推荐微信扫码登录以获得更好的体验!另:🎉建议更新到最新版,以体验新功能!🎉 版本更新记录→

黄仁勋最新断言:别再单打独斗了!未来企业必须靠“开放智能体联

字幕摘录

时间英文中文
0:11Excited to be here with Jensen.很高兴和Jensen在一起
0:13There has been a ton of advancements in AI and agents over the past year, but few months在过去的一年里,大赦国际和特工机构取得了大量进步,但几个月
0:18in particular, I feel.特别是,我觉得。
0:19We've seen a lot of these advancements come in the form of better performance, but at我们已经看到很多这些进步 表现得更好,但是...
0:24the same time, we've also seen that openness and control and trust in a lot of these models同时,我们也看到 开放、控制和信任 很多这些模型
0:30and systems around them has become more and more important.周围的系统变得越来越重要。
0:34So the first thing I want to start with is how and why are you guys at NVIDIA investing首先我想说的是 你们在NVIDIA投资的方式和原因
0:40in an open agent ecosystem and stack?在开放的代理生态系统和堆积?
0:44Well, first, before I answer, I want to congratulate you for all the work that you do.首先,在我回答之前,我要祝贺你所做的一切工作。
0:49In fact, if you look at the last six months, we could both agree that although we've been事实上,如果你看看过去的六个月, 我们可以都同意,虽然我们已经
展开字幕全文(452 条)
序号英文中文
1Excited to be here with Jensen.很高兴和Jensen在一起
2There has been a ton of advancements in AI and agents over the past year, but few months在过去的一年里,大赦国际和特工机构取得了大量进步,但几个月
3in particular, I feel.特别是,我觉得。
4We've seen a lot of these advancements come in the form of better performance, but at我们已经看到很多这些进步 表现得更好,但是...
5the same time, we've also seen that openness and control and trust in a lot of these models同时,我们也看到 开放、控制和信任 很多这些模型
6and systems around them has become more and more important.周围的系统变得越来越重要。
7So the first thing I want to start with is how and why are you guys at NVIDIA investing首先我想说的是 你们在NVIDIA投资的方式和原因
8in an open agent ecosystem and stack?在开放的代理生态系统和堆积?
9Well, first, before I answer, I want to congratulate you for all the work that you do.首先,在我回答之前,我要祝贺你所做的一切工作。
10In fact, if you look at the last six months, we could both agree that although we've been事实上,如果你看看过去的六个月, 我们可以都同意,虽然我们已经
11working in AI for 15 years, the last six months changed everything.在大赦国际工作了15年 过去6个月改变了一切
12And so all of the technology, all the large language model advances, all the scaling,所以所有的科技,所有的大型语言模型 进步,所有的缩放,
13all of the breakthroughs, all the Omni models, multi-modality stuff, all that stuff is fantastic.所有的突破,所有的Omni模型, 多式联运的东西, 所有的东西是惊人的。
14But in the end, it was the last six months that where everything kind of came together.但最后是过去6个月, 所有的东西都走到了一起。
15And now finally, AI is useful.现在终于,AI有用了.
16And when AI is useful, every company in the world, every enterprise in the world wants当AI有用时 世界上每一个公司 世界上每一个企业都想要
17to get their hands on it.以获得他们的手。
18And now the question is how?现在的问题是如何?
19This is where Lang chain comes in.兰连锁店就在这里
20And so you always had a vision that the large language model was the essential ingredient,所以,你一直有一个愿景, 大语言模型是基本成分,
21the essential enabling technology.关键辅助技术。
22But in order to turn it into a useful product, you have to surround it with what is now known但是为了把它变成有用的产品 你必须用现在已知的东西来包围它
23as a harness.做为绳索
24There's so much more.还有很多
25So much more.还有这么多
26And back in the old days, we used Lang chain to help us turn a large language model into回到过去,我们用Lang链 帮助我们把一个大语言模型变成
27a promptable API.一个快速的API。
28And we would turn, we use Lang chain to build our rags and we use Lang chain to step by然后我们转身,我们用Lang链 来建造我们的破布,我们用Lang链 步入
29step by step, which led to today's agents.一步一步,这导致了今天的特工。
30And really what happened in the last six months, the big breakthrough are these agentic systems其实在过去六个月里发生的事情 最大的突破是这些代理系统
31that are grounded on knowledge that can use tools to do search and has memory that it以可以使用工具进行搜索并具有内存的知识为基础
32manages and it has safeguards and has the ability to iterate until it gets the job done.管理并有保障措施,并有能力延展,直到完成任务。
33But it ultimately needed some models that have reached a level of capability where everything但它最终需要一些模型 已经达到一个水平的能力 在那里,一切
34comes together into that flash point.一起进入闪光点
35And that's where cloud code really kind of brought the imagination of agentic systems,这就是云码真正带来 代理系统想象力的地方
36OpenClaw, of course, was a big deal and all the work that you did, right, with deep agentsOpenClaw当然是个大问题 和你所做的所有工作 对 深层特工
37and we use that ourselves.我们用自己。
38And all of that kind of came together and bam, here we are with agentic systems.所有这种东西都聚集在一起,砰,我们有了代理系统。
39The reason why we do it, we've dedicated ourselves for many, many years to build open systems.我们这样做的原因是,我们多年来致力于建设开放的系统。
40And the reason for that is because ultimately, AI is a fundamental technology.其原因是,最终,AI是一种基本技术.
