OpenAI has established a team to train AI models for bots, and the company behind the chatbot ChatGPT confirmed the Forbes report in a statement.
This news comes about four years after the company disbanded a team that researched ways to use artificial intelligence to teach robots new tasks.
Forbes reported that several former members of the team still work at the company, which could facilitate its efforts to return to the market.
According to a job posting on its website, OpenAI plans to develop a set of large multimodal language models for robotics use cases.
A multimodal model is a neural network that processes text and other types of input. This input could include, for example, data from sensors within the robot.
A number of other companies are integrating large language models into autonomous machines. Covariant, a startup founded by former OpenAI researchers, previously launched a large language model called RFM-1 for the robotics market. The model allows robots to ask humans for help when faced with a difficult task.
The RFM-1 also lends itself to other use cases, according to Covariant.
Before performing a task, such as moving a box, a robot equipped with the large language model can generate a short video clip depicting how the action is performed. The robot can then study this video to find the efficient way to move the box.
The OpenAI job posting did not specify how the large language models it intends to develop for the robotics market will work.
The company explained that the position’s responsibilities include exploring new model architectures. This suggests that the large language models the company seeks to develop may rely on a different design than the decoder architecture supporting most advanced language models.
According to Forbes, OpenAI founded the robotics team about two months ago. The company is expected to make the large language models the team may develop available to third-party robot manufacturers.
OpenAI competes with a number of companies in this market by seeking to hire engineers who can support development efforts.
OpenAI’s launch of its robotics unit comes four years after the dissolution of an internal team that was focused on the same market.
In 2017, the company’s original robotics team demonstrated an artificial intelligence system that allows robots to learn through demonstration.
Engineers can teach the system new tasks by showing them how to perform these tasks using virtual reality headsets.
In 2018, OpenAI released a set of simulation tools to teach AI programs new skills.
