ChinaChina
HuxiuFEATURE · TRANSLATED

Translated from Chinese · 1/21/1970 · 17 min read · 降噪NoNoise

Original: 谁在赚机器人的“学费”? · https://www.huxiu.com/article/4887809.html

Who Is Earning the "Tuition" of Robots?

Who's Earning the "Tuition" of Robots? After doing household chores for decades, Gao Bo, a full-time mother in her 50s, has discovered for the first time that her movements can also be sold for money.

Who's earning "tuition" from robots? She lives in Shandong and normally has to take care of her teenage son, making it difficult for her to work outside the home. Now, Gao Bo records her actions with her phone every day while cooking, washing clothes, and cleaning. She told the media that she can earn 120 yuan for six hours of work per day.

Who's paying the "tuition" for robots? These first-person videos, after being processed, will become training materials for robots to learn household chores.

Who's paying the "tuition" for robots? Gao Bo is not an isolated case. On social media, similar scenes are becoming increasingly common: in a McDonald's, a girl wears a collecting device to clean tables; in a large restaurant, a chef records the angle and movement trajectory of his wrist while stir-frying dishes...

Who's earning the "tuition" for robots? Embodied data is entering the era of "crowdsourcing". In the past, robot data was mainly produced by collectors in training centers through remote control of real machines. Today, with the popularity of first-person perspective cameras and other non-entity collection devices, data collection workstations are no longer limited to training centers. Any place where people work can become a classroom for robots.

Who is paying the "tuition" for robots? This wave of crowdfunding fervor is backed by a huge data gap in the robotics industry. Robots need to "supplement their brains" to work in the physical world, but the current situation is that there is a severe lack of data for understanding the physical world.

Who's Earning the "Tuition" for Robots? A supply chain surrounding robot data collection and training has rapidly emerged, with enthusiasm from the primary market following closely.

Who's Earning the "Tuition" for Robots? But behind the excitement, the math is not easy to do. Who is making money first in this industry chain? How much of the continuously expanding data capabilities can actually be turned into a robot's capabilities and customer orders?

Who's Profiting From Robot 'Tuition'? The Entire Industry Is Racing to Teach Robots

Who's profiting from robots' "tuition"? According to Interact Analysis data, as of the end of April 2026, a total of 64 data collection and training centers have been put into operation nationwide, with at least 90 centers either under construction or in the planning stages, among which at least 13 centers have deployed over 100 robots.

Who is earning the "tuition" of robots? ▲Interact Analysis's "Inventory of 90+ Domestic Humanoid Robot Data Collection Centers"

Who's paying the "tuition" for robots? In an area of several thousand square meters, homes, supermarkets, warehouses, and factories are replicated on a 1:1 scale. Dozens or even hundreds of robots line up for "classes", where pickers wear VR headsets and manipulate mechanical arms to repeatedly move boxes, sort goods, and operate tools.

Who is making a "tuition" for robots? On recruitment websites, embodied intelligent data collectors have also started to emerge in batches, with some positions not requiring work experience.

Who is earning the "tuition" of robots? A data collection service provider, which previously served internet clients such as Baidu Maps, added a new team of over 20 people at the beginning of the year to focus on binocular data collection for general-purpose applications. According to the person in charge, two leading robot manufacturers and one elderly care real estate company have taken the initiative to cooperate, "The demand is still relatively large, and we have caught the trend."

Who is profiting from the "tuition" of robots? This wave of enthusiasm stems from a shift in the development focus of the robotics industry. As advancements continue in areas such as entity, joint, and whole-body control, robots are now capable of running, dancing, and boxing, prompting the industry to devote more energy to models and practical tasks.

Who's paying the "tuition" for robots? A new bottleneck has emerged: robots can repeatedly complete an action in a training environment, but their success rate may plummet when the object is changed, the placement angle is adjusted, or they enter a room with different lighting and environment, meaning their generalization ability is relatively weak.

Who is paying the "tuition" for robots? The robotics industry is therefore attempting to replicate the Scaling Law of large AI models: when the model architecture and training methods are relatively stable, it exchanges increased data, parameters, and computing power for enhanced capabilities.

Who is profiting from the "tuition" of robots? This rule is still in the verification stage for robots, but it has already changed the industry's expectations for data scale.

Who is paying the "tuition" for robots? According to industry estimates, the entire industry has accumulated around 500,000 hours of high-quality training data, while to achieve intelligent emergence, 1 billion hours of training data may not even be enough. By this scale, the difference between the two is at least around 200 times.