41It can only be useful if applied in a whole bunch of different use cases.只有在一整批不同的使用案例中应用才能有用.
42Now, of course, the first use case is just language and cognitive intelligence and that's现在,当然,第一个使用案例 仅仅是语言和认知智能,这就是
43very important, of course.当然很重要
44But we imagine a world where scientists and digital biologists and designers and roboticists但是,我们想象的世界 科学家和数字生物学家 设计师和机器人学家
45and students and researchers, enterprise IT, all of us could use agentic systems, AIs,和学生和研究人员,企业IT, 我们所有人都可以使用代理系统,AI,
46to solve domain-specific problems.用于解决特定域的问题。
47And many of the problems that we want to solve, either we have specialized domain knowledge我们想要解决的许多问题 要么我们有专业领域的知识
48that is just simply not available outside, that we have to embed into, imbue into our AI,我们不得不嵌入 嵌入我们的AI,
49or it's because we believe that AI becomes ultimately great, become a super agent, when或者是因为我们相信AI最终会变得伟大 成为超级特工
50we put it into a flywheel where we use it, it gets smarter, it becomes more useful, we我们把它放在飞轮上 用它,它变得更聪明,它变得更有用,我们
51get smarter, kind of like us, kind of like humans, learns over time.变得更聪明,有点像我们, 有点像人类, 学习随着时间的推移。
52And so we kind of imagine this future where AI has a foundation and the work that Anthropic我们想象着这个未来 AI有一个基础 还有Anthropic的作品
53OpenAI和Google都做得很棒
54But there's specialized AIs and domain-specific AIs and proprietary AIs that people want to但有专门的AI和专有域的AI和专有AI是人们想要的.
55build and we want to enable that world to happen.我们想让这个世界得以实现。
56Maybe digging into that for a second, on this topic of specialization, how exactly do you也许研究一下, 关于这个专业话题,你到底如何
57think it's best to specialize these systems?最好专门使用这些系统?
58Is it going to be purely the model, is it going to be the harness as well, the context会不会是纯粹的模型 会不会是绳子 环境
59outside?外面吗?
60Like what goes into the specialization?像什么进入专业?
61Well, the specialization starts with you need to have intelligence that's good enough.好吧,专业的开始,你需要 拥有足够的智能。
62And this is why we worked on Nemotron and really love the fact that you're part of the这就是为什么我们工作 在尼莫特隆 并真正喜欢的事实 你的一部分
63founding team of Nemotron Coalition.尼莫特龙联盟创始团队.
64我们让Nemotron Ultra非常不可思议。
65Now Nemotron Ultra is a great model as a start, but it becomes an incredible model when you现在Nemotron Ultra是一个伟大的模型 作为开始,但它成为一个不可思议的模型 当你
66put the Lang chain framework around it, the Lang chain harness around it, so that you把连锁框架放在周围 连锁绳套在周围
67ground it on information that is domain-specific and a person is an intelligent person becomes以特定领域的信息为基础 一个人是个聪明人
68super useful when we give them access to particularly important information.当我们让他们获得特别重要的信息时 是非常有用的
69And so access to information is important.所以获得信息很重要。
70Putting it into a flywheel where maybe you're even training the model, post-training the把它装到飞轮上 在那里你甚至正在训练模型 训练后
71model inside the Lang chain harness against the harness so that the model becomes good兰链内模型与绳索对齐 使模型变得良好
72at applying the harness around it.在它周围应用绳索。
73What you want it to do for that task.你希望它为这个任务做什么
74What you want it to do.随其所欲.
75Yeah.对
76And so I think that this moment has now arrived, but we need an open harnessing system that所以我认为这一刻已经到来 但我们需要一个开放的驾驭系统
77we can build ourselves, that we can apply, and then of course improve against over time.我们可以建立自己, 我们可以应用, 然后当然改善 与时间。
78I love what you said about the model being good enough.我喜欢你说的模特够好
79I feel like that threshold was crossed, I don't know, maybe a year ago by some of the我觉得那个门槛已经越过了 我不知道 也许一年前
80frontier models, six months ago by some of the open weight models.前沿模型,半年前 一些开放重量模型。
81You talked about Nemotron 3 Ultra.你说过尼莫特龙3号超人
82We've done a lot of work with that to make that really good in deep agents.我们为此做了很多工作, 使这在深层特工中是很好的。
83Some of the things we did is tweak the harness to make it best for this model because we我们所做的一些事情是 调整绳子 使其最好的 这个模型,因为我们
84found that different models need different prompts, different tools.发现不同的模型需要不同的提示,不同的工具.
