FIGURE 01 AI Robot Update w/ OpenAI + Microsoft Shocks Tech World (THEMIS HUMANOID DEMO)

Summary

OpenAI and Microsoft are investing $500 million in Figure AI, the creator of the Figure One humanoid robot, known for its advanced dexterity and versatility. The 167cm tall robot can handle up to 20kg payloads and is designed for industrial and service applications. With a lightweight build and impressive runtime, Figure One's development is accelerated by a team from tech giants like Boston Dynamics, Tesla, and Apple. The company focuses on system hardware, advanced AI for autonomous task completion, affordability, safety, and practical industrial applications, even considering a robotics-as-a-service model. The humanoid robot market is projected to surpass the personal computer market, with other companies like Westwood Dynamics and Nvidia also making significant advancements in AI robotics and simulation technologies.

Takeaways

• 🤖 OpenAI and Microsoft are investing $500 million into Figure AI, the company behind the highly capable Figure One humanoid robot.
• 🌟 Figure One is recognized as one of the most intelligent bipedal machines globally, with impressive dexterity and versatility exceeding human capabilities.
• 📈 In just over a year since its inception, Figure AI has made significant progress towards creating a general-purpose robot with human-like finger movements.
• 🏭 The Figure One robot stands at 167 cm and can handle payloads of up to 20kg, making it valuable for industrial and service applications, including BMW's production facilities.
• 🏃‍♂️ Weighing only 60kg, the robot is lightweight, agile, maneuverable, and has a battery life of five hours, with a top walking speed of 1.2m/s.
• 💡 Figure AI has assembled a 51-member elite team with employees from tech giants like Boston Dynamics, Tesla, and Apple, contributing to the robot's rapid development.
• 🛠️ Figure AI focuses on five key areas: system hardware, advanced AI, affordability and volume, safety, and real-world industrial applications.
• 💰 The company aims for high-volume manufacturing to reduce costs and make the robot accessible to a broader market through a potential robotics-as-a-service business model.
• 🔄 Nvidia's dual computer AI model, including the AI factory and runtime environment, is set to transform robot development and deployment.
• 🌐 China's Vary Toy is a compact large vision language model designed for standard GPUs, addressing the demand for efficient image perception in AI and overcoming computational challenges.

Q & A

How much investment is OpenAI and Microsoft planning to make in Figure AI?

OpenAI and Microsoft are planning to invest $500 million into Figure AI.

What makes the Figure One humanoid robot unique?

The Figure One humanoid robot is unique due to its capability and intelligence, being one of the most advanced bipedal machines in the world. It replicates and exceeds human finger movements in dexterity and versatility.

What record has Figure AI set in terms of development progress?

Figure AI has made world record progress towards creating a general-purpose humanoid robot, achieving significant advancements in just over a year from its inception.

What is the height and payload capacity of the Figure One robot?

The Figure One robot stands at a height of 167cm and is engineered to handle significant payloads of up to 20kg.

Which company is testing the Figure One humanoid in their production facilities?

BMW is testing the Figure One humanoid in their production facilities.

How does the Figure One robot's weight affect its performance?

The Figure One robot's lightweight design, at just 60kg, enhances its agility, maneuverability, and battery life, allowing for an impressive runtime of five hours on a single charge.

What is the size of the team assembled by Figure I and what is its composition?

Figure I has assembled an elite 51-member team, with many employees coming from tech giants like Boston Dynamics, Tesla, and Apple.

What are the five key areas Figure I is focusing on for its development?

The five key areas Figure I is focusing on are system hardware, advanced AI, affordability and volume, safety, and a robotics as a service model for its business.

What is the potential market size for humanoid robots compared to the personal computer market?

The humanoid robot market is poised to be even bigger than the personal computer market.

How does Nvidia's dual computer AI model impact the development and deployment of robots?

Nvidia's dual computer AI model, consisting of the AI factory and a runtime environment, transforms how robots are developed and deployed by leveraging data center resources for simulation and training, and providing adaptable runtime environments based on the robot's application.

What is China's Vary Toy and how does it address the limitations of existing vision vocabulary networks?

China's Vary Toy is a compact large vision language model designed for standard GPUs. It addresses the limitations of existing vision vocabulary networks by scaling up the vocabulary through training with a smaller autoregressive model and integrating it with the existing vocabulary, offering an efficient and effective solution for image perception in AI.

