Name: 中英字幕】NVIDIA GTC 2025發表機器人12分鐘全精華！黃仁勳 Keynote 點名的下一個大市場 (【中英字幕】NVIDIA GTC 2025發表機器人12分鐘全精華！黃仁勳 Keynote 點名的下一個大市場)
Uploaded: 2025-03-21T03:39:28.000Z
Duration: 14 min 44 s

我們來談談機器人技術，好嗎？

Well, the time has come, the time has come for robots.

Robots have the benefit of being able to interact with the physical world and do things that otherwise digital information cannot.

機器人的優勢在於能夠與物理世界互動，做一些數字信息無法做到的事情。

We know very clearly that the world has severe shortage of human laborers, human workers.

我們清楚地知道，世界嚴重缺乏人類勞動力和人類工人。

By the end of this decade, the world is going to be at least 50 million workers short.

到本十年末，全世界將至少缺少 5000 萬名工人。

We'd be more than delighted to pay them each $50,000 to come to work.

我們非常樂意為他們每人支付 5 萬美元，讓他們來工作。

We're probably gonna have to pay robots $50,000 a year to come to work, and so this is going to be a very, very large industry.

我們很可能要支付機器人每年 5 萬美元的工資來工作，所以這將是一個非常非常龐大的產業。

你們的基礎設施將是機器人式的。

Billions of cameras in warehouses and factories, 10, 20 million factories around the world.

倉庫和工廠裡有數十億個攝像頭，全球有 1,000 萬、2,000 萬個工廠。

Every car is already a robot, as I mentioned earlier, and then now we're building general robots.

正如我之前提到的，每輛汽車都已經是一個機器人，現在我們正在製造通用機器人。

讓我告訴你我們是如何做到這一點的。

Everything that moves will be autonomous.

Physical AI will embody robots of every kind in every industry.

物理人工智能將體現在各行各業的各類機器人中。

Three computers built by NVIDIA enable a continuous loop of robot AI simulation, training, testing, and real-world experience.

由英偉達™（NVIDIA®）公司製造的三臺計算機實現了機器人人工智能模擬、訓練、測試和實際體驗的連續循環。

Training robots requires huge volumes of data.

Internet-scale data provides common sense and reasoning, but robots need action and control data, which is expensive to capture.

互聯網規模的數據可以提供常識和推理，但機器人需要行動和控制數據，而獲取這些數據的成本很高。

With blueprints built on NVIDIA Omniverse and Cosmos, developers can generate massive amounts of diverse synthetic data for training robot policies.

利用基於 NVIDIA Omniverse 和 Cosmos 構建的藍圖，開發人員可以生成大量不同的合成數據，用於訓練機器人策略。

First, in Omniverse, developers aggregate real-world sensor or demonstration data according to their different domains, robots, and tasks, then use Omniverse to condition Cosmos, multiplying the original captures into large volumes of photoreal diverse data.

首先，在 Omniverse 中，開發人員根據不同的領域、機器人和任務彙集真實世界的傳感器或演示數據，然後使用 Omniverse 對 Cosmos 進行調節，將原始捕獲數據乘以大量逼真的多樣化數據。

Developers use Isaac Lab to post-train the robot policies with the augmented dataset, and let the robots learn new skills by cloning behaviors through imitation learning, or through trial and error with reinforcement learning AI feedback.

開發人員利用 Isaac 實驗室使用增強數據集對機器人策略進行後期培訓，讓機器人通過模仿學習克隆行為，或通過強化學習人工智能反饋進行試錯，從而學習新技能。

Practicing in a lab is different than the real world.

在實驗室裡練習與現實世界不同。

Developers use Omniverse for software and hardware-in-the-loop testing, simulating the policies in a digital twin with real-world environmental dynamics, with domain randomization, physics feedback, and high-fidelity sensor simulation.

開發人員利用 Omniverse 進行軟件和硬件在環測試，在數字孿生中模擬具有真實環境動態、域隨機化、物理反饋和高保真傳感器模擬的策略。

Real-world operations require multiple robots to work together.

現實世界中的操作需要多個機器人協同工作。

Mega, an Omniverse blueprint, lets developers test fleets of post-train policies at scale.

Mega 是 Omniverse 的一個藍圖，可讓開發人員大規模測試車隊的列車後政策。

Here, Foxconn tests heterogeneous robots in a virtual NVIDIA Blackwell production facility.

在這裡，富士康在虛擬的英偉達 Blackwell 生產設施中測試異構機器人。

As the robot brains execute their missions, they perceive the results of their actions through sensor simulation, then plan their next action.

機器人大腦在執行任務時，會通過傳感器模擬感知行動結果，然後計劃下一步行動。

Mega lets developers test many robot policies, enabling the robots to work as a system, whether for spatial reasoning, navigation, mobility, or dexterity.

Mega 可讓開發人員測試多種機器人策略，使機器人作為一個系統工作，無論是空間推理、導航、移動還是靈巧性。

Today, we're introducing NVIDIA iZake Groot N1.

今天，我們要介紹的是 NVIDIA iZake Groot N1。

Groot N1 is a generalist foundation model for humanoid robots.

Groot N1 是仿人機器人的通用基礎模型。

It's built on the foundations of synthetic data generation and learning in simulation.

它建立在合成數據生成和模擬學習的基礎之上。

Groot N1 features a dual-system architecture for thinking fast and slow, inspired by principles of human cognitive processing.