85And with that tweaking, we managed to get Nemotron 3 Ultra in deep agents.有了这个变化, 我们设法让Nemotron 3 Ultra进入了深层特工。
86We have an internal benchmark and managed to get up to like 86% on that, Claude Opus我们有一个内部基准 并设法上升到86% 在那,克劳德·奥普斯
87for comparisons at 87.比较87。
88里面有DeepSeq和Minimax的模型 大约82,83
89So we're starting to see that some of the more recent open weight models are really因此,我们开始看到一些 最近的开放重量模型 真的
90getting to frontier performance.进入前沿表演。
91I know.我知道,我记得
92I am so proud.我很骄傲
93It is so incredible.太不可思议了
94Thank you.谢谢
95It is so incredible.太不可思议了
96One of the just as important thing is it's 10 times as cheap as Opus.同样重要的一件事是 它比奥普斯便宜十倍
97And I think open weight models are starting to really strike a good balance between performance我觉得开放式体重模型 正在真正在性能之间 取得良好的平衡
98and cost.和成本。
99So I'd be curious how you see this cost part changing the equation for builders.所以我想知道你是如何看待这个成本部分 改变建造者的方程式的。
100The benefit of cost, it comes in a couple of different ways.成本的好处, 它有几种不同的方式。
101I happen to think that when you have cost effective intelligence, people just use more我碰巧认为,当你有 成本效益高的智能, 人们只是使用更多
102of it.对
103When you have a cost effective agent, then you can iterate across a larger search space.当你有一个成本效益高的代理, 然后你可以跨越一个更大的搜索空间。
104And as a result, the answer could actually be better.结果,答案实际上可能更好。
105And in the case of Nemotron, it's cost effective because it's so fast.而对于尼莫特龙来说,由于速度如此之快,所以成本效率很高.
106It's so computationally efficient.计算效率很高
107When it's computationally efficient, it could explore larger spaces.在计算效率时,它可以探索更大的空间.
108And it's no different than when somebody can think fast, you could explore more space.和人们能快速思考的时候一样 你可以探索更多的空间
109When you can try things more quickly, you can find a better answer.当你能更快地尝试事物,你可以找到更好的答案.
110And so this is the incredible benefit of Nemotron 3 Ultra inside the Lang chain framework and因此,这是尼莫特龙3号超超强星在朗链框架内 令人难以置信的好处
111the Lang chain harness inside DeepAgent.朗链系在深潜特工体内
112It could think so quickly, it could explore so quickly.它能迅速思考,能迅速探索.
113It could iterate so quickly and efficiently that it's going to find better answers.它能快速高效地移动 从而找到更好的答案
114And so I'm just really, really excited that we created a model that was near the frontier.所以我非常非常非常兴奋 我们创造了一个靠近前沿的模型。
115But adapting the environment around Nemotron, you made it deliver frontier capabilities.但是适应尼莫特龙周围的环境, 你让它提供了前沿能力。
116Now insight, you know, for humans, it's kind of the same.现在的洞察力,对于人类来说,是一样的。
117Of course, we like to hire the smartest people in the world.当然,我们喜欢雇用世界上最聪明的人。
118But beyond that, we also give them access to tools, we give them access to information.但除此之外,我们还让他们获得工具,让他们获得信息。
119And we also create the world around them so that we, you know, we allow them, enable them我们还创造周围的世界 这样我们,你知道,我们允许他们,让他们
120to create the conditions for them to achieve their full potential.创造条件,使他们充分发挥潜力。
121And so you adjust the environment, not just the model.所以你要调整环境, 而不仅仅是模型。
122And this is where Lang chain came in.这是Lang连锁店的入口
123What you said about using more of the intelligence as it's cheaper and faster.你说的多用一些智能 因为它更便宜更快
124We see that to be so true.我们认为这是事实。
125I think one of the things that I like to think I'm AI forward, one of the things that I've我觉得我喜欢的东西之一 认为我是AI向前, 我曾经的东西之一
126estimated is just the demand for intelligence and for tokens and how big and massive that估计只是对情报和信物的需求 以及数量和数量
127market is.市场是。
128And I think that's become especially true recently with these models getting good and我觉得最近这些模型越来越好
129being really fast and really cheap.很快而且非常便宜。
130How should we think about using frontier models?我们该如何考虑使用前沿模型?
131Should we just use these open source models all the time?我们是否应该一直使用这些开源模型?
132Is there is there a time and place for both?二者有无时有处.
133Well, the frontier models are getting better all the time.边疆模型越来越好
134And I fully expect the frontier models to be unbelievably good.我完全期望边疆模型会非常优秀。
135And they still have a long, long runway of improving the models that scaling scaling而且他们还有一条长长的跑道 来改进那些缩放模型
136laws, of course, are going to sustain their harnesses are improving all the time.当然,法律将维持他们的优势 不断改善。
137Their technology for dealing with memory, whether it's working memory or long term他们处理记忆的技术,无论是工作记忆还是长期记忆
138memory is advancing incredibly quickly.记忆在快速前进
139And the compaction technologies, they're all of the advancements in retrieval,而压缩技术,它们都是 检索的进步,
140retrieval augmented generation and knowledge graphs.检索增加了生成和知识图表。
141There's still a lot of incredible advances that are being implemented into these还有很多令人难以置信的进步 正在落实到这些
142frontier model APIs.前沿模型 API.