Outlines

00:00

🤖 Investment in Figure AI and the Figure One Robot

OpenAI and Microsoft are investing $500 million in Figure AI, the creator of the Figure One humanoid robot, recognized for its intelligence and capabilities. The robot, standing at 167cm, is designed to handle up to 20kg payloads and is valued for its industrial and service applications, with BMW testing it in their facilities. Weighing 60kg, the robot offers agility, maneuverability, and a 5-hour battery life. Figure AI's 51-member team includes experts from Boston Dynamics, Tesla, and Apple, focusing on system hardware, advanced AI, affordability, safety, and a robotics-as-a-service model. The company is also considering real-world industrial applications for the future. Meanwhile, Westwood Dynamics' Themis robot and Nvidia's dual computer AI model, along with China's Vary Toy, represent significant advancements in the field, emphasizing the growing potential of the humanoid robot market.

05:00

🚀 Nvidia's Generative AI and Robotics Innovations

Nvidia is revolutionizing the field of simulation and synthetic data generation with its dual computer AI model, which includes the AI factory for development and improvement of AI models, and a runtime environment tailored to the robot's application. The Isaac platform leverages Nvidia's data center resources for simulation and training, enhancing the AI's accuracy, performance, and adaptability. Nvidia's generative AI, integrated with the Isaac platform, allows for the creation of detailed scenes from text prompts and new 3D assets, significantly reducing the time and resources needed for simulation and data generation. This integration leads to more intuitive human-robot interactions and superior accuracy in various modalities. Nvidia's strategic partnerships demonstrate the impact of generative AI across sectors, shaping the future of robotic operation and interaction.

Keywords

💡Investment

The term 'investment' refers to the act of committing money, time, or other resources to a particular venture with the expectation of achieving a profit or benefit. In the context of the video, it highlights OpenAI and Microsoft's $500 million financial contribution to Figure, emphasizing their confidence in the company's potential and the growing significance of AI robotics in the industry.

💡Humanoid Robot

A 'humanoid robot' is a robot designed in the shape of a human, often mimicking human movements and capabilities. The video emphasizes the Figure One humanoid robot as a highly capable and intelligent bipedal machine, showcasing its advanced dexterity, versatility, and ability to handle significant payloads, which positions it as a valuable asset for various applications.

💡Dexterity

Dexterity refers to the skill and ease of performing tasks that require fine motor skills, precision, and control. In the context of the video, the Figure One robot's dexterity is highlighted as surpassing human capabilities, indicating its advanced ability to perform complex tasks with precision and agility.

💡Payload

Payload refers to the weight or items that a vehicle, such as a robot, is designed to carry or transport. In the video, the Figure One robot's ability to handle significant payloads of up to 20kg is emphasized, showcasing its utility in industrial and service applications where heavy lifting is required.

💡Agility

Agility refers to the ability to move with quick and smooth movements, often associated with flexibility and responsiveness. The video describes the Figure One robot as being relatively lightweight at just 60kg, which enhances its agility and maneuverability, allowing for efficient and dynamic operation in various environments.

💡Advanced AI

Advanced AI, or Artificial Intelligence, refers to sophisticated systems designed to perform tasks that typically require human intelligence, such as learning, problem-solving, and decision-making. In the video, Figure AI's development of an AI system capable of autonomous task completion by humanoid robots is one of its most ambitious goals, highlighting the company's focus on creating intelligent agents that can adapt to complex real-world environments.

💡Affordability and Volume

Affordability and volume pertain to making a product economically viable and accessible to a wide range of consumers by producing it in large quantities. In the video, Figure AI aims to reduce the robot's cost through high-volume manufacturing, indicating a strategy to make the technology more accessible and broaden its consumer market.

💡Safety

Safety in the context of robotics refers to the design and operation of robots in a way that ensures minimal risk to human users and the environment. The video emphasizes Figure I's commitment to operational safety as a cornerstone of their designs, ensuring that their robots can work alongside humans without compromising safety.

💡Robotics as a Service

Robotics as a Service (RaaS) is a business model where robots are provided to customers on a subscription basis, rather than being sold outright. This model allows smaller operations to utilize humanoid robots without the need for significant upfront capital investment. The video suggests that Figure I is considering this model, which could revolutionize the way businesses adopt robotic technology.

