Boston Dynamics angrily criticizes Musk, is it bragging or coming true? Three major conjectures about Tesla robots

Speaking of Tesla robots, let’s first take a look at the key timeline.

In fact, the concept of Tesla robot has been born for a year. At the first Artificial Intelligence Day event held on August 19, 2021, Musk It introduced its AI-focused ambitions in detail, and also stated that the actual prototype of Tesla’s humanoid robot, Tesla Bot, will be released in mid-2022. The live PPT live cosplay at that time convinced many netizens that robots would be another feat after Musk committed to exploring Mars.

On June 3, 2022, Musk stated on Twitter that Tesla Bot’s name was changed to Optimus Prime (Optimus), and in 2022 The Artificial Intelligence Day event will be postponed to September 30.

#On June 21, Musk’s latest tweet was: AI is getting better and better. Some brokers interpreted in a public event that Musk is hinting that Tesla AI Day will have shocking new breakthroughs. Musk is very likely to launch Optimus ("Optimus Prime" on September 30). ") prototype.

Although it is old news and new speculation, the market is just switching between high and low, looking for low-level and high-odds sectors. With the help of this news, cars and robots related to that day Stocks surged. More than ten stocks, including New Star, Aolian Electronics, Mahe Holdings, Xinsong Robot, Baiao Intelligence, Paislin, Lanying Equipment, and Greatoo Intelligence, hit their daily limit. Amid the general surge in the concept of robots, some companies have also issued good news to continue to stimulate stock prices.

#After a period of rising prices, some analysts discovered that behind the crazy rise in robot stocks, some brokers were shipping large amounts of money, and pointed out that when the stock price rose, As the Skerbot craze begins to subside, maybe someone will finally be swimming naked.

Don’t let the clouds obscure your eyes. Let’s go back to the most fundamental question. What is the probability of Musk’s robot success from a technical perspective?

Is the hardware all based on deception?

Friends who are familiar with Musk know that every time Tesla’s big move is announced, Tesla’s stock will rise significantly. Many netizens have ridiculed that a project The success of is positively related to how many stocks Musk cashes out in the capital market. Let’s talk about the matter. Whether it’s Musk’s Mars plan, completely driverless cars, or brain-computer interfaces and humanoid robots, they are all projects that cost a lot of money.

Although Musk has successfully introduced concepts such as fully driverless driving and brain-computer interface to ordinary people, every There is often a long way to go between a project and its actual commercialization. In a word: it is difficult for science and technology to be realized through the "Great Leap Forward", and R&D investment and output require a long period of precipitation.

#So what are the difficulties in implementing Tesla’s Optimus? We can analyze it from the hardware and software levels.

There are several well-known targets of humanoid robots around the world. The first one must be Boston Dynamics’ bipedal robot Atlas, followed by Ford. Digit, and then Ubisoft Walker, these companies currently have actual robot products in the market.

According to the information released by Musk, Optimus is 1.72m tall, weighs 57kg, and has a load of 20kg (the arms add 5kg ), with an operating speed of up to 8 km/h. Compared with the 1.5m Atlas, which weighs 75kg, the Digit robot which weighs 42.2kg, is 1.55m tall and has a payload of 15kg, and the Walker which weighs 63kg and is 1.30m tall, the height of a human is undoubtedly a bit awkward.

Because if Optimus is positioned as a service robot as Musk said, this robot may be a bit too similar to a human design, causing discomfort. Nowadays, robotics technology is having a huge impact on manufacturing. Affected by issues such as the uncanny valley, the mainstream voice has always been that there is no need for robots to look like humans. Height may be a very troublesome little problem in the first place.

The weight of 57kg is almost the same as that of an adult. In terms of hardware technical characteristics, because the weight and height mean the lightweight of the components inside, the more compact the product, the more difficult it is to assemble and integrate. The requirements for equipment are also higher.

Increasing energy density and reducing weight ratio have always been unchanging propositions in the machinery industry. It is necessary to insert multiple transmission mechanisms and Energy, and at the same time there is a certain degree of compatibility between various hardware, which requires a lot of high-density debugging. Optimus can achieve similar performance but is lighter than Atlas. Perhaps Musk is really building a car robot "Optimus Prime" that can run for three minutes.

