Reports from the Heart of the Machine
Machine Heart Editorial Department
The starting point for robots to change society may start from the field of mobility.
In 1956, John John McCarthy once proposed at the Dartmouth Conference: "Artificial intelligence is to make machines behave like intelligent behavior exhibited by humans."
This concept was later defined as weak artificial intelligence, in which machines do not have autonomous consciousness, but can actively make decisions and take actions through specific procedures, rather than just implementing simple mechanical actions.
At present, mainstream scientific research is generally concentrated on weak artificial intelligence, and many technology giants have regarded intelligent robots based on this as the commanding heights of future technological development. At the same time, the commercialization of intelligent robots has never stopped, not only the traditional industrial robots that assist or replace human work, but also the new exploration of improving the life experience and combining with more kinds of tools.
The era of "robots" has arrived
In the last month alone, we've seen several impressive robot-related stories. Among them, the humanoid robot "Ameca" from the British company Engineered Arts can show extremely realistic and natural expressions and movements, as if it has a real life.
Researchers at the University of Vermont and Tufts University in the United States used the epidermal cells and cardiomyocytes of frogs to create the world's first living robot "Xenobot", which resembled beans and were able to achieve self-"reproduction".
The quadruped robot, produced by Swiss-Mile, a spin-off company of the ETH Zurich, is not only extremely balanced, but also very fast, and even openly returned to a Tesla Model 3.
It can be seen that the form of robots can vary widely, from satisfying fantasy humanoid bionics to tools or simple scientific research purposes, the results of robots have almost become the ultimate carrier to show the progress of AI technology.
The layout and exploration of technology giants in the field of robotics have never stopped, and the more prominent Amazon and Google are also relatively close in the pace of development.
In 2012, Amazon acquired kiva Systems, a robotics company that originally had a 1,000-pound lifting capacity to move shelves full of goods in a warehouse for $775 million in cash. After the acquisition, Kiva changed its name to Amazon Robotics, and the product line has now covered robots from warehouse picking to terminal delivery.
Unlike Amazon's industrial robot research and development route, Google also has bionic robots. By the end of 2013, Google's parent company Alphabet had acquired eight robotics companies, including Boston Dynamics, Industrial Perception, and Schaf.
Among them, the most well-known is Boston Dynamics, whose products include Spot, a robot dog that can perform tasks in harsh environments, SpotMini, a cheetah robot that can run fast, and Atlas, a humanoid robot that is nearly 1.5 meters tall and can do triple jumps, rotate in situ 360°, turn over, and carry.
It's worth noting that while Alphabet found Boston Dynamics difficult to commercialize, its product concept never lacked followers. For example, Xiaomi released a four-legged bionic robot dog this year, CyberDog, which was called "Iron Egg" by Lei Jun. Although it is still not yet in the mature mass production stage, this also shows Xiaomi's exploration and emphasis on the field of robotics.
It is also a bit surprising that Tesla also released a humanoid robot Bot at this year's AI DAY. Musk said that in the future, it can also be used in Tesla's self-driving cars, which can learn all the operations when driving. But obviously, this is just a concept.
The problem here is that it is more difficult for cars to operate fully autonomously than bionic or industrial robots, not only to achieve more sets of intelligent classes, but also to deal with countless Corner cases. However, if the "robotization" of the car can be realized, it will also mean that the driver can be freed from such a complex job, and even more services can be provided through AI, bringing a revolutionary experience.
And, in addition to unlocking the value that the driver receives, application scale is another key element of the value of the technology. In the domestic market alone, the number of motor vehicles in 2020 has reached 372 million, and the market demand is far from being comparable to other directions.
In fact, there is no essential difference between intelligent cars and robots, the action of the car is the action of the robot, its perception and decision-making system is also the same as the robot, only the actuator of the terminal is different, and the interactive experience of the intelligent cockpit is also one of the directions pursued by many technology companies such as Amazon Astro and Samsung Bot Handy.
At the 2021 World Internet Conference, Baidu unveiled for the first time the concept product "Auto Robot" in the form of a future car, which can serve people through three capabilities: L5 autonomous driving ability, multi-mode interaction ability, and self-learning ability.
At present, the route of automotive robots has gradually gained industry consensus. For example, Apple CEO Cook, who has been working on titan projects for many years, has proposed that "self-driving cars are robots that can achieve a high degree of automation".
Li Deyi, an academician of the Chinese Academy of Engineering and an expert in command automation and artificial intelligence, also said: "The future car will be a 'wheeled robot', not only driving, but also learning, interacting, personality, and understanding."
Cars are being redefined
If you dispel the stereotype that robots are equal to bionic bodies, you will find that car robots are not suddenly proposed. Configurations such as ADAS systems and speech recognition that have been around for a long time can be seen as signs of robotics, but they are not strong enough.
In October last year, Waymo, a self-driving subsidiary of Google, announced that it would open a driverless taxi service without safety officers to the public in Phoenix Phoenix, and achieve unmanned autonomous driving through bicycle intelligence, realizing one of the most important basic functions of car robots.
However, "automatic driving" is only the basic function of car robots, and if it is a Robotaxi that simply recognizes the road and can run, it may not be the final form of "robot".
When the concept of "automotive robots" was first proposed, many people directly associated it with highly automated driving. But the thing to know, automatic driving is only one of the attributes, and the expandable function of the robot is the expansion of the infinite scene after the free hands, and this is also the automotive robot connection technology and scenario proposed by Baidu.
It is not difficult to find that many types of robot research and development ideas point to "humanoid" and "scene-oriented". Because only robots that understand humans better, empathize more, and are better at learning can expand to more scenes, and then truly solve the problems of rigid needs, pain points and high frequencies in travel.