143The thing that the way I think about it is, frankly, I always start all of my work老实说,我总是开始所有的工作
144starting with the frontier.从边境开始
145And the reason for that is because because it's useful.其原因是因为它有用。
146I know what's the potential cost a little bit more money, but it's incredibly我知道什么是潜在的 花费多一点钱,但它令人难以置信
147my time to getting to work done is fast.我工作的时间很快
148However, you know, over time, I find that I want to add sub agents to them.然而,你知道,一段时间后,我发现我想在他们中增加副特工。
149I want to connect sub agents that are super agents at certain skills.我想联系那些具有某种技能的超级特工
150And so we have, you know, we have optimization problems inside our company that所以我们有,你知道, 我们公司内部的优化问题,
151that relates to supply chain.这与供应链有关。
152Maybe it's related to chip design optimization, floor planning optimization.可能与芯片设计优化,地板规划优化有关.
153And these problems, these optimization problems are insanely hard.这些问题,这些优化问题都非常困难。
154And so you're not going to just have an a general AI go off and crunch on it and所以,你不会让 一个将军AI去 并压缩它,
155think that you're going to find a great answer.认为你会找到一个伟大的答案。
156We create super sub agents and these super up sub agents we would create with我们创建超级子代理 和这些超级高级子代理 我们会与
157deep agents, Lang chain deep agents with Neymotron 3 inside.深层探员 朗把深层探员和内森三号连在一起
158And we'll even connect them to specialized tools.我们甚至会把他们和专用工具联系起来。
159And that thing is built for one job.那东西是为一份工作而建造的。
160That super agent is not trying to, you know, book me travel, travel appointments.那个超级经纪人并不想,你知道, 订我的旅行,旅行预约。
161It's just trying to optimize our supply chain.它只是想优化我们的供应链。
162And in that case, I really do need to have Lang chain.这样的话 我真的需要Lang的锁链
163我真的需要有Neymotron 3 Ultra。
164And I connected to a lot of proprietary knowledge and proprietary skills.我和很多专有知识和专有技能联系在一起。
165I've got a whole team who's just dedicated to refining that.我有一队人 致力于完善它。
166Now, I think that that defines a company.现在,我认为这定义了公司。
167A company is really about a collection of a whole bunch of these super, super公司其实是一大堆超级超人
168proprietary, super important workflows.专有,超重要的工作流程.
169And now we can have Lang chain with deep agent and ultra, you know, Neymotron 3现在我们可以让Lang 带着深层特工和超超能力特工
170inside and it gives them all of the control they need.里面有他们需要的所有控制
171Super efficient access to incredible tools.超级高效的进入 不可思议的工具。
172You know, that's the future.你知道,这就是未来。
173Do you have any advice for enterprises if they're following your practice of你对企业有什么建议吗 如果他们遵循你的做法
174starting with the frontier and then starting to specialize?从边境开始,然后开始专攻?
175When should they think about specializing?他们什么时候该考虑专业?
176What are some triggers that you look for that?你在找什么触发器?
177As soon as it gets good enough, you know?一旦它变得足够好,你知道吗?
178So I would start with cloud code and codecs and use it for as long as I can.所以我从云码和解码器开始 尽可能长时间使用
179And frankly, for a lot of things, you never have to replace, you know,坦白地说,对于很多事情, 你永远不用替换,你知道,
180because they're getting better all the time.因为他们越来越好 所有的时间。
181And they're on a trajectory that's going to bring capabilities insanely fast.并且它们正在一个轨迹上, 将带来能力 疯狂的快速。
182And so I think that in the future, just like companies are today,所以我认为,在未来, 就像今天的公司,
183we have employees that we hire because their domain specialization我们有员工,我们雇用他们 因为他们的领域专业化
184and the refinement of the work and the work process and all of their life改进工作、工作过程和一生
185learnings here in the company is just too valuable.公司里的学习太有价值了
186But we also hire consultants and we license external tools and we outsource但我们也雇了顾问 给外部工具发许可证 然后外包
187work to other people and so on and so forth.对其他人的工作等等。
188I think this is the future for AI, you know?我认为这是AI的未来,你知道吗?
189And so are we going to continue to use frontier models?那么我们要继续使用前沿模型吗?
190Absolutely, and tons of it.当然,还有很多
191But are we also going to create specialized, you know, super agents但我们也要创造出专业的特工吗?
192with Lang chain and Nemotron 3 Ultra that, in fact, arguably could be your crown jewels?兰连锁店和尼莫特龙3号超市 可以说是你的皇冠珠宝?
193And the answer is absolutely true.答案是绝对正确的。
194I think even for the consultants that you bring in,我想即使你请来的顾问
195just like when you bring in a consultant, you need to get them up to speed就像你请来一个顾问的时候, 你得让他们加快速度
196in your organization and give them context on the organization, how things work,在你的组织, 并给他们的背景 组织,事情如何工作,
197what tools do they need that have access to data that's only inside your organization.他们需要哪些工具 能够获取你组织内部的数据
198And so I think like one of the things we've seen is as enterprises start to adopt AI,所以我想,就像我们所看到的一样 企业开始采用AI,
199there's all of these kind of like systems that they have to build around them in order to make他们必须围绕这些系统来制造
200the agentic systems as a whole trustworthy and safe and proper kind of like governance.整个代理系统都是值得信赖的 安全而恰当的治理
201I'm curious, how do you see?我很好奇 你怎么看出来的?