Who is paying the "tuition" for robots? After demand increases, the industry still needs to solve the problem of where the data comes from. Traditional real-machine collection is difficult to fill this gap. Insiders revealed that a trainer can usually only remotely control one robot, and one robot often requires two people to cooperate, completing only a few dozen real-machine collections per day, because the robot's "work" speed is much slower than that of humans.

Who is profiting from the "tuition" of robots? In 2026, the concentrated heating of non-entity collection has lowered the threshold for expanding data production capacity.

Who is profiting from the "tuition" of robots? This route emerged around 2024 and has begun to transition from papers and prototypes to complete products this year. For example, the EGO device (note: first-person perspective data device) worn on the head records body movements, while the UMI (note: universal operation interface) held in the hand records hand movements, rotations, and openings and closings; wrist cameras and tactile devices supplement close-up images and contact information.

Who is earning the "tuition" for robots? Technology has opened up supply, and policies have further amplified construction demand. In June 2026, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission of the State Council launched a special action for realistic training, requiring 10 provinces and cities to each select no fewer than 20 key scenarios, with a plan to form over 100 high-value application scenarios by the end of the year, driving the landing of capabilities on a scale of tens of thousands of units.

Who's Profiting From the "Tuition" of Robots? Multiple forces have combined, and the robotics industry has quickly become crowded with companies from different backgrounds.

Who is earning the "tuition" for robots? Companies like OriBit and TuJian Technology are the most similar to "selling shovels". OriBit started with 3D vision, extending its 3D camera, calibration, and large-scale manufacturing capabilities to EGO, UMI, and wrist cameras. TuJian, on the other hand, has entered the field with flexible electronic skin, using tactile gloves to record changes in contact force and stress that are difficult to capture on video.

Who is earning the "tuition" of robots? ▲On the official website of OriObot, a video demonstrates a person wearing an EGO device to collect data.

Who is earning the "tuition" for robots? Companies like Ubitus and Zhihuang Technology, which focus on ontology and models, are also entering the data collection phase to secure equipment orders and training data. When local governments build data collection centers, they often bundle the procurement of robots, remote operation equipment, and training systems. Ubitus has won two projects in Huizhou and Hohhot, with a total amount exceeding 130 million yuan. Zhihuang Technology is simultaneously developing models, ontology, and data collection tools, which can determine what data to collect next based on model performance, thereby shortening the cycle of data collection, training, and testing.

Who is earning the "tuition" for robots? Zhiyuan-incubated Miqu Tech is attempting to turn its data capabilities, originally designed for its own robots, into an independent platform, providing external services for data collection, governance, training, and evaluation, and acting as an organizer for the entire data production chain.

Who is earning the "tuition" of robots? Jingdong and Xian Intelligent have existing advantages in scenarios. Jingdong has logistics, retail scenarios, and organizational capabilities, allowing it to extend its reach into warehouses, shopping malls, factories, and ordinary households. Xian Intelligent already has a large number of robots operating in warehouses and factories, hoping to generate data while these devices work. The company disclosed that its robot brain installation exceeds 50,000 units, accumulating over 500,000 hours of multi-format real-machine data.

Who is earning the "tuition fees" for robots? The entire industry is accelerating the process of "compiling textbooks" for robots, but having more "textbooks" does not necessarily mean the robots can learn.

Who is earning the "tuition fees" of robots? What kind of "textbooks" do robots need?

Who is earning the "tuition" of robots? In June this year, robot company XDOF shared an experiment.

Who's Paying the "Tuition" for Robots? A team preparing to teach robots to fold T-shirts used a common imitation learning method for training, and after testing 20 times, the robot succeeded 20 times. Later, the team continued to add more completed task demonstrations, thinking that the increased data would improve the model's performance, but the success rate first dropped to 2 times, and eventually resulted in 20 complete failures.

Who's paying the "tuition" for robots? The problem lies in those seemingly qualified videos where the clothes are eventually folded, but the process is interspersed with pauses, hesitations, re-grasping, and invalid adjustments. Humans watch it once and think it's just normal operation; robots, however, will take in the entire process, including the hesitations, as the standard answer.

Who is paying the "tuition" for robots? This feeling is similar to that of domestic practitioners. An industry insider told NoNoise that starting this year, the demand for data collection has expanded rapidly, but it was often discovered that the collected data was of little use after it was returned.

Who is earning the "tuition" for robots? The industry's true scarcity is gradually becoming the ability to judge what robots should learn.

Who's Earning the "Tuition" for Robots? To turn human movements into robotic capabilities, the first step is to determine what to capture. Picking up boxes requires stability and rhythm, screwing in screws demands precision and force control, while folding clothes involves handling deformation, occlusion, and multi-step operations. With different task definitions, the requirements for collecting camera images, motion trajectories, tactile sensations, and joint states also change.