For example, Tesla's Tesla Bot released this year is composed of FSD on-board computers and visual perception systems, as well as 40 micro motors, which are clearly labeled with a series of automotive hardware and software, but they are the way to present humanoids. More than what it can do, we see more from "transportation" to travel partners who know how to accompany and communicate. As Robin Li said, in addition to automatic driving, car robots must become intimate intelligent assistants.
Musk has also stressed on many occasions that Tesla is an artificial intelligence robot company: "Everyone looks at us just building cars, but in fact, we are also the world's largest robot company." Because Tesla's cars are like robots on wheels, we build real robots."
This kind of thinking, Apple's Cook has also expressed. He has publicly stated that the future of smart cars will definitely transform into the nature of robots.
Among the car companies, Jidu Automobile is the first company to determine the mass production of real car robots, planning to be delivered in 2023, and its head of intelligent driving Wang Weibao once said: "Automotive robots we think we must have three main capabilities, free movement, natural communication and self-growth." Among them, free movement refers to L4 level automatic driving, natural communication is the voice ability to achieve natural dialogue with people, and self-growth is based on AI and big data self-learning iterations.
Under this trend, the nature of the car is already beginning to change. Whether luxurious or not, the most important use of the mechanical age car is still a means of transportation. However, in the era of intelligence, because of the emergence of large computing chips and intelligent systems, cars have also put forward more additional needs.
The most core point here is that the car should change from a passive driving tool to an intelligent robot that actively serves, and it must expand to more scenes and complete more work outside of "transportation".
For example, in closed operation scenarios such as ports and mines, there are Yishi Technology, which focuses on factory logistics, and companies with all-field layouts such as Baidu Apollo. The latter has cooperated with Huaneng Yimin Coal and Power Co., Ltd. to launch a special operation robot and a smart mining solution Huaneng Yimin autonomous driving mining card.
In the field of travel, Baidu's forward-looking vision of the car robot released in August this year can also indicate the intelligence level of the future cockpit: it not only has multi-mode interaction capabilities such as voice and face recognition, but also analyzes the potential needs of users and actively provides services. More importantly, it also has the ability to learn itself and continuously upgrade, which can be called a multi-scenario intelligent body.
Industrial demand ushered in the wind outlet
Nowadays, "car robots" have become the consensus of the industry, and technology giants such as Baidu, Waymo, Cruise, and Apple, which are in the forefront, have begun to enter the stage of commercial exploration. Moreover, it is worth noting that because the "automotive robot" has completely changed the product form, it has also spawned new industrial needs: terminal distribution, factory logistics and other unmanned services.
However, mobility is still the main competitive direction for autonomous driving companies. Ark Investment pointed out in the "Big Ideas 2021" report that the current ride-hailing market has reached a global transaction size of $150 billion, and in high-performing cities, the profit margin is as high as 50%.
Such an exaggerated market size and profit margin means that if the autonomous driving company can put the unmanned fleet into commercial operation, it will get a more exaggerated yield. Ark Investment predicts that the enterprise value of autonomous driving platform operators may expand to $3.8 trillion by 2025. At present, Radish Run, Waymo, and Cruise have entered the stage of commercial fee operation of Robotaxi.
In addition to taking the lead in carrying out the commercial charging operation of autonomous driving in Yizhuang, Beijing, Baidu's autonomous driving platform Radish Run has been opened for normal operation in Beijing, Shanghai, Guangzhou, Chongqing, Changsha and Cangzhou. In the third quarter of this year, the order volume of "Radish Run" reached 115,000 orders.
Freight robots are more important areas in recent years, compared to the open road environment, trunk lines, warehouses, field environment is more single, if the automatic vehicle instead of manned driving, will play a greater economic benefit in this high-frequency and fixed application scenario.
Taking trunk logistics as an example, the completion of the driving task of highway sections through the automatic driving system will bring great changes to the logistics industry in terms of safety, operational efficiency and personnel costs. At present, Baidu Apollo, Tucson Future, Wincher Technology, Zhijia Technology and other companies have gradually formed a first-mover advantage, and the commercialization route is also clear.
More representatively, in addition to determining the mass production of the first L3-level smart heavy truck two years later, Baidu ecological company DeepWay also proposed the H2H (Hub to Hub) trunk logistics model, which handed over the driving task of the high-speed section to the deepWay automatic driving tractor front through the transportation scheme of separating the container from the front of the car.
In addition, terminal logistics scenarios such as express delivery and retail, which have long occupied a lot of manpower, are also turning to "robotization". Especially during the COVID-19 epidemic, the demand for low-speed unmanned logistics with life services as the core has been further highlighted. During the epidemic in Guangzhou this summer, Baidu Apollo deployed unmanned vehicles for logistics and distribution with a load of 500kg and unmanned vehicles for cooked food delivery to the front line of the epidemic to provide all-weather and contactless delivery services.
No matter what the shape and function of the robot are different, when it comes to the final application, it is the same way to serve humans. Whether it is companionship, labor, creation, entertainment or assistance, when robots are integrated into real society, the root of their meaning is to further improve human life, and the reason why car robots are valued is the same.
The trend is irreversible, the relationship between robots and humans, and even the whole society will only become closer and closer, and the car robot will be one of the most important roles, and it is also the pioneer who can let consumers experience the functions of high-end robots.
Because the development trend of automotive robots not only represents the demand of the basic travel level, but also reflects the consumer's demand for humanized service functions, and also represents the AI industry's exploration of multi-scenario unmanned services and commercialization models. In the future function of the automobile, perhaps "transportation" and "transportation" are just added values, and their achievable functions and impacts may be far more than everyone imagines.