202And just to add on that, just, you know, today,顺便说一句 今天
203most companies are built on business processes.大多数公司都以业务流程为基础。
204In the future, most companies will be built on harnesses.未来,大多数公司都将建立在利用的基础之上。
205And so the idea, you know, Lang chain would just become the tool that creates the operating所以这个想法,你知道,兰链 将只是成为工具 创建操作
206system for the company.公司的系统。
207And everybody will be using Lang chain to create their specialized harness,每个人都会利用Lang连锁公司 来创造他们的专用绳索
208which represents a workflow of the past.这代表了过去的工作流程。
209And now this harness inside that workflow becomes autonomous, agentic, you know,而现在,这个拉带 在工作流程中 变得自主, 代理,你知道,
210much more efficient.效率要高得多。
211I think we see that these things are, there's the harness, there's the model,我认为我们看到这些东西是, 有绳子,有模型,
212and then there's all the context around it.然后是周围的所有背景。
213And all of these can be optimized at different points in time.而所有这些都可以在不同的时间点优化.
214That's right.没错
215And so the work that we did with Nemotron 3, I think was a great example of doing因此我们与尼莫特龙3的工作 我认为是一个很好的例子
216some pretty high ROI things around the harness, changing the prompt, changing the tools.一些相当高的ROI 事物在拉带周围, 改变迅速, 改变工具。
217One of the things we're looking forward to is experimenting with post-training我们期待的一件事是试验 后训练
218Nemotron to, it takes a little bit more time, but I think it really raises the ceiling尼莫特隆需要多一点时间 但我觉得这真的能提高天花板
219of what this overall system can do.这整个系统能做什么。
220This is incredible.这是不可思议的。
221This is the big breakthrough.这是大突破
222And so what you just described is a future where, once you get the harness built,所以你刚才描述的是一个未来 一旦你得到了这个带子,
223and it's doing the work, and it's now part of the business process,而它正在做工作, 现在它的一部分 业务流程,
224and it's very successful.它非常成功。 和它非常成功。
225Now the question is, how do we get it even better than that?现在的问题是,我们如何得到它甚至更好?
226Of course, you can keep improving the information that you provided.当然,你可以不断改进你提供的信息。
227You can tune the harness.你可以调谐绳索。
228But you can now also improve the AI model, the large language model,但你现在也可以改进AI模型,大型语言模型,
229Nemotron 3 Ultra, inside the harness.尼莫特龙3号超极,在绳索内
230I think that's a complete breakthrough, a capability that's never existed before.我认为这是一个完全的突破,一种从未有过的能力.
231And I'm super excited about that.我对此非常兴奋
232And it's really going to take all of these enterprise-specific business processes并且它真的要采取 所有这些企业特有的业务流程
233and really start to tune this flywheel.并开始调整这个飞轮。
234And I think one of the things that we've heard when talking to enterprises is我想当我们跟企业交谈时听到的一件事就是
235the demand or need for this to be built on an open ecosystem.建立开放生态系统的要求或需要。
236This is all of this enterprise's knowledge and processes that they're putting in there.这是企业所有的知识 和流程 他们在那里。
237Having full control over that seems paramount to a lot of them.完全控制着这件事对很多人来说都至高无上
238So I'm curious if you can touch on how you see open stacks所以,我很想知道,如果你可以接触 如何看到打开的堆
239really empowering enterprises going further with AI.真正赋予企业更大的权力
240Every company is built fundamentally on domain-specific每个公司都以特定领域为基础
241or some specialized intellectual property.或某些专门的知识产权。
242The reason why we call it intellectual property, intellectual, it's intelligence.我们之所以称之为知识产权,知识,就是智力.
243Every single company is built on intelligence,每个公司都靠情报工作
244some foundation of intelligence that's specialized.某种专门的情报基础
245Our company is specialized in something.我们的公司是专门从事某事的
246We're not good at everything, but we're very, very good at one thing.我们并不擅长一切,但我们非常,非常擅长一件事.
247And every company is built that way, you know?每个公司都是这样建的,你知道吗?
248And so that specialization, your company's intelligence is who you are.所以,专业,你的公司智能就是你是谁。
249You can't possibly not continue to control it, improve it, make it better, right?你不能不继续控制它, 改进它,使它更好,对不对?
250And somehow outsourcing that intelligence, whether you're a person, company,不知何故外包了情报 不管你是个人还是公司
251country, makes no sense to me.国家,对我来说毫无意义。
252And so, of course, there's general intelligence.因此,当然还有一般的智能。
253And there are general things that we all do.也有一些一般的事情 我们都做。
254And it's a lot of the economy, you know?和它很多的经济,你知道吗?
255And for example, software coding is actually a general thing.举例来说,软件编码其实是一般的事情.
256我们都在Python计划。
257We all program in C++.我们都用C++程序
258We all program, right?我们都在计划 对吧?
259And so you're applying it to different things,所以你把它应用到不同的事物上,
260but the skill of coding is largely the same.但编码技巧基本相同.
261And that's a general skill.这是一般的技巧
262Writing is a general skill.写作是一种一般的技能.
263But those are foundational skills that we then apply for our specialized domain intelligence.但是,这些是基础技能, 然后我们申请 我们的专业域情报。
264And that's where Langchain and Neemotron comes in.兰辛和尼莫特隆就从这里进来了
265I think the foundation of society is going to have these foundational models,我认为社会的基础 将会有这些基础模式,
266and they're going to be general, and they're going to be available in the cloud,且将作大将军,云中可用.
