ChatGPT Speaks Through Figure 01 Robot Following $675M Funding
Robotics startup Figure has showcased a new voice feature for its Figure 01 line, incorporating ChatGPT and OpenAI’s text-to-speech model, as seen in the video above. The demonstration comes just a couple of weeks after Figure and OpenAI inked a deal to work on new AI models for robots and the OpenAI Startup Fund joined a new $675 million funding round for Figure.
Robot ChatGPT
The video shows Figure 01 accurately identifying objects within its visual range, offering an apple to a person requesting food based on logical deduction, and executing tasks like dish storage and trash disposal. Impressively, these tasks are accompanied by the robot’s ability to articulate its thought process in a natural-sounding voice, underscoring the advanced level of its learned behaviors and the absence of remote operation.
“We’ve always planned to come back to robotics and we see a path with Figure to explore what humanoid robots can achieve when powered by highly capable multimodal models,” OpenAI vice president of products and partnerships Peter Welinder said when announcing the deal with Figure. “We’re blown away by Figure’s progress to date and we look forward to working together to open up new possibilities for how robots can help in everyday life.”
The collaboration leverages the robot’s onboard cameras to feed visual data into a large vision-language model. This model processes the conversational history to generate appropriate verbal responses and determine the most suitable learned behaviors for executing commands. These behaviors are powered by neural network visuomotor transformer policies, showcasing a sophisticated integration of AI and robotics.
The partnership between Figure and OpenAI is expected to significantly shorten the commercial development timeline for Figure 01. With BMW as a notable client planning to deploy general-purpose robots on its production lines, the advancements in Figure 01’s capabilities could soon see BMW employees working alongside robots capable of handling mundane tasks and providing companionship, marking a significant milestone in the integration of AI and robotics in the industrial sector.
Generative AI for robots is fast becoming a new focus for the technology. That includes plenty of investment, such as the $100 million raised by robotics startup 1X recently, including from OpenAI. Meanwhile, Nvidia is using GPT-4 to train robots to perform delicate tasks more quickly and cheaply than previously feasible. There are also University of Tokyo researchers who linked GPT-4 to a humanoid robot named Alter3 and successfully taught it to dance and make other movements purely through conversation.
Follow @voicebotaiFollow @erichschwartz
Google DeepMind Shows Off New Generative AI-Based RT-2 Robot Command Language