💡Generative AI

Generative AI refers to AI systems that can create new content, such as text, images, or videos, based on learned patterns and data. In the video, Nvidia's generative AI is showcased as a transformative technology that can create intricate and detailed scenes from simple text prompts, significantly reducing the time and resources needed for simulation and data generation in robotics development.

💡Large Language Models

Large Language Models (LLMs) are AI models trained on vast amounts of text data, enabling them to understand and generate human-like language. The video mentions the use of LLMs, such as ChatGPT, in Nvidia Isaac to create detailed scenes rapidly, illustrating the integration of language models in enhancing the interaction between humans and robots and improving the accuracy of AI systems.

Highlights

OpenAI and Microsoft's $500 million investment in Figure, AI, the company behind the Figure One humanoid robot.

Figure One humanoid robot is considered one of the most capable and intelligent bipedal machines globally.

Figure AI has made world record progress in creating a general-purpose humanoid robot in just over a year.

Figure One robot replicates and exceeds human finger movements in dexterity and versatility.

The robot stands at 167cm, balancing size and functionality.

Figure One can handle payloads of up to 20kg, making it valuable for industrial and service applications.

BMW is testing the Figure One humanoid in their production facilities.

The robot is lightweight at 60kg, enhancing agility, maneuverability, and battery life.

Figure One has an impressive runtime of five hours on a single charge.

The robot's top speed is 1.2m/s when walking.

Figure I has assembled an elite 51-member team with employees from tech giants like Boston Dynamics, Tesla, and Apple.

Figure I focuses on five key areas: system hardware, advanced AI, affordability and volume, safety, and robotics as a service model.

The company aims to match the physical capabilities of an average human with its electromechanical humanoid design.

Figure I is working on AI systems for autonomous task completion by humanoid robots.

Operational safety is a cornerstone of Figure One's designs, ensuring safe human-robot collaboration.

Figure I is focusing on real-world industrial applications and constant testing of new designs.

Westwood Dynamics' Themis humanoid robot is an advanced, general-purpose machine capable of various real-world tasks.

Nvidia's dual computer AI model is set to transform robot development and deployment.

Nvidia's Isaac platform integrates generative AI, enabling more efficient and effective robotic development.

China's Vary Toy is a compact large vision language model designed for standard GPUs, addressing the demand for efficient image perception in AI.

Vary Toy's innovative approach scales up the vision vocabulary for large vision language models, offering an accessible solution for researchers.

Transcripts

Openai and Microsoft want to invest $500 million into figure

AI, the company behind the figure one humanoid robot, arguably the most

capable and intelligent bipedal machine in the world.

But what makes this robot so special?

In fact, just over one year from its inception, figure AI has already made

world record progress towards creating this general purpose marvel that not only

replicates human finger movements, but even exceeds them in both dexterity and

versatility. And in terms of size.

The figure one I robot stands at a height of 167cm

to strike the perfect balance between size and functionality.

Even more importantly, the robot is engineered to handle significant payloads of

up to 20kg, making the robot a valuable asset across various

industrial and service applications.

So much so that even BMW is already testing the figure one humanoid in

their own production facilities.

But despite the robot's human like size, it's still relatively lightweight at

just 60kg, enhancing its agility, maneuverability, and

battery life. Boasting an impressive runtime of five hours on a single

charge and being able to sustain continued performance even in demanding

environments. With a top speed of 1.2m/s when walking

impressively. Figure I has assembled an elite 51 member team, with

many of its employees having been brought in from tech giants like Boston Dynamics,

Tesla and Apple, allowing for the robot's record setting development speed.

Furthermore, figure I maintains an edge on its design time by having its

very own in-house prototyping and production facility, which allows the company to

rapidly iterate through hardware implementations as they focus on five key

areas. Number one system hardware.

Figure I aims to create a fully electromechanical humanoid with ultra

dexterous hands and by setting benchmarks in motion range, payload,

torque, and energy efficiency, the company believes it can match the physical

capabilities of an average human.

Number two advanced AI.

The development of an AI system capable of enabling autonomous task completion

by humanoid robots is one of figure's most ambitious goals, with the company

already hard at work creating intelligent agents that can adapt to and navigate

complex real world environments.

Number three affordability and volume.

Aiming to reduce the robot's cost, figure I is focusing on high

volume manufacturing as its strategy to economize the robot and grant

access to a broader consumer market.

