news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

"the sensors inside, including some ranging sensors and lidar sensors, are the same on both sides, and some parts suppliers may also be the same." recently, an engineer from zhumi technology told reporters about the sweeping machine and the car. the similarities between them.

the wave of artificial intelligence has brought sweeping robots and self-driving cars in the same development direction. they may be equipped with the same chips with artificial intelligence computing power, and be equipped with large models developed by the same supplier but with different parameters. they follow similar technical paths and even share the same supply chain system. during this year's yunqi conference, alibaba group ceo wu yongming said that in the future, all movable objects will become intelligent robots.

in the era of artificial intelligence, a sweeper and a car have much more in common than they appear on the outside.

the ideal car equipped with lidar

they can all be robots

artificial intelligence, especially large models, has developed rapidly in the past two years. artificial intelligence is beginning to understand all aspects of the real world, from molecular motion to human running, from using tools to painting and composing music.

the outbreak of artificial intelligence may be the prelude to an industrial revolution, and it happens at a transition moment in the automobile industry.

in the chinese market, nearly half of the transition from fuel vehicles to new energy vehicles has been completed. the latest monthly statistics show that 2.453 million cars were sold in the chinese market in august, of which the new energy penetration rate was 44.8%, and the new energy penetration rate in the passenger car market exceeded 50%. the electrification wave of automobiles paves the way for intelligence. when the engine is replaced by the battery and the chip is taking over the car, the second half of intelligence has also begun.

"the automotive industry is undergoing such changes. previous self-driving technology relied on people to write algorithm rules. with hundreds of thousands of lines of code, it still cannot cover all driving scenarios." at this year's yunqi conference, wu yongming said that using after "end-to-end" large-model technology training, the ai ​​model learns massive amounts of human driving visual data, giving the car a driving ability that surpasses that of most drivers.

he believes that robots will be the next industry to undergo great changes. in the future, all moving objects will become intelligent robots. it can be a robotic arm in a factory or a crane on a construction site. in the future, there will be many robots in factories, producing robots under the command of large ai models. in the future, there may be two or three robots in each home to help people improve the efficiency of their lives.

in the era of ai blessing, more and more technology companies and entrepreneurs are beginning to view the entire physical world from the perspective of robots.

"you can think of tesla cars as robots on wheels, and anything connected to the internet is a terminal of artificial intelligence." musk said at a recent forum that the development speed of artificial intelligence is faster than he has ever seen. any technology is faster.

the focus is on the universality of the technology. "we can take the same technology and put it into a humanoid robot to make it more useful." this is why tesla is also developing the bipedal robot optimus prime. if humans can successfully transform robots, it will completely change the production methods of global manufacturing.

many car manufacturing processes and core components can also be seen in household sweeping robots.

"because it is the chassis movement of the whole machine, how to adjust the center of gravity, how to match the center of gravity when overcoming obstacles, how to make the impact smaller when coming down, including the chassis structure, actually refer to the design of the car." an engineer from zhuimi technology introduced the new generation product stated that this product has stretchable bionic mechanical feet and took nearly two years of research and development.

american companies invented the sweeping robot, and chinese companies have developed it. according to idc data, in the second quarter of this year, chinese-funded companies accounted for nine of the top ten sweeping robot companies in the world. roborock technology, ecovacs, and zhuimi technology are all in the top ten.

these sweeping robots basically realize semi-automatic operation in consumers' homes. the ground conditions are understood through sensing devices such as cameras and lidar, and built-in algorithms realize automatic navigation and other functions. now the leading sweeping robot companies are looking for large models to be built into the machines.

"anything powered by a computer is really a robot," musk said.

chasing how sweepers sense roadblocks

interoperable supply chain

sweepers and cars have many similarities in their core components.

the low-end versions of the hardware and algorithms used in 400,000-yuan cars can be migrated to 4,000-yuan sweepers and humanoid robots.

"the first one is the sensors inside, including some ranging sensors and lidar sensors. the two sides are connected, and the parts suppliers may also be connected. the second one is the processing of maps and routes, path planning, and driving algorithms. it's the same." an engineer from zhuimi technology told china business news. the third is that large models are being embedded in sweepers and smart cars.

of course, the sensing equipment and chip processing capabilities used in sweepers and cars are different.

vehicle-level driving speeds are high-speed or medium-speed driving, and the sensor's long-distance and accuracy response will be higher, and the corresponding cost will be higher. however, the sweeping robot belongs to indoor semi-autonomous driving, so in terms of sensor selection category, it will be a more cost-effective low-speed sensor.

these sensing devices with different performance may come from the same supplier.

for example, listed companies sagitar juchuang and hesai technology specialize in the production of lidar. their lidar is widely used in xpeng motors, huawei cars, etc., and can also be used in a series of sweepers and industrial equipment. the principles of these lidars are the same, but the two industries have different selections of sensor accuracy.

compared with sweepers used in households, commercial cleaning robots have higher sensor requirements. commercial cleaning robots can weigh up to about 1,000 kilograms and have high speeds. the sensing devices they require are closer to car-grade sensing devices.

the similarities between sweepers and cars are also obvious in terms of intelligence. they are all catching up with the wave of intelligence, embedding large models, and may even cooperate with the same large model company.

