The next step for AI is clear! The direction predicted by Huang Renxun, Tesla and OpenAI are crazy to catch up

Author: Shiina Editor: Zuri

Source: Silicon Rabbit Race

Image credit: Generated by Unbounded AI tools

People who failed to catch the first AIGC train are most concerned about how the next wave of AI will develop.

Continuing with the big model?

All in AIGC application?

Or return to the logic of "selling shovels"?

Musk, who is jealous of ChatGPT, told us that the number of humanoid robots in the future will far exceed that of electric vehicles;

The largest shovel-selling company in the AI world Invidia CEO** Huang Renxun** pointed out that the next wave of AI is "embodied AI" (Embodied AI);

The top AI company OpenAI also secretly sent out a signal-invested in a "embodied AI" company.

This is an intelligent system that can understand, reason and interact with the physical world. Just like its English name, it has one more "body" than the current AI form. How should we understand the importance of "embodied"? Why, in Huang Renxun's view, AI needs to have a body in the next step? Which start-ups are ready to go on this track where challenges and prospects coexist?

01 tacit understanding between the bosses

The first generation of AI, is a symbolic reasoning system based on rules and logic; it relies on artificially programmed rules and knowledge to solve specific problems, and is more limited in dealing with complex real-world problems. The second generation of AI appeared after the rise of big data and statistical methods. The key to this stage is to input large amounts of data into machine learning algorithms, so that machines can learn patterns and laws from data without explicit programming . This includes traditional machine learning algorithms (such as decision trees, support vector machines, etc.) and neural network techniques. The third generation AI refers to the current and future artificial intelligence technology, emphasizing the comprehensive use of different methods and technologies to solve more complex problems. This includes combining technologies in multiple fields such as machine learning, deep learning, natural language processing, computer vision, and reinforcement learning to improve the capabilities and effects of intelligent systems, of which ChatGPT is a typical representative. Now, Embodied has created a new way for machines to learn autonomously. It can perceive and learn the physical world from a first-person perspective, and understand and perceive things like humans. The learning of the physical world is no less than the learning of virtual information on the Internet. . Jensen Huang sees robots, self-driving cars, and even chatbots made smarter by understanding the physical world as "embodied AI."

The humanoid robot that Musk showed at the Tesla shareholder meeting, Optimus, is also an "embodied AI". Musk's expectation for it is to remember the surrounding environment and easily pick up any object.

Optimus Prime "Optimus"｜Source: ABC

These two giants in the industry, one dominates the underlying chips of AI, and the other dominates terminal equipment. The two bigwigs who understand technology and business have chosen to bet on "embodied AI". AI systems have the potential to bring significant technological and commercial breakthroughs in many fields, and more importantly, to continue the new business growth curve of these two companies. In fact, "embodied AI" has experienced continuous technological breakthroughs and theoretical developments over the past few decades. Early research focused on the perception and motor control capabilities of robots. Advances in sensor and actuator technology have led researchers to design and build robots that can mimic human motion and behavior. These robots are able to perceive the environment through vision, hearing and touch, and perform tasks autonomously. However, due to technical and cost constraints, "embodied AI" has been limited to research laboratories and industry in the past. Until recent years, with the improvement of computing power, the decrease of sensor cost and the advancement of artificial intelligence algorithms, "embodied AI" has ushered in new development opportunities. Startups and research institutions have begun to pay attention to the "embodied AI" Translate into commercial opportunities for practical application. Among them is another existence that can never be bypassed-Boston Dynamics.

02 Existence that cannot be bypassed

To understand the current state of "embodied AI," one must start with Boston Dynamics' past. This "embodied AI" company that came out of MIT was established in 1992, and the earliest funding came from US Defense Advanced Research Projects Agency (DARPA). Later, they continued to launch new robot series, famous for their amazing embodied artificial intelligence technology, whether it is Atlas, a robot that can walk like a human, or Spot, a robot dog that can run and jump at amazing speeds. It has demonstrated unprecedented dynamic stability and maneuverability, and can perform difficult actions and tasks. At the heart of the technology behind these robots is the Boston Dynamics team's mastery of dynamic balance and motion control. In addition to its outstanding performance, Boston Dynamics continues to improve and optimize the robot's perception and decision-making capabilities, enabling it to better understand and respond to complex real-world scenarios.