Number four safety figure I claims that operational

safety is a cornerstone to their designs, ensuring that the figure one robots

can safely work side by side with people.

But in terms of the future figure, I is taking a pragmatic approach that focuses on

real world industrial applications, with the company constantly testing and verifying

new designs. Additionally, figure I is also considering a

robotics as a service model for its business, a business model that could provide

smaller operations with humanoid robots without requiring hundreds of thousands of

dollars in capital to get started.

And with the humanoid robot market poised to be even bigger than the personal

computer market, both Microsoft and OpenAI are doing their best to

secure a firm position for themselves in the robotics hardware industry, and that's just

the beginning of what's happening with AI robots, as Westwood Robotics has also

recently given the world a sneak peek.

Look at its newest Themis humanoid robot, which is yet another advanced

general purpose machine that's already been seen in the wild on multiple occasions

doing various real world tasks.

While not much is known about Westwood's Themis, it appears to already be

capable of walking across various terrains as well as carry out several other

tasks based on the company's previous open source.

Bruce robot, which costs just over $15,000, has 16

degrees of freedom integrated proprioceptive extremities and liquid cooled

actuators. It's likely that the Themis robot will also integrate

some of these design paradigms into its newest model.

While the Themis robot's release and price are still unknown, it appears as though the

robot will incorporate multiple LCD screens to communicate with humans in

its vicinity. Having already shown multiple video demonstrations of Themis

operating in the wild.

Next, Nvidia has revealed its dual computer AI model that's set to

transform how robots are developed and deployed in this new dual computer

model. The first computer is known as the AI factory, and plays a vital role

in the ongoing development and improvement of AI models, leveraging Nvidia's

data center resources and platforms for both simulation and

training. This aspect of the Isaac platform plays a vital role in

refining the accuracy, performance, and adaptability of the AI that's

responsible for powering various types of autonomous mobile robots.

Next, the second computer of the dual computer model effectively

complements the AI factory as a runtime environment and varies based on

the robot's application ranging.

From cloud based systems to on premises machine processors like Nvidia's

Jetson, which act as an edge device equipped with an array of sensors and

cameras. And the introduction of Nvidia's generative AI is revolutionizing

the field of simulation and synthetic data generation.

By leveraging large language models such as ChatGPT, Nvidia Isaac

enables the creation of intricate and detailed scenes from simple text prompts in

mere minutes. Furthermore, the introduction of Nvidia's text to 3D asset

generation via Picasso even further elevates this capability by

producing new, lifelike assets on demand.

This advancement drastically cuts down the time and resources traditionally needed for

simulation and data generation, paving the way for more efficient and

effective robotic development in the runtime environment.

The integration of large language models and language vision models facilitates a

more natural and intuitive interaction between humans and

robots. In fact, robots equipped with a generative AI model

trained across various modalities exhibit superior accuracy compared to

conventional CNN based computer vision models.

Nvidia's strategic partnerships in this domain exemplify the profound

impact generative AI is having across multiple sectors, reshaping the

way robots operate and interact in diverse environments.

Overall, Nvidia's foray into blending generative AI with robotics through the

Isaac platform is a visionary approach to creating smarter, more adaptable

robots. Finally, China has unveiled Vary Toy,

its pioneering compact large vision language model, also known as an

Llvm, which is designed for standard GPUs.

This breakthrough addresses the growing demand for efficient and effective image

perception in AI, overcoming the challenges posed by existing vision,

vocabulary networks, and the high computational costs

associated with optimizing complex parameters.

The new model emerges as a response to the limitations of popular large

vision language models, which have excelled in combining computer vision

and natural language processing tasks.

These include image captioning, visual question answering,

meme comprehension, and scene optical character recognition.

These successes are largely attributed to advanced vision vocabulary networks

like Clip. However, the true potential of these models is often capped

by the limitations of the vision vocabulary network in encoding visual

signals. Effectively.

Most of all, Very Toy stands out with its innovative approach to scaling up the

vision vocabulary for large vision language models.

It involves training a new visual vocabulary network using a smaller

autoregressive model, such as OP one and 25 M, and integrating it

with the existing vocabulary.

Ferry toys compact size not only makes it a potent tool in the large vision

language model landscape, but also offers an accessible solution for researchers

with limited resources.