267and it's going to be incredible.这将是不可思议的。
268But on top of that platform, we're going to have to build our own specialized capabilities,但是在这个平台上 我们必须建立我们自己的专业能力
269and you need open tools for that.你需要打开的工具。
270And you can't outsource it.则不能外包.
271You know, I can't imagine calling a third party when I need to enhance my intelligence.你知道,我不能想象 打电话给第三方 当我需要增强我的智力。
272You know, I need to enhance it right here inside the company.我需要在公司内部加强它
273And so I think that that future is not one or the other.所以我认为未来不是一对一。
274It's a completely complementary vision.这是一个完全互补的愿景.
275And really what we're doing is just making sure that we automated intelligence is integrated我们要做的就是确保我们自动智能集成
276into all aspects of everything that we do.我们所做的一切
277And as a result, you know, we're all going to be better.结果我们都会好起来的
278Completely agree.完全赞成。
279And I think it's still hard to get that integration up and running.我认为仍然很难实现整合。
280And so one of the things that we're announcing today is a blueprint with deep agents所以我们今天宣布的一件事 是一份有深层特工的蓝图
281and open shell inside of the Neemoclaw blueprints out there.并打开壳 在尼莫茨瓦夫蓝图 在那里。
282And so this will let enterprises run deep agents with Neemotron 3 Ultra这样企业就能和Neemotron 3 Ultra 一起经营深层代理
283inside of open shell, which is a secure and open runtime and take advantage of that.打开的外壳里面,这是一个安全开放的运行时间 并利用它。
284This is one of just such a huge deal.这是一个如此巨大的交易。
285Hopefully it makes it way easier for enterprises to get up and running.希望这能让企业更容易起来和运行。
286Yeah.对
287All of the key ingredients necessary for you to build your personal domain specific为您构建个人领域所需的所有关键要素
288proprietary, you know, your super agent, all of the technologies, all the components,独家,你知道,你的超级代理, 所有的技术,所有的组件,
289all the tooling, all of the harnessing, everything and the blueprint, a great example,所有的工具,所有的操纵, 一切和蓝图,一个伟大的例子,
290all put together for you.都为你准备好了
291How do you guys think about blueprints?你们觉得蓝图怎么样?
292You have many of them.你有很多这些
293This is obviously the best one.这显然是最好的。
294I won't make you say that, but I'll say that this is the best one out there,我不会逼你这么说的 但我会说这是最好的
295but you have a ton of blueprints.但是你有一堆蓝图
296Why invest so heavily in them?为什么投资这么大?
297Well, because the tools are arcane still, you know, and there are a lot of pieces to it,嗯,因为工具仍然是 arcane 仍然,你知道, 有很多块它,
298you know, building an agentic system, building AI is not simple.你知道,建立一个代理系统, 建立AI不是简单的。
299And there's a lot of different pieces of technology.并有许多不同的技术.
300And we already talked about some of them.我们已经谈过其中一些了。
301There's the large language model.有大语言模型.
302There's the tools it uses and, you know, the knowledge graph that it has to deal with.有它使用的工具 还有它必须处理的知识图
303It's memory system and it's guard railing system and it's fine tuning system.它是记忆系统,也是警卫栏杆系统,也是精细的调制系统.
304Now, you know, the technology, you're going to create the post-training against the harness.现在,你知道,技术, 你要去创建 后训练与牵引。
305And then, of course, there's the harness itself.当然,还有绳子本身。
306But what about the runtime?跑步时间呢?
307When you're done, you still have to run time.完成后,仍须运行时间.
308You have to keep it in a sandbox so it's secure, it's private.你必须把它藏在沙盒里 这样它才安全,是私人的
309That access control is something that IT organizations can control.这种访问控制是IT组织可以控制的.
310Is that the hardest thing about the runtime?跑步时间最难吗?
311You think inside of enterprises, all the security things that go alongside it?你想想企业内部, 所有安全的东西 一起走?
312Without solving the security, the access control, it's impossible to deploy.如果不解决安全,出入控制问题,就无法部署.
313It's no different than it's impossible to hire a new employee into the company.这与公司聘请新员工是不可能的一样.
314If you don't onboard them, give them access control.如果你不在船上, 给他们进入控制。
315We don't give every employee access to every file and every network, right?我们不给每个雇员 访问每一个文件和每个网络,对不对?
316And so you have every single employee based on their job and their responsibility因此,你每个员工 基于他们的工作和责任
317and what they need to have access to.以及他们需要获得的东西。
318We give them access to tools, the laptops and, you know, design tools and, you know,我们给他们提供工具 笔记本电脑 设计工具
319programming tools and whatnot.编程工具什么的。
320We give them access to certain parts of the network.我们让他们进入网络的某些部分
321We give them access to information.我们让他们获得信息。
322We connect them to other agents.我们把他们和其他特工联系起来
323We connect them to other colleagues that they work within.我们把他们与他们在其中工作的其他同事联系起来。
324And we provide them a skills file, you know?我们给他们一个技能档案 你知道吗?
