OpenAI's robotics derivate 1X tests voice interface in service robots
1X controls its service robot Eve by voice. The robot can tidy up an office environment on request, for example.
(Image: 1X (Screenshot))
OpenAI's robotics company, 1X, has built a natural language interface into its semi-humanoid service robot Eve. This enables the robot to perform pre-trained tasks, such as tidying up a room, when called upon.
The natural language interface developed by 1X is based on the semi-humanoid robot Eve, which has existed since around 2019 and was developed by 1X's predecessor company Halodi Robotics before the company changed its name to 1X as part of the partnership with OpenAI. Eve has a human-like upper body with two arms and hands, but has no legs. It moves on a wheeled platform. Although this restricts movement in environments made for humans, it simplifies the hardware considerably and saves the complicated process of learning to walk.
1X has now published a video showing how the Eve service robot can be called upon to tidy up an office environment using voice commands. 1X has now integrated the necessary voice interface into Eve in an initial version. It enables an operator to control several robots simultaneously by voice. The robots can then string together a series of previously learned actions and execute them in such a way that even more complex tasks can be completed.
AI behavior model
In March 2024, X1 had already developed an autonomous model that packs many tasks into a single AI behavior model. These include, for example, unpacking a shopping bag and deciding where to put the products, wiping up spilled liquids and folding shirts.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
However, the model also has its weaknesses, as 1X itself notes: Improving the behavior of a single task in a small multitask model harms the behavior of other tasks within the model. This can only be remedied by increasing the number of parameters. However, this means a longer training time, which in turn results in slower development.
According to 1X, the robots in the video act completely autonomously via a neural network. There is therefore no teleoperation. The video also does not contain any hidden cuts, accelerated images or animations.
(olb)