that time even agents as a concept was still early. So we call it a workflow. And basically we can use the large language model to control the microscope and to segment the the the the the image captured and we use a meta segment anything model and then we can use some supervisor clustering which is a very basic machine learning. Don't forget machine learning is it's the beginning of AI. Okay. So, so, so, so, so, so we can teach the AI to understand what materials are there. Is that clean? Is there defects? Or is there something just nothing there? Okay.这非常有意思。翻页笔现在没反应了，也许我们该把它换成非硅基的设备。屏幕似乎卡住了。但总的来说，我们所做的就是让显微镜能够自主执行研究任务。这种探索更偏向于AI for Science范畴，也可以看作是如何利用AI去观测半导体制造过程。最重要的是，当时大家还在质疑大模型是否存在大量幻觉——尽管最近很多研究人员不再常提起这个词，但在当时幻觉确实是个问题——但我们验证了利用AI来识别材料的可行性，能够仅凭图像对材料进行最终分类，区分出单层、多层、裸硅或者空白区域。
英文原文
So, in that case like we we we can see it's very interesting. Uh, this is not responsive now. Maybe we should replace it to not silicon based thing. So, so I think we're frozen. But but but again like like basically what we did is like we can let the microscope to do the task research talk by itself. So by doing this basically uh it's it's it's possibly to to we can see this is more like in the category of AI for science or we can say like it's actually uh uh like like in in in more like a structure of like how to use AI to actually see the manufacturing of the semiconductors. Okay. So and and most importantly we notice uh it's possible to uh uh so so so at that time like people still questioning is ch or or still have a lot of hallucination okay this word recently not heard by by many of the research but at that time the hallucination is a problem but we we validated that it's possibly like to use AI to identify the material and also to finally classify the results by just seeing the material what are single layers, what are multi- layers, what are bare silicon or or what are the nothing over there. Okay.我们发现这个模型表现非常出色，而且目前依然运转良好。与肉眼相比，该模型的良率识别准确度可以达到99.7%。尽管延迟仍然是一个问题——人类看一眼就能得出结果，而系统仍需要6秒钟，不过最近我们正在降低延迟，效果很不错——但在某些任务上它甚至比人类更出色。例如左侧的图像中，如果让人类去标注所有可能的二硫化钼薄片，人肉眼标注的往往就像红框所示；但模型不仅能识别整个薄片，还能识别出所谓的畴界（domain boundaries）。畴界是不同微区之间的交界，本身就是一种缺陷。这里并没有使用特殊的技巧，而是因为人类依赖肉眼直觉，而模型进行的是逐像素分析。在某些特定任务中，AI的表现绝对优于人类。我们还测试了半导体领域的多种不同材料，比如诺奖材料石墨烯以及各种二维半导体，该模型表现都非常好。最重要的是，我们甚至不需要对其进行训练，也不需要数据中心，只需构建智能体系统并设计提示词，就能驾驭它来完成我们的应用。
英文原文
So, so basically we found it's this model works really well and it's working now. Yeah. So, so this this model works really well. So we can see like comparing with the human eyes this the yield of the model can reach to 99.7 even though the latency is still a problem like human just look at it in and then you can get result but we still need six seconds. So recently we are we're lowering the latency is working pretty well but again like like like uh in some task is even better than human. So if I show the left image showing here basically I'm telling you like asking you to to label all the possible like like like uh MS2 flakes is sometimes the human will just label like this in the red. Okay. But model cannot just see the the the whole flakes but also identify some we call it like a do domain boundaries. It's a boundary of different domains and these boundaries are actually a kind of defect. So, so there's not no tricks here but because a human we just use intuation by our eyes but but this uh like model they have pixel by pixel analysis. So in some certain task like I'm I try to convince you that AI can work definitely better than human. Okay. So we also try different kinds of like materials in semiconductor for example showing here like different kinds of for example graphine which is Nobel prize winning um material and all kinds of 2D semiconductors. This model works really well and most importantly we even don't need to train it. Can you believe it? We don't need data center. we just we we can just build a aentic system and and and uh and do the prompts and then we can we can manage to to harness it to to for our for our application. All right.AI驱动的端到端自主物理实验与配方迁移