Of course, Boston Dynamics has publicly criticized this unrealistic fantasy: "Boston Dynamics already has a long history in robots and their bipedal movements. After more than ten years of accumulation, Atlas still has errors such as stumbles, and parts have limited commercialization. This shows how ridiculous it is for Musk to think that he can surpass our work in one year."

In terms of external surface materials, Optimus uses "lightweight materials", which Musk believes can be smoother than human skin. But this is contrary to common sense in the robot industry. Due to the high price of full-skin materials, and in order to solve the problems of structural stability and internal heat dissipation, many humanoid robots currently do not use skin and want to cover the whole body with sensors. For skin, a precise mold needs to be made on a stereolithography 3D printer and then filled with silicone. This price will cost more than an airplane.

Most common robot shells on the market are made of metal or plastic materials such as aluminum, and are equipped with different sensors on key actuators such as robot hands to achieve High-resolution, low-cost haptics make it easier to complete some human flexibility tasks.

# Of course, there are also cases where robots adopt full skin.

Previously, for example, China's D&T has demonstrated a bionic robot that covered itself with skin. Its skin was made of bionic silicone, and D&T's self-developed internal technology was used. Intelligent flexible actuator SCA, but this robot Dada does not show much movement characteristics.

Another company that uses all-light material skins is a Japanese company called tmsuk. They have exhibited at CES A super-underworld pediatric training robot has been developed, but this robot does not have much mechanical structure and can only lie down and make fixed movements. It is mainly used for pediatric medical students to practice.

The video of the Ameca robot from the British Engineered Arts company has been widely circulated on the Internet. This gray-skinned robot also has a head. The skin of the face and hands is covered, but in order to achieve the surface characteristics of the skin, the Ameca robot has a total of 45 precision motors on the face alone, and the cost is so high that it will take off.

If Optimus wants to adopt a full-skin robot, it may need greater breakthroughs in materials and heat dissipation Safety and stability are indispensable, and battery and sensor technology also need to break through. Otherwise, it will be very embarrassing if the skin attached to it falls off or cracks when the robot performs some actions.

Some netizens ridiculed that if Musk couldn't do it, he would replace all the outer skin of Optimus with silicone. It would also be a good choice to lie flat and make a high-end inflatable doll. Of course, if Musk announces the direct use of alien technology in liquid metal, it will shock people's attention, and it will be at least ten years more advanced than the entire robot market.

In terms of freedom, Musk said Optimus has hands with human activity levels, its limbs are controlled by 40 electromechanical actuators, and its feet can sense feedback to achieve Balance and agility movements. The number of 40 mechatronic actuators is relatively small, and the solution of bipedal sensor feedback has many applications in exoskeleton robots to cooperate with intention perception, but most of them do not adopt similar solutions in humanoid robot solutions.

For example, my country's Walker robot adjusts its body posture through the compliant control of its legs, or obtains information through cooperation with its own visual perception and force perception. The position and attitude information of external moving objects is used in real time to coordinate the corresponding position and attitude compensation of the moving objects. The communication delay problem of sensor feedback from both feet may make it difficult to achieve the hand-eye synchronization envisioned by Musk.

The software may be feasible

Of course, truth and falsehood are often clever lies, and Musk has publicly Many concepts mentioned also use ambiguous words such as "maybe".

Putting aside the aspects of sensors, batteries and drives, Tesla’s robots really have certain capabilities in terms of the technical solutions of the operating system, especially the computing power part. Realize the possibility.

The car robot concept proposed by Baidu and other companies some time ago has been buzzing. The head of the Optimus robot will be equipped with the same intelligent driving vision and sensors as Tesla cars. At the same time, Optimus has a screen on the face that can display specific information. Since it has a built-in FSD chip and shares the AI ​​system with the car, it means that Optimus should be fully capable of supporting it in terms of algorithm and computing power.

Because effective control of precision movements often requires complex operations, which further affects response speed and flexibility. After all, before Musk said, “It can be said that Tesla is the world’s largest robotics company, and our cars are almost semi-sentient robots with wheels.” This means that the Optimus robot will have enough computing power and data volume in the future. The core problem may be the number and commonality of programmers. After all, programming of both robots and artificial intelligence requires a large number of engineers for training.