325We essentially, you know, give them a document about this is your mission.我们基本上,你知道, 给他们一个文件 关于这是你的任务。
326This is how it's previously been done.之前就是这样做的
327And now, you know, help do it even better than that.现在,你知道,帮助做它甚至更好。
328And so in a lot of ways, we are creating an HR system, if you will, you know, for AI因此,在许多方面,我们正在创建一个HR系统, 如果你愿意,你知道,为AI
329that allows the IT organizations and all of the different business units inside the companies信息技术组织和公司内部所有不同的业务单位
330to be able to build, improve and deploy these agents inside companies.在公司内部建立、改进和部署这些代理人。
331This is more of a philosophical question, but you're talking and I think a lot of people这更是一个哲学问题 但你在说话 我认为很多人
332talk about these agents and anthropomorphize them a lot, bring them into human systems.说到这些特工 并把他们变成人类 带入人类系统
333But agents aren't human and they have some things that are better than what humans are.但特工不是人类 他们有些比人类更好的东西
334And they have other places where they are, you know, very different他们有其他地方 他们在那里,你知道,非常不同
335and maybe not as good as what humans are good at.也许不如人类擅长的好
336What is the right level to anthropomorphize these agents?将这些人化为人类的正确层次是什么?
337Well, it's electrons, you know, it's electrons, not atoms.是电子,是电子,不是原子
338And it's not biological, has no consciousness, you know, it's not awake.而且它不是生物的,没有意识的, 你知道,它没有清醒。
339And so it's not any of that.所以它不是任何这样的。
340It's a tool that it's kind of like my vacuum cleaner that's roaming around the house.这是一个工具,它有点像我的吸尘器 在房子周围游荡。
341And it's, of course, roaming around the house, cleaning up the house,当然,它在房子周围游荡, 清理房子,
342doing something that I used to do.做我以前做过的事
343And, you know, have autonomous lawnmowers and you have, you know, and so I,而且,你知道,有 自主的草坪和你有,你知道, 所以我,
344you could just imagine a hundred years ago when the first dishwasher came along.你可以想象一百年前 第一个洗碗机来的时候
345And now it's doing dishes by itself.现在它自己做菜
346It must have been magical to watch it and we call it a dishwasher,看着它一定很神奇 我们称之为洗碗机
347which is a little bit like a human.有点像人类
348And so we have dishwashers.所以我们有洗碗机。
349When my first job, I was a dishwasher.当我第一次工作的时候,我是一个洗碗机。
350And so, you know, and so in a lot of ways, I mean, you know, we'll get used to it.所以,你知道, 所以在许多方面, 我的意思是,你知道,我们会习惯它。
351I think it's right now, we tend to imbue too much human properties to it.我认为现在,我们倾向于把 太多的人性化为它。
352It's nothing close to that.何等为近.
353It's software, it's computers.是软件 是电脑
354We know exactly how it's working because obviously we created the harnesses around it.我们非常清楚它是如何工作的, 因为很明显我们创造了周围的绳子。
355We obviously know how it works because it's getting better all the time.我们显然知道它是如何工作的,因为它越来越好。
356If we don't understand how something works, how do we make it better every time?如果我们不明白 东西是如何工作的, 我们如何使它更好?
357And so if we don't understand how something works, how do we improve it?所以,如果我们不明白 东西是如何工作的, 我们如何改进它?
358You know, how do we fix it?你知道,我们如何修复它?
359And so obviously we understand how these things work.很明显,我们理解这些事情是如何运作的。
360And I think that we ought to keep it there, you know.我认为我们应该把它放在那里,你知道的。
361And meanwhile, one of the things that we know is that the more AI we use,与此同时,我们所知道的一件事 就是我们使用的AI越多
362somehow the more people we have to hire.不知何故,我们不得不雇用更多的人。
363And the reason for that is because these agentic systems are new skills.这是因为这些代理系统是新的技能。
364And now we have a lot of software engineers building agents.现在,我们有许多软件工程师 建筑代理。
365They used to code software, but now they're building agents.他们以前用代码软件,但现在他们正在建造代理.
366If you ask me, every one of my software engineers prefer to be building agents如果你问我,我的软件工程师都喜欢做建筑代理
367而不是写Python代码
368Coding is like typing.编码如打字.
369And so they're going to do less typing.所以他们会少打字。
370They're going to be more systems engineers, more building engineers, you know, building他们将会是更多的系统工程师, 更多的建筑工程师,你知道,建筑
371and creating these autonomous systems that are super cool.并创建这些超酷的自主系统。
372They're creating evals.他们正在创造Evals。
373They're creating benchmarks.他们创造了基准。
374They're creating guardrails, isn't that right?他们在制造护栏 不是吗?