take zhuimi technology and nissan motor as examples. they are cleaning appliances and vehicle companies respectively.

the sweeping machine x40 pro ultra released by chuimi in may this year is equipped with dreamegpt jointly developed with baidu qianfan wenxin large model. this is a large model based on baidu's large model and fine-tuned for specific usage scenarios. it allows the sweeper to better understand the user's voice intentions and interact with the user more smoothly.

products launched by nissan in the chinese market will also be equipped with baidu’s generative ai technology. during the recent "enjoy e-park" future travel carnival, li jun, senior deputy general manager of nissan (china) investment co., ltd., told reporters that the implementation of this technology will accelerate the pace of nissan's intelligence.

nissan and zhumi are representative cases in the industry. the two industries, sweeping machines and automobiles, are rapidly moving toward intelligence. in addition to zhumi technology, cloud whale has also launched its own nargpt multi-modal large model, and ecovacs' large model algorithm has also passed national registration; chinese automobile companies such as xpeng motors and nio are also developing large models to promote autonomous driving. drive.

even cars and sweepers face similar competitive environments in the chinese market.

chinese clean appliance companies such as ecovacs and zhumi technology are facing continuous price wars, accompanied by rapid innovation and industrial upgrading. automobile companies in the chinese market have not only undergone harsh transformation, but also been trapped in a price war. nissan and others will need to respond quickly to stay competitive. it previously launched "the arc nissan electric arc plan", aiming to accelerate the transformation to intelligence and electric drive. nissan plans to renew 73% of its product lineup by fiscal year 2026 and launch 8 new energy vehicles.

what are china’s advantages?

the core hardware supporting intelligence is the chip.

smart cars and sweepers also have a lot in common in terms of core chips.

for example, horizon robotics started by developing car-grade chips, but it also has a new brand of sweet potato robots that are used in a wider range of applications such as sweepers, lawn mowers, family companions, hotel business services, robotic arms, wheels, and even humanoid robots. chip for robotics scene. both chips can support the implementation of large models, and can be used in both a 400,000-yuan ideal car and a 4,000-yuan sweeper.

industries that can scale up quickly are particularly attractive to the sweet potato robots that are pursuing commercialization. sweepers and lawn mowers, as the most mature consumer-grade robot scenarios, are an important source of millions of sweet potato robots shipped every year.

"after all, we are making chips, so we will pay close attention to areas with large quantities." wang cong, ceo of digua robot, said in a recent interview with the media, "many leading companies in the fields of sweepers, lawn mowers, and family companions have relationships with us. very deep collaboration.”

nvidia is also a clear case. the gpu it develops is currently the most efficient tool for large model training and inference. its products can not only be implemented in self-driving cars, but can also be widely used in sweepers, bipedal robots and other fields. for example, xpeng motors, nio, etc. are equipped with nvidia's smart driving chips, and many bipedal robot companies also rely on chips provided by nvidia.

the investment in chip development is huge, and large-scale applications are of great significance to companies such as horizon, nvidia and intel to expand revenue and share costs.

taking horizon as an example, although automotive solutions are its main source of revenue, non-automotive solutions will bring it more than 80 million in revenue in 2023. horizon's overall revenue is growing rapidly, but it still lost 1.6 billion yuan in 2023. the huge investment in research and development is an important reason. in 2023, it will invest 2.3 billion yuan in research and development expenses.

the investment for large models is also huge. the investment in large parameters, large computing power and big data consumes openai more than 1 billion us dollars every year. it is also currently preparing a new round of financing worth billions of dollars. domestic technology companies such as baidu, alibaba and toutiao are all transforming in the direction of artificial intelligence and have made huge investments to develop their own large models such as wen xin yi yan and tong yi qian wen. they all hope that these models can be applied to a wider range of application scenarios, covering cars, sweepers, four-legged robots, etc.

in terms of high-end chips and top-end large models, american companies are at the forefront. the advantage of chinese companies lies in their exposure to a wider range of application scenarios and the resulting economies of scale.

for example, the united states will sell 15.58 million cars in 2023, but china will sell 30.16 million cars, nearly twice that of the united states. moreover, nearly 30% of the 30.16 million cars sold in china are new energy vehicles, which opens up more possibilities for the development of intelligence.

let’s look at industrial robots. china's installed capacity of industrial robots ranks first in the world, and more than half of the world's industrial robots are installed in chinese factories. in terms of household sweeping robots, china’s annual sales exceed 4 million units, while china’s local production capacity is close to 15 million units.

"when it comes to hardware products like robots, in china you can find the entire supply chain of robots. there are a lot of companies in china, especially in shenzhen. we draw a picture in the morning and send it to the factory to start working on it in the afternoon. this is where china is uniquely endowed.”

these large-scale sweepers, automobiles, and industrial robot industries are all expected to be equipped with artificial intelligence chips and large models with different parameters in the wave of artificial intelligence. with the iteration of technology and the achievement of economies of scale, some originally high-precision products have become increasingly mature and widely used, and their r&d and manufacturing costs have become increasingly thin. chips that originally cost 400,000 yuan for cars can now be used on 4,000 yuan for sweepers.

such an evolutionary trend can only occur in a small number of economies.