The human robot takes sound orders to help all over the house

Founder of Figure and CEO Brett Adikok Thursday open A new model for automatic learning for human robots. The news, which arrives two weeks after AdCock The decision to move away from Openai’s cooperationIt is concentrated on Helix, a “general expert” model-VLA.

Vlas is a new phenomenon for robots, and the use of vision and linguistic orders to process information. Currently, the most famous example of the category is Google DeepMind’s RT-2Which is trained by robots through a group of video and large language models (LLMS).

Helix works in a similar way, combining visual data and language demands to control a robot in actual time. The shape writes: “IUD displays a strong circular to the organisms, the ability to capture thousands of new home tools with shapes, sizes, colors and the characteristics of changing materials that have not been previously faced in training, simply by asking in the natural language.”

In an ideal world, you can simply tell a robot by doing something and will do so. This is where the snail comes, according to the shape. The platform is designed to bridge the gap between vision and language processing. After receiving a natural language sound wave, the robot visually evaluates its environment and then does the task.

Figure provides examples such as, “Delivery of a cookie bag to the robot to your right” or, “Receive a cookie bag from robot to your left and put them in the open stairs.” Each of these examples involve a pair of robots that work together. This is because Helix is ​​designed to control a robotin simultaneously, with one other help to perform various home tasks.

The VLM shape is shown by highlighting the work that the company does with 42 Humanoid Robot in the home environment. The houses are very difficult for robots, given that they lack the structure and consistency of warehouses and factories.

The difficulty of learning and controlling are major obstacles that stand between complex and home robot systems. These problems, along with price signs from five to six numbers, are the reason that the home robot has never preceded for most human robotics companies. In general, the approach is to build robots for industrial customers, improve reliability and reduce costs before housing processing. Homework is a conversation for a few years from now.

At techcrunch A tour of the offices of the Gulf region In 2024, ADCOCK showed some of the lines that the company was putting at home. At that time it seemed that the priority was not given to work, as the number focuses on pilots in the workplace with companies like BMW.

With the announcement of the IUD on Thursday, the number shows that the house should be a priority in itself. It is a difficult and complex preparation for testing these types of training models. Teaching robots to carry out complex tasks in the kitchen – for example – open them to a wide range of procedures in different settings.

“In order for robots useful in captivity, you will need to be able to generate new smart behaviors upon request, especially for things that have not been unprecedented before.” “Teaching robots even one new behavior currently requires a great human effort: either hours of handicrafts experts at the doctoral level or thousands of demonstrations.”

Manual programming will not expand home. There is simply a lot of unknown. Kitchens, living rooms and bathrooms vary greatly from one to another. The same can be said to the tools used in cooking and cleaning. Besides, people leave chaos, rearrange the furniture, prefer a range of different environmental lighting. This method takes a lot of time and money – despite the number Certainly he has a lot of the latter.

The other option is training – and a lot of it. Automated weapons trained to choose and put objects in laboratories often use this method. What you do not see is that hundreds of hours of repetition takes it to make a strong illustration enough to take over to take very variable tasks. To capture something properly the first time, the robot needs to do so hundreds of times in the past.

Like many human robots surrounding the current time, working on Helix is ​​still at a very early stage. Viewers should be advised that a lot of work behind the scenes to create the well -produced short videos that were seen in this post. Today’s announcement, in essence, is an employment tool designed to bring more engineers to help develop the project.