尽管我们教会了AI如何观察这种材料，但距离实际制造、正确生产还有很长的路要走。过去两年AI for Science非常重要，而我认为很快就会转向“AI for Engineering”，工程AI将变得至关重要。对于这些新型半导体，我们认为可以利用AI来加速发现和制造过程。我们已经看到AI非常擅长提出假设和构思。谷歌去年发布的Co-Scientist利用AlphaEvolve等结构，让AI能够提出更多创意。正如昨天的AI for Science圆桌所提到的，现实中的物理实验和真实世界验证非常困难。我们在大约四周前刚发布的一篇论文中，实验室与Google DeepMind合作，尝试采用这种Co-Scientist架构并将其引入物理世界，在物理实验中开展端到端的验证与材料发现。
英文原文
So um so so so then like still even though we teach AI to see this material still like we are far from actually how to manufacture it how to get it right. So I will say like in last two years AI for science is very important and I will say like soon we'll change to I will I'm thinking like maybe the AI for engineering uh will be very important. So in that case like we think like for this novel semiconductors it's possible to use AI to speed up the discovery and speed up the manufacturing. Okay. So so actually we already see the AI is very good on hypothesizing ideas. So, so Google's co-scientists I actually uh published last year and uh and and they they realized they using some some some structure like like ARA alpha evolve so AI can give more ideas. Okay. So and also the problem is like as as mentioned yesterday in the A4 science panel actual physical experiment and also the actual real world validation is really difficult. Okay. So, so in that case like what we did this paper just released a lot uh four four weeks ago. So, so my lab collaborated with Google deep mind and we try to use this co-scientist structure and add it into the physical world in physical experiments to do the end to end validation and discoveries. Okay.这项工作让Co-Scientist不仅能查阅文献，还能提出构想并制定实验或执行方案；我们还将仪器和自建的实验装置连接起来，让AI直接控制它自主完成实验。大约一年前，人们还在说：“Harry，你以后不需要博士生了。”看完这项工作后，我觉得也许很快就不需要教授了，因为AI提出的构想比我的更好，而且它还能自己写论文。或许我们都应该去数据中心工作。这项工作可以阅读文献、设计实验、完成端到端验证；这里的视频展示了其运行方式：Co-Scientist架构能够预测前驱体，生成实际执行的操作方案，并控制软件中的操作面板。目前人类依然在控制回路中：我们还需要用手称量前驱体、装入反应管并进行密封。但也许很快就可以用走廊上的机械臂来取代这一步骤。系统生成了非常优秀的配方，我们在论文中确信，它有可能生长出以往从未成功制备的新型半导体；即便是对于二硫化钼这类传统二维半导体，也展现出了极佳的结果：只需让AI查阅文献并控制仪器，就能获得成果。
英文原文
So basically the uh the idea is I want to introduce the work a little bit. So the co-scientist now cannot just do the the the the literature review. It can give ideas and also can do the give the experimental or the execution plans and also we connect our uh our instrument and a self-built experimental setup and let AI to control it and now it can do the experiment by itself. Okay. So in that case like in in about one year ago people are thinking in the future uh Harry you don't need a PhD student. Okay. Then after I saw this work, I I'm saying like maybe you don't need a professor very [laughter] soon because uh because AI can give ideas better than I so why need a professor [laughter] and and also it can write paper by itself, right? So um so maybe we should all work on uh work on data center. [laughter] Yeah. So, so basically this work can do the the the read literature design experiment validate end to end and more importantly here this video shows like how it works is the co-scientist geometry can can can pre predict the precursor and also generate a protocol to actually execute it and also like controlling some panel in the software and human still in the loop right now. So, so, so we have to use hand to weigh in the precursor, put it into the to the tube and and and and also like seal it. Okay. But maybe soon like we can use a robot arm in that that hallway [laughter] to replace this step. Uh but but again like it's generate a really good recipe and can even like in the paper we we we highly uh truly believe like it's possible to grow some new semiconductor never grown before. Okay, even theoretically predicted but but it's very hard to grow and for the conventional 2D semiconductors as I showed you MS2 and everything. So it's also shows really good results. We just let AI to control the instrument. It looks at the the literature and get the results.在学术界，许多实验结果往往难以复现，因为不同实验室的设备几何结构存在差异：我的实验室有这种炉子和管道，但已发表的论文采用的却是另一套装置。我们发现AI在复现实验结果方面表现极为出色。如果我们把实验室特有的几何构型输入给AI，让它查阅文献并针对我们定制的仪器生成专属配方，效果非常理想。输入始终是提示词和任务描述，输出则直接对接定制仪器。这里展示了一个演示：我的实验室长期制备这种材料，但我一直催促学生尝试生长另外两种材料。学生告诉我需要查阅文献，得花上几周时间。有一天我说，不妨用大语言模型生成一个配方试试看能不能成功。结果一次就成功了。我跟学生开玩笑说，未来甚至不需要查阅文献，直接按照模型生成的方案操作即可。这为我节省了几个月的时间，也免去了催促学生的精神折磨。这表明AI仅凭一次尝试、单次试错即可达到工业级制造水准。当你需要复现实验且面对全新设备时，AI能够节约时间并加速研发与科学发现。