Carl Berry, a lecturer in robotics engineering at the University of Central Lancashire in the UK, believes that it is too early for Musk to deploy artificial intelligence on robots. He said: "Frankly speaking, every speech by Musk is usually exaggerated hype. It is feasible to deploy robots with AI in intelligent manufacturing, but the more complex the robots, the more difficult it is to add artificial intelligence." CarlBerry also mentioned , most humanoid robots on the market only display simple body movements. This is not because the hardware cannot complete it, but because it is difficult to achieve random optimization of complex movements without AI, which requires more excellent and stable algorithms.

The universal robot application plan introduced by Musk at the press conference aims to make the Optimus robot a companion robot with diversified applications and forms, eliminating dangers and duplication And boring tasks, such as bending down to pick things up, or going to the store to buy groceries, and these can be completed by Atlas, Digit or Walker. It is not a very novel concept. The core is how to go from a DEMO to a stable robot product. .

Some netizens predict that there is a high probability that Musk’s robot will use the Mesmer system and the Tritium system. , or combine it with several other mature hardware products to create a prototype that can run two steps, and then announce that all the boasts have been realized.

Many world-class companies have expressed caution when it comes to adding robots to AI. Previously, the lecture hall had discussions with relevant persons in charge of companies such as UBTECH in a roundtable event. Regarding whether robots should be added to AI and what level of machine intelligence humans can accept, each company is taking small steps to trial and error. Some companies It is being developed through marginalized application attempts. The implementation of artificial intelligence in robot applications is currently mostly limited to industry and software optimization. Artificial intelligence without extensive training often becomes an obstacle.

But some analysts believe that Tesla’s manufacturing technology innovation ability is very strong. Tesla's dare to die-cast a complete vehicle and launch a fully automatic production line can already be seen that Tesla has been making industry breakthroughs.

Therefore, the core parameters of Tesla robot Optimus were announced last year. It can be predicted that the prototype has been finalized 3-5 years ago, and now it has passed After one year of scenario application and computing power training, considering Tesla’s strong engineering capabilities, the robot’s maturity can be expected.

Maybe car robots are the end game?

The news that Tesla will build robots has been mentioned by Musk from last year to this year. In an interview in April this year, Musk even revealed that Tesla will If Optimus Prime is mass-produced in the next two years, due to the cost reduction effect of scale, its cost will eventually be lower than that of a car. However, Musk did not specify the specific car model at the time. Interview host Anderson estimated at the time that it would be about US$25,000 ( (approximately 159,500 RMB), which led some netizens to judge that perhaps Tesla also built a car robot.

In June, Jidu Automobile, a joint venture between Baidu and Geely, released its first automotive robot ROBO-01. Jidu CEO Xia Yiping announced that the car Deliveries will begin at the end of the year. If Tesla really turns to the concept of car robots, and if Tesla's strong supply chain capabilities are combined, Tesla's car robot pricing is expected to be more affordable in the future, and its promotion is expected to accelerate under Tesla's brand effect.

Therefore, in terms of the truly specific investment direction of Tesla robots, some analysts have given four more down-to-earth dimensions:

• Robot upstream industry chain;
• Tesla automobile upstream and downstream industry chain;
• New material industry chain represented by carbon fiber reinforced resin;
• artificial intelligence technology path industry chain.

But as a public figure, Musk’s attention to any industry will inevitably bring huge traffic and business opportunities to it, and also create opportunities for related manufacturing. business attracts more funding and attention. For example, the famous Sophia robot once frequently appeared on various chat shows and magazine covers. This has aroused the imagination of a large number of people about artificial intelligence, and also made people continue to work in this direction to create more intelligent robots. Products close to reality.

This process and direction of constant exploration into the unknown, just like human curiosity, is the real force for society to move forward. As some netizens mentioned, Musk's robot may not come true, but what it teaches us may not be the ability to do things, but innovation and imagination.

The above is the detailed content of Boston Dynamics angrily criticizes Musk, is it bragging or coming true? Three major conjectures about Tesla robots. For more information, please follow other related articles on the PHP Chinese website!