History of Boston Dynamics Robotics ｜ Source: Synced

However, this top technology company has experienced twists and turns and changed owners many times because of its technology is too advanced and its commercial value is a mystery: In 2013, Google (now Alphabet Inc.) acquired Boston Dynamics. This acquisition provides Boston Dynamics with greater resources and financial support, enabling the company to further expand its R&D and innovation capabilities. However, in 2017, Google transferred Boston Dynamics to Japan's SoftBank Group. As we all know, Softbank founder Sun Zhengyi has always had a soft spot for robots, but the Pepper robot supported by Softbank is not considered a success and has been discontinued. In 2018, Boston Dynamics completed a round of financing worth $250 million, led by Softbank. The purpose of this financing is to further expand the application fields of Boston Dynamics robots and accelerate the commercialization of products. Finally, in 2020, South Korean automaker Hyundai announced it was buying a 20% stake in Boston Dynamics for $1.1 billion. These three “sell-in” also reflect to a certain extent that the “embodied AI” technology has never stopped improving, but whether it can find a foothold to realize commercial value is uncertain.

But this does not prevent a number of emerging companies from pouring in.

03OpenAI also bet

Recently, the tranquility of a small Norwegian town was shattered, and a start-up here has been making headlines.

1X Technologies, the humanoid robotics company, announced the completion of a $23.5 million round of financing led by OpenAI Venture Fund.

The company says their mission is to create robots with practical applications to complement the global workforce. Founded in 2014, the robotics company produces robots that mimic human movements and behaviors. 1X Technologies intends to use the funding to ramp up research and development on their upcoming bipedal robot model NEO and plans to mass-produce their first commercial robot EVE in Norway and North America. 1X Technologies currently has about 60 employees. Because of the participation of OpenAI in the financing, this news immediately attracted widespread attention, and people began to imagine the infinite possibilities of the combination of ChatGPT and robots. Of course, this goal may still be a long way from the actual implementation.

In the past May, an American artificial intelligence start-up company called Figure announced that it had obtained a huge financing led by Parkway Venture Capital - 70 million US dollars. Figure believes that this fund will provide them with Paving the way to create a general-purpose humanoid robot, its products are expected to be launched in the next few months. Although the company is only a year old and its valuation has not been disclosed, people familiar with the matter say it is valued at more than $400 million. Figure’s founder and CEO Brett Adcock has $20 million of his own pocket, and other investors include Aliya Capital and Bold Ventures. Figure's goal is to develop a general-purpose humanoid robot that can adapt to different environments and handle a variety of tasks, whether it's in warehousing or retail.

The company believes that general-purpose humanoid robots have greater potential than single-purpose robots. Deploying them in the labor market can help address labor shortages.

Another robotics company Agility Robotics established in 2015 has a faster commercialization process. The founders Damion Shelton and Jonathan Hurst are both Carnegie Mellon University robotics graduates. The company's products focus on logistics and warehousing robots with walking capabilities, and invented Digit, a humanoid robot with mobility and manipulation capabilities, which has been commercially available since July 2020. Agility Robotics completed a US$150 million Series B round of financing last year, led by Playground Global, a leading deep-tech investment firm DCVC and "Father of Android" Andy Rubin, and participated by Amazon Industrial Innovation Fund .

Robots from Agility Robotics｜Source: ZDNET

In general, as a leading artificial intelligence company, OpenAI's investment in "embodied AI" companies shows that they value and are optimistic about this field. "Embodied AI" has great potential in realizing truly intelligent robots, which can bring a more human-like interactive experience and a wider range of application scenarios.

Coupled with the optimism of technology giants such as Nvidia and Tesla, it will undoubtedly further accelerate the development of this field. The next step of AI, perhaps it is time to hand over the torch of innovation to "embodied AI".