The vice president to 1x technology shared a number of important updates on its human robot.
While the race to create the hymotium robot, 1x is firmly fire in the robot in the house, addressing daily tasks, while others focus on industrial applications.
In places today, we see Neo realize functions as pick up the leaves and put it in a bag. For those who have trees usually edit their leaves as seasons changes, this task is something familiar with. Showing a robot that perform functions like this shows the abilities as their design.
When we see the videos of the hymped robanoid videos, it is important to understand exactly what we see. There are a video where a remote operator is controlling the robot movements, however is not the case there.
As jang highs in their post, the robot works 100% self-employed. I would love to see Walk to the leaves, identify the sheet of leaves, that solves the full bag or more of the pay is able to carry. The man wasted to leaves in a bin, so while the final task may not be solved today, this is a promising sign for the future.
In another place, let’s see the robot that shops one of the most hated tasks at home, stacking the diswasher. When we analyze the complexion of this task, is easily one of the hardest to establish a robot to.
The objects are a dimensions and sets of different and they can only fail in a set of places in the flavapers hunts, also has to treat objects that can be slippery. The specific demo is relatively basic, in whom the dishwasher is already open and waiting for a cup to be loaded.
Because this will be practice in each home, solution must generalize, the solution to all the various dishforce, or go to a phase of learning to your house, with your valuable devices.
What 1x is attempting to prove the objection of the objects, recover the cup from the sink, moving between the hands to reduce the body movement. It also seems to the drawing to allow you to place the fee successfully.
Again, it is important to consider what is not shown in videos to appreciate the progress today and the list of make.
Finally see you a third video where NOO walks with confidence to the couch and put a couch pillow. The task may be a little less common, if those with children and animals, have cusps that returns in their place at the end of the day it would be a big help.
This task is required the robot to figure out what a cuspid is, where the couch is and that the pillow usually in the corner of a couch. The robot resembles put the pillow without toppling above, making the necessary adjustments for counter-balance.
These are promising signals that robot could a day clean your home with the right training. There is also tap and other exacts to treat and when in a shared environment, it is important to the robot is resilient and may be dynamic in receipt. Imagine an animal or child moves in the target position for the pillow, the robot will reply to how much, like to sit on an alternative couch.
Yang shares a thread on X that helps us best understand what we have seen today.
Everything is learned from the data – Contracting network makes the body motion, including the rl body controller direct and the spine books at the same time.
At @ 1x_tech We make humanids for home. A short edge on why we go for the house market, instead of the last.
The house is the “robotic’s final polickey, because it is so radically not borrowed. Every home is different and there are mile of sub-maturations involved in something simply making the laundry (cask underscores, in neighborhood).
Although the house is where truding the tescul latest of different data necessary to create a general intelligence. Several data through several tasks and environments is a required internet ingredient.
Most robotic companies assumes the autonomy must be resolved a task at one time, because autonomy is hard. Solve a task, and then use the money generated by that folder to expand in most valuable jobs.
However, adding capionition when Vior pure chilant – Although in arm to a new theater every quarter, it will only be 20 skills in 5 years. Home acquisition asks thousands of ability, not a little.
In autonomous cars, a few autonomous truck on the highway would be easier than urban guide because the problem does not require general intelligence.
The paradox, doing a narrow work in moderately unstructured situations it is in order to make it more difficult than to make a general you are in situations very non-structures. The moderately unstructured environments are more general than you think.
You need to climb different data in order for something to work, even on tight problems. However, it is hard to collect failures in the wild for something as a high pal. As moving a truck. And the “structured environment” you truly work against rare events are also rarr then it is difficult for my data that helps make mad.
That’s why the Tesla Adpilot Adpilot – Trouble-trained – great works, while there is not yet a single solution autonomously autonomous today.
The same is true for humanoid robots. One could think that fix factory work that goes to the auto at home, but the tasks that are always discovered a physical understanding that is stringed
Chatgpt showed the world that autonomy should not be solved a task at the moment. Is an incredible business case study in the consumer distribution strategy for ai. Solve a wide-set of computers for consumer users, and after the adoption of the business quickly follows.