英文原文
And most importantly we found like um I will say like in academia many of the results are very hard to repeat because different labs have different geometry. So my lab have this furnace have that tube but but the published result has a different setup. So in that case like we found uh like AI can uh work really good on how to repeat the results. That means that if we have the geometry of our lab and give it to AI and let it look at the literature and try to see like can you have a customized recipe using my customer's instruments. It works really well. So it's it's actually showing here like the the the in is always prompt and the task description and the out is can be the directly get from the customers customized instruments. So basically here shows a demo like I try to push my student to grow my my lab is growing this material for for a long time but these two material I try to push my student to to can you grow it can you grow it and my my student say oh Dr. I have to look at literature. It takes a few weeks or something. Then one day I just say, can you just try use large angle model and try a recipe and let's see whether it's work. Then it's it's working. Okay. Then then it's I joke student like like like in the future like that you don't need to even like look at the literature. You can just follow what models tells you. Okay. So that saved me a few months. So I I try to push my my my students there's a lot of me mental mental torture for myself right so but uh but again here shows like it's the AI can do the uh try the the the industrial scale manufacturing just with one shot or we say there's just first trial okay this is also something if you want to repeat something and you have just a new setup and AI can save your time and speed up the the discovery and your research.AI原生制造：从缺陷筛选到“自生智能”闭环
如果我们再往前走一步：我们已经能够检测材料并合成材料，那么能否直接制造出真正的芯片？这张尚未发表的图片显示，我们可以将显微镜与制造环节结合。显微镜拍摄到的生长材料并不理想，这是我们特意选择的非理想材料，上面有很多黑点缺陷。如果把这种材料交给工业界合作伙伴，他们会认为这是废料，根本不可能在上面生长芯片。但如果让AI来进行制造，AI会专门挑选完好无损的区域。
英文原文
So uh basically here shows if we go one step beyond we already see the material and also we can grow the material. How about we directly build the actual chip. So here image showing like not published yet. So we can combine this microscope. You can see the microscope saying some materials we grown which is not ideal. We on purposely choose a non ideal material. You can see a lot of black dots are actually defects. If I give this material to the industry partner, they will say this is crap. Okay, you can never grow any chips on top of that. But if we use our AI and let AI to build it. So what AI will do is just choosing the good area.在传统制造方式中，我们需要数十亿个整齐一致的晶体管；但如果借助AI，可以看到每个晶体管的布局都可以因地制宜。目前我们正将AI与超净间内的所有半导体制造设备连接起来，实际制造芯片。尽管这颗芯片尚未达到数十亿晶体管的规模，目前只集成了数千个晶体管，但这代表了一个非常好的开端。让AI接管半导体制造能够显著提高良率：原本这种材料自身的良率只有65.6%，但在这种低良率材料的基础上，我们制造出了高良率的晶体管阵列（这是最基础的芯片），良率提升了约30%。我认为这将在未来颠覆并彻底改变整个半导体制造行业。
英文原文
