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

Summary

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

Outlines

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

Mindmap

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

Keywords

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

Highlights

The video is abnormal, and we are working hard to fix it.
Please replace the link and try again.

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.