375And so the amount of work that we have to do to bring AI into the world因此,我们必须做大量的工作 让AI进入世界
376is really quite incredible.真是不可思议
377And so it's creating a whole bunch of jobs.所以它创造了一整批的工作。
378And my software engineers love this.我的软件工程师很喜欢这个。
379I think we've seen, you mentioned evals briefly.我想我们已经看到了, 你简短地提到了Evals。
380I think we see that being a key part to unlocking a lot of agentic usage我认为,我们看到,这是一个关键 部分解锁 很多代理用法
381inside an enterprise.企业内部
382You need to have some sense of how it's doing.你需要有一点感觉 它是如何做的。
383And quantifying whether it's good or not is oftentimes best done by subject matter并量化它是否好 通常是最好的时间 通过主题
384experts who already live inside the enterprise and can easily give feedback已经住在企业里并且可以轻易提供反馈的专家
385and work with these systems to automate a lot of the tedious parts of their job和与这些系统合作 自动化 很多乏味的工作部分
386and then spend time on the really intellectually stimulating parts然后把时间花在 真正的智力刺激部分
387and the creative parts.和创作部分。
388And so that's right.所以这是对的。
389In a lot of ways, whether you're a doctor or a designer or a software engineer,在很多方面,无论是医生,设计师还是软件工程师,
390you are creating an agent and you're taking all the mundane work你正在创造一个代理 而你要承担所有的工作
391and you're trying to get this agent to do it.你试图让这个特工来做
392But meanwhile, we're all trying to get our agents elevated但与此同时 我们都在试图提升我们的特工
393to do things with us that we couldn't do before.做我们以前不能做的事
394And so that requires imagination.而这需要想象力。
395That requires creativity, you know, a lot of technology.这需要创造力,你知道, 很多技术。
396I think that's spot on.我想这是位置。
397I think currently, a lot of the best usages that we see of agents我认为,目前,很多最好的用途 我们看到的代理
398are giving ourselves more leverage to do more things.给自己更多的影响力 做更多的事情。
399But I think a lot of that approach is thinking about what did we do previously但我认为很多方法 是在思考我们以前做了什么
400and can we automate that?我们可以自动化吗?
401But I think a lot of the unlock will come in the future但我觉得将来会有很多解锁的
402of what couldn't we do before.我们以前不能做的
403That's right.没错
404That now we can do.现在我们可以做到这一点。
405And so maybe-所以也许...
406Ambition helps.野心有帮助。
407A hundred percent.百分百都一样
408Ambition, agency.野心,机关。
409Ambitions help.野心有帮助。
410Maybe on that vein, wrapping up, as you think about how to help drive towards this future,也许在那个脉络, 结束, 当你想怎样 帮助驱动未来,
411what are some of the missing pieces of this agentic stack?这个代理堆的残骸是什么?
412Well, today we're announcing a very big deal.今天我们要宣布一件大事
413I mean, this is a very big deal thing that we're doing today.我的意思是,这是一个非常重大的事情 我们今天做的。
414We are providing the basic building blocks, the foundation,我们正在提供基本的构件,基础,
415all of the key ingredients.所有关键成分
416All of the key ingredients to build super agents.建造超级特工的关键成分
417When I say super agents, I mean, they're domain specific.当我说超级特工时,我是说,他们是特定领域。
418They belong to you.他们属于你。
419You build them, you improve them, you refine them over time.你建造它们,改进它们,并随着时间的推移加以完善。
420You give them access to proprietary information, knowledge.你让他们获得专有信息 知识
421Maybe it's super private to you.也许对你来说是超级隐私
422And as a result, this super agent will be able to do things that you can imagine.由此,这个超级特工将能够做你能想象的事情.
423And it will be extremely, extremely good.这将是极好的,极好的。
424We've created all of the key parts.我们创造了所有的关键部分。
425A world-class language model.世界一流语言模式.
426A framework called Langchain Deep Agents that has now been also fine-tuned一个叫做"Langchain深层特工"的框架 现在已经做了微调
427to expose the full potential of the Nemotron 3 Ultra.揭开尼莫特龙3号超极星的全部潜力.
428A blueprint that helps everybody do that.帮助每个人这样做的蓝图。
429And of course, the runtime, the open shell runtime that keeps it secure.当然,运行时间, 开放的壳运行时间 保持它的安全。
430The acceleration stacks that are all integrated into it.全部融入其中的加速度堆栈.
431And so every company in the world should be able to,所以世界上的每一个公司都应该能够
432every developer in the world should be able to now create these super agents,世界上每个开发商 都应该能创造出这些超级代理
433deploy it anywhere in the cloud, on-prem.在云中任何地方部署它,在预告。
434A good friend of mine just built one for DGX Spark.我的一个好朋友刚为DGX Spark造了一台
435And so now you have these agents running on DGX Spark right next to your laptop.现在你让这些特工 运行在DGX Spark 在你的笔记本电脑旁边。
436You could have it running on a DGX station.你可以让它运行在一个DGX站。
437You could build your own supercomputer inside your company,你可以在你的公司里建造自己的超级计算机
438if you like, or do it in the cloud.如果你喜欢,或者在云中做。
439We now have agentic capabilities that you can now build for yourself everywhere.我们现在有你现在可以为自己建立的能力。
440All the pieces are now here.所有碎片都在这里
441There are no excuses not to engage it.没有借口不参与。
442I think that's a perfect way to end it.我想这是结束这一切的绝佳办法
443You got me so pumped up when you were speaking.你说话的时候让我很兴奋
444A great motivational speach.伟大的动力弹簧。
445So I'm going to go out and build some.所以我要出去建造一些。
446I'm going to go down and build some agents.我要下楼盖些探员
447Thank you, Johnson.谢谢你,约翰逊。
448Congratulations.祝贺你
449Good job.干得不错
450Proud of you guys.为你们骄傲
451Job, proud of you guys.Job,为你们骄傲。
452Thank you.谢谢
该视频共有字幕 453 条。解锁更多字幕为会员功能,请移动到 价格