So you can see conventional way we have uniform transistors billions. Okay. But if we use AI, we can see every transistors are different. And now we are connecting AI with all the uh clean room manufacturing instruments and we can build like it's actually uh uh like actually build the chip. So even though the chip is not billion scale okay so we can have still have thousands of transistors but we think it's a very good beginning. So basically like we also see like by taking the uh let AI control the whole semiconductor manufacturing it's increase the yield. So originally this this material can only have 65.6 uh 65.6% which is crap but built upon this low yield material we can have a high yield like transistor arrays which is a very basic uh chips. So this 30% around 30% increase we think it will wholly like like will change uh and disrupt the whole manufacturing semiconductor manufacturing in the future. That that's my belief. Okay.基于这一理念，我们大约一个月前创办了一家初创公司，探索利用AI变革半导体行业。我们认为AI原生的半导体制造将变得越来越重要，我们将尝试提供良率智能（yield intelligence）或从实验室到晶圆厂（lab-to-fab）的解决方案，甚至考虑为新型二维半导体建立一条小型中试线。今天大家都在讨论AI，但最终所有的资金都流向了英伟达，而英伟达最终要把钱付给台积电，对吧？制造这些芯片才是成本最高的环节。尽管大家都在谈论数据中心等方案，我现在给出了一种更便宜的替代路线。最近很多人都在讨论递归自我改进（Recursive Self-Improvement, RSI），而我们认为，补齐缺失的环节——即让AI制造属于它自己的大脑——至关重要。
英文原文
So, so in that case, so in that case, basically we recently built a startup about a month ago and we think like using AI to to to interrupt the the semiconductor industry will be very interesting. We think AI native uh semiconduct m manufacturing will be more and more important. We think we will try the yel intelligence or lab to fab solutions or we think of like like maybe we want to build a small pilot line for the new 2D semiconductors. But again like like uh for everyone talking about today so all your money goes to Nvidia in the end [laughter] and Nvidia have to pay to TSMC in the end right so in the end is how to build these chips actually takes the most right even though we talk about AI we talk data center and everything now I give you alternative solution which is cheaper okay so [laughter] so so so again like I will say like AI has already write some models uh recently a lot of people talk about RSI recursive self-improvement and then like like we think like building the missing link how to use AI to manufactur its own brain is very important.这构成了一个飞轮：AI能够设计芯片、制造芯片，芯片又能在核心数据的支撑下训练AI。这形成了一个更高效的飞轮，模型与芯片最终能够融为一体，类似于我们的生物大脑——人类的思维与大脑从未分离，但目前模型与硬件芯片在某种程度上是割裂的。如果我们将芯片与模型结合，用模型制造芯片，再用芯片运行模型，我认为这代表了一种全新的生命形式，或者可以称之为硅基生命形式的新形态。如果在传统AI和RSI之后迈向下一步，也许有人称之为超级智能（Superintelligence, SI）——最近SI似乎比AI更受推崇——但如果将物理制造真正融入芯片构建，我认为这是一种全新的智能。既然还没有人为此命名，为了让它回归AI的缩写，我把它命名为“自生智能”（Autogenic Intelligence），完成整个闭环。
英文原文
So this uh it shows like a flywheel like if AI can design the chip can build a chip can train the AI chips can train the AI with the data in the middle so it's actually become a more efficient like flywheel or I will say like like like the model and chips can finally combine to each other similar to our biological brain. So for example our minds and our brain are never separated right. So but currently the model and chips are kind of separated right so if we combine this chips and model by using model to build chips and using chip to run the model so I will say this is the new format of life >> maybe we can say it's like a new new format of the silicon life format whatever you call it okay and I will say like this could be the next step if we have a conventional AI and then we have RSI and then this RSI maybe we can call it a super intelligence and recently like it seems like SI is more favorable than AI but but but again if we realize put the manufacturing into the the building the chips and I will say this could be new intelligence and because no nobody has has give a name of that maybe I can try to give a name to bring it back to AI I'll call it autogenic intelligence is go back to AI again to finish this closed loop okay so将思维与大脑、芯片与模型结合在一起就是一种新的生命形态。作为教授，我总在思考这最终会走向何方。我想借用一个1916年（一百多年前）被发现的生物学概念——趋同演化中的“蟹化”（Carcinisation）。它的含义是，人们发现无论海洋生物最初是什么形态，经过数百万年的演化往往都会变成螃蟹的形态。比如鱼演化数百万年可能变成像帝王蟹的样子，乌龟演化数百万年可能变成类似青蟹的样子。这种现象在大自然中已经存在了数亿年、数百万年。因此我想向大家提出一个问题：你们认为AI与芯片的结合最终会发散还是收敛？这种硅基生命形态最终是否也会演化出某种“螃蟹”的形态？下次吃螃蟹时大家不妨想一想，也许你吃的是一种高度智慧的生命体。我的演讲到此结束，非常感谢大家。