Who is paying the "tuition" for robots? These contents can be converted into hours on the "production schedule", but after entering the model, they play completely different roles.

Who's paying the "tuition" for robots? After data collection is completed, it undergoes the first round of inspection: whether the task was completed, if the image was obstructed, if the sensor was disconnected, if the image and control commands can be synchronized, and if any human faces, mobile phone screens, or trade secrets were captured.

Who is earning the "tuition" of robots? A data service provider claims that currently, they can achieve at least one valid piece of data out of every three or four collected, which is relatively efficient in the industry.

Who's Profiting From Robots' "Tuition"? After being cleaned and labeled, the data still has to go through an even more troublesome hurdle: the same batch of data may yield completely different verification results when handled by different companies.

Who's paying the "tuition" for robots? Embodiment data is deeply tied to the robot's body and model architecture. Six-axis robotic arms have different degrees of freedom than seven-axis arms, and two-finger grippers have different motion spaces than five-finger dexterous hands. Camera positions, sensor configurations, coordinate systems, and control frequencies also vary. The same trajectory that can be directly trained at one company may need to be remapped or even become completely unusable when switching to a different robot or model.

Who is earning the "tuition" of robots? So, embodied data has at least three layers of effective standards.

Who is earning the "tuition" for robots? The first layer involves collecting effective data, confirming task completion and file availability; the second layer involves ensuring the validity of the dataset, confirming that the data has been cleaned, annotated, and aligned, and can enter the client's training pipeline; the third layer is where the model becomes effective, and after adding this batch of data, whether the robot's success rate, generalization ability, and failure recovery are truly improved.

Who is paying the "tuition" for robots? Currently, these three layers of "effectiveness" are easily confused. Vendors deliver 1000 hours of effective data, which may only complete the first two layers; however, customers expect the robot's capabilities to increase accordingly.

Who is earning the "tuition" for robots? This also determines that embodied data is temporarily difficult to trade like ordinary commodities. It is hard for clients to simply place an order for 1,000 hours of data. A more common demand is: a certain robot has a success rate of only 60% in performing specific tasks, and the supplier needs to determine which scenarios and actions are missing, and then design a plan for data collection, cleaning, training, and supplemental collection.

Who is paying the "tuition" for robots? Validity is thus hard to become a fixed attribute of data itself, it depends on whether the data can match with a certain model, a certain embodiment, and a certain task.

Who is paying the "tuition" for robots? Yu Shu founder Wang Xingxing has pointed out that for robots, every input and output may produce deviations and losses, which is an important reason why the generalization ability and task success rate of current robot models are still insufficient.

Who is paying the "tuition" for robots to learn? This means: from capturing a video of an action to a robot actually learning it, there is a data loss chain in between.

Who is paying the "tuition" for robots? This has resulted in a mismatch in the industry chain: the costs have already been paid upfront, while the effectiveness of the final training remains uncertain. There is no formula for robots to automatically become intelligent after being fed for a certain number of hours.

Who's paying the "tuition" for robots? As the output of data cannot be directly converted into capability gains, a dilemma arises in the data collection business: suppliers calculate costs based on equipment, labor, and collection time, while clients are only willing to pay for model effectiveness. The intermediate loss must be borne by someone, which has become an account that the data collection industry must settle.

Who is paying the "tuition" for robots? A survival question for the data acquisition industry

Who is earning the "tuition" for robots? Looking along the industry chain, the first to "cash in" is still the money from laying the foundation. Cameras, gloves, clamps, remote operation systems, and the robot bodies themselves can be settled on a per-unit or per-set basis, while training venues can be accepted based on area, number of devices, and construction period.

Who is earning the "tuition" of robots? Once it comes to the data trading link, uncertainty surges.

Who's Paying the "Tuition" for Robots? Multiple insiders in the embodied intelligence industry revealed to NoNoise that some local governments are willing to pay for robots and build training facilities, but during project negotiations, they require robot companies to buy back the data.

Who is paying the "tuition" for robots? According to the "Embodied Intelligence Training Ground Research Report (2026)" jointly released by the China Academy of Information and Communications Technology's Artificial Intelligence Institute, training grounds require heavy asset investment. While selling data products can generate revenue the fastest, relying solely on data sales is difficult to cover the heavy asset investment, and the payback period is relatively long.