Groot N1 採用雙系統架構，既能快速思考，也能慢速思考，其靈感來源於人類的認知處理原則。

The slow thinking system lets the robot perceive and reason about its environment and instructions and plan the right actions to take.

慢速思維繫統可讓機器人感知和推理其所處的環境和指令，並計劃採取正確的行動。

The fast thinking system translates the plan into precise and continuous robot actions.

快速思維繫統可將計劃轉化為機器人精確而連續的動作。

Groot N1's generalization lets robots manipulate common objects with ease and execute multi-step sequences collaboratively.

Groot N1 的通用性可讓機器人輕鬆操縱普通物體，並協同執行多步驟序列。

And with this entire pipeline of synthetic data generation and robot learning, humanoid robot developers can post-train Groot N1 across multiple embodiments and tasks across many environments.

有了這一整套合成數據生成和機器人學習管道，仿人機器人開發人員就可以在多種環境下對 Groot N1 的多種實施方案和任務進行後期訓練。

Around the world, in every industry, developers are using NVIDIA's three computers to build the next generation of embodied AI.

在全球各行各業，開發人員都在使用英偉達™（NVIDIA®）的三臺計算機構建下一代人工智能。

Physical AI and robotics are moving so fast.

物理人工智能和機器人技術發展如此之快。

This could very well be the world's first and very well likely be the largest industry of all.

這很可能是世界上第一個，也很可能是最大的產業。

At its core, we have the same challenges.

其核心是，我們面臨著同樣的挑戰。

As I mentioned before, there are three that we focus on.

正如我之前提到的，我們的重點有三個。

How, where do you create the data necessary to train the AI?

如何、在哪裡創建訓練人工智能所需的數據？

How can we scale either the data, the compute, or both so that we can make AIs smarter and smarter and smarter?

我們如何才能擴大數據、計算或兩者的規模，從而讓人工智能變得越來越聰明？

And those two, those fundamental problems exist in robotics as well.

機器人技術中也存在這兩個基本問題。

In robotics, we created a system called Omniverse.

在機器人領域，我們創建了一個名為 "Omniverse "的系統。

It's our operating system for physical AIs.

這是我們的物理人工智能作業系統。

You've heard me talk about Omniverse for a long time.

今天，我要向你們展示兩樣東西。

One of them is so that we could scale AI with generative capabilities and generative model that understand the physical world.

其中之一是，我們可以利用生成能力和能夠理解物理世界的生成模型來擴展人工智能。

Using Omniverse to condition Cosmos and using Cosmos to generate an infinite number of environments allows us to create data that is grounded, grounded, controlled by us, and yet be systematically infinite at the same time.

用 "宇宙 "來調節 "宇宙"，用 "宇宙 "來生成無限多的環境，這樣我們就能創造出腳踏實地的數據，這些數據由我們控制，但同時又是系統性的無限數據。

Okay, so you see Omniverse, we used candy colors to give you an example of us controlling the robot in the scenario perfectly, and yet Cosmos can create all these virtual environments.

好了，你看，"宇宙"，我們用糖果色給你舉了一個例子，我們可以完美地控制場景中的機器人，而 "宇宙 "卻可以創造出所有這些虛擬環境。

The second thing, just as we were talking about earlier, one of the incredible scaling capabilities of language models today is reinforcement learning, verifiable rewards.

第二件事，就像我們之前談到的，如今語言模型令人難以置信的擴展能力之一是強化學習，即可驗證的獎勵。

The question is what's the verifiable rewards in robotics?

問題是，機器人技術的可驗證獎勵是什麼？

And as we know very well, it's the laws of physics.

And so we need an incredible physics engine.

是以，我們需要一個令人難以置信的物理引擎。

Well, most physics engines have been designed for a variety of reasons.

大多數物理引擎的設計都有各種原因。

They can be designed because we want to use it for large machineries, or maybe we design it for virtual worlds, video games and such, but we need a physics engine that is designed for very fine-grained, rigid and soft bodies, designed for being able to train tactile feedback and fine motor skills and actuator controls.

它們可以是為大型機械而設計的，也可以是為虛擬世界、視頻遊戲等而設計的，但我們需要的物理引擎是為非常精細的剛體和軟體而設計的，是為能夠訓練觸覺反饋、精細運動技能和執行器控制而設計的。

We need it to be GPU accelerated so that these virtual worlds could live in super linear time, super real time, and train these AI models incredibly fast.

我們需要 GPU 加速，這樣這些虛擬世界就能以超線性時間、超實時時間運行，並以難以置信的速度訓練這些人工智能模型。

And we need it to be integrated harmoniously into a framework that is used by roboticists all over the world, Mojoco.

我們需要將它和諧地整合到一個框架中，而這個框架正是全世界機器人專家都在使用的，那就是 Mojoco。

And so today we're announcing something really, really special.

是以，今天我們要宣佈一件非常非常特別的事情。

It is a partnership of three companies, DeepMind, Disney Research, and NVIDIA, and we call it Newton.

它由 DeepMind、迪斯尼研究院和英偉達三家公司合作開發，我們稱之為 "牛頓"。

Tactile feedback, rigid body, soft body, simulation, super real time.

觸覺反饋、剛體、軟體、模擬、超實時。

Can you imagine just now what you were looking at as complete real-time simulation?

你能想象你剛才看到的完全是實時模擬嗎？

This is how we're gonna train robots in the future.

這就是我們未來訓練機器人的方式。

Just so you know, Blue has two computers, two NVIDIA computers inside.

你要知道，"藍色 "有兩臺電腦，裡面有兩臺英偉達™（NVIDIA®）電腦。