Dji dock 3 being used in action for emergency situation.
SHANGHAI/BEIJING, May 13 (Reuters) - In a sprawling warehouse in a Shanghai suburb, dozens of humanoid robots are manoeuvred by their operators to carry out tasks like folding a T-shirt, making a sandwich and opening doors, over and over again. Operating 17 hours a day, the site's goal is to generate reams of data that its owner, Chinese humanoid startup AgiBot, uses to train robots it hopes will become ubiquitous and change the way humans live, work and play.
Reuters spoke to more than a dozen people, including Chinese humanoid manufacturers, investors, customers and analysts, who described how breakthroughs in developing robot "brains" will allow these metallic machines to go from mere spectacles to productive and autodidact workers that could revolutionise the world's pre-eminent manufacturing power. China aims to build its edge by focusing on data training and the , the people said, with some saying the prowess of DeepSeek was a big aid.
Humanoids need to interact with a physical environment and train on datasets focused on tasks, such as stacking boxes or pouring water into a cup.
Last year, Shanghai authorities assisted in setting up AgiBot's data collection site, providing premises rent-free where about 100 robots operated by 200 humans work every day. AgiBot's facility enables it to collect high-quality, targeted data, which it can use to train its embodied AI model, said Yao. Similar sites are being built by governments in Beijing and Shenzhen, according to announcements. Widening deployment of humanoids, especially into factories, is likely to accelerate data collection.
Self driving car's main problem was/is trying to achieve deterministic safty with a stochastic system, which becomes a asymptotic convergence problem.The hype behind these humanoid robots kind of reminds me of the hype behind self-driving cars five to six years ago. The reality is that maybe the technology might have a longer way to go then people realize. Another thing to consider, is that you want to make sure that the people you are replacing with humanoid robots have other job opportunities, otherwise there is going to be a massive backlash against this technology.
Self driving car's main problem was/is trying to achieve deterministic safty with a stochastic system, which becomes a asymptotic convergence problem.
For humanoid robots there are easier ways to achieve determinstic safety, eg. torque limiters, so it should be easier to adopt.
IMO the main problem of humanoid robot is it by definition must be a robust, adaptive and multi purpose tool, since any single application task can already be done with single application robots. So for humanoid robots to make sense it has a far higher capability and intellegence requirement than classic robots, and theres uncertainty in how long it will take for technology to reach those requirements.