Who's Earning the "Tuition" for Robots? According to publicly available estimates, a professional remote operator working for 8 hours can only produce 2 to 3 hours of effective data on average. The price for domestic real-machine data is approximately 500 yuan to 1,000 yuan per hour, while some data collectors who require a technical background or on-site presence can earn a monthly salary of 8,000 yuan to 15,000 yuan.

Who is paying the "tuition" for robots? ▲Data collection positions on recruitment platforms

Who is profiting from the "tuition" of robots? Requiring data buyback is equivalent to adding an "insurance policy" to the already high investment costs.

Who is profiting from the "tuition" of robots? However, manufacturers of robot bodies and modeling companies also have their own struggles —

Who's paying the "tuition" for robots? As the cost of Ego and binocular devices continues to drop, general tasks such as cleaning, storage, and sorting are the easiest to mass-produce and repeat. Different collectors wiping tables in different rooms do increase the duration; however, if the objects, actions, and environments change are limited, the new capabilities added to the model may not be as numerous as the new dataset appears to be.

Who's paying the "tuition" for robots? Figure, a US humanoid robot unicorn, once tried to procure data from external suppliers, but found it difficult to meet the requirements of its own models in terms of data scale, diversity, and quality. As a result, it turned to building its own data collection system - the Index platform, which collects data through crowdsourcing, soliciting videos of daily chores or work from ordinary users around the world. To date, Figure has paid creators $15 million.

Who is paying the "tuition" for robots? Real-machine data is also hard to sell. It is more closely tied to target tasks and the robot's body, with slower production speeds and higher costs, resulting in fewer customers who can make a purchase. An industry insider told NoNoise that when model companies reach the procurement stage, they often opt for lower-cost simulated data first. Only when tasks are close to actual delivery, such as moving boxes or assembly, will clients supplement with targeted real-machine data.

Who is paying the "tuition" for robots? These problems are forcing the industry to change its approach to production. What the industry needs to reduce is the overall cost, from the cost of collecting data per hour to the total cost of a robot learning a skill.

Who is paying the "tuition" for robots? One approach is to reserve the expensive physical machines for where they are needed most. For example, using body-less data collection devices to expand the scale of human demonstrations, using simulations to batch manufacture environmental changes and failure scenarios, and then having the physical machines complete task adaptation and final calibration with concentrated data. Nvidia once used a small amount of human demonstrations to generate 780,000 synthetic trajectories, equivalent to 6,500 hours of manual demonstration data, with the entire generation process taking only 11 hours.

Who's paying the "tuition" for robots? Another approach is to reduce the cost of repeated adaptation for different clients. Google has partnered with 33 labs to launch the Open X-Embodiment dataset, which brings together 22 types of robots, over 500 skills, and more than 1 million task trajectories, in an attempt to unify action representations and data formats.

. " tops of a " tops of a " top of a " top of a " " " " " " " " " " " " " " " " " " " " " " " " " ", " " " " " " "." " " " " " " " " " " " " " "." " " " " " " " " " " " " " " ". " " " " " " "." " no of " " " " " " " " " " " " " " " " " " " " ". " " " " " " " " " " " " " " " " " " " ". " " " " " " " " our " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " ". " " " " " " " " a " " " " " " " " " " " " " " " " " " " " " " " " ". " " " " " " " " a " " " " " " " " " " " " " " " "

Who is earning the "tuition" for robots? These attempts cannot eliminate the physical differences of robots, but can first unify the data recording and reading methods, reducing the workload of service providers in defining formats from scratch, and making it easier for customers to determine whether a batch of data can be imported into their own training pipeline.

Who is earning the "tuition" of robots? The domestic standard construction has also been launched. Since March 2026, several somatic intelligent data and training specifications have been released or initiated.

Who is profiting from the "tuition" of robots? From an industry perspective, the data collection industry, as part of the data infrastructure for "upgrading the intelligence" of robots, is on the verge of a breakthrough. In the future, once the technical direction of basic models converges and the industry reaches a consensus on data standards, the value of high-quality data may be revalued.

Who is earning the "tuition fees" of robots? Source:

Who is earning the "tuition" for robots? [1] "China is training a robot future—one folded shirt at a time", Rest of World, June 2026

Who is earning the "tuition" for robots? [2] "Lowering the Noise Floor on Robot Data", XDOF, June 2026

Who's Earning the "Tuition" for Robots? [3] "Inside the Digital Harvest Factory: In-Depth Discussion on the Desert of Robot Data, the Four-Layer Pyramid, and the Tree Planter", Silicon Valley 101, May 2026

Who's Paying the "Tuition" for Robots? [4] "Investigation into the Industrial Chain of Embodied Data Collection: Humans Collected by Robots", Jiazi Lightyear, June 2026

← Back to Latest