A humanoid robot developed in 2016 is now capable of more human-like gestures thanks to the incorporation of GPT-4 — a large language model created by OpenAI.

Researchers from the University of Tokyo updated the latest iteration of the robot, dubbed Alter3, using GPT-4, which reportedly enabled the bot to perform simulations like taking a selfie, tossing a ball, eating popcorn and playing air guitar.

Source: University of Tokyo

The team explained that before the inclusion of GPT-4, such tasks would have required specific coding for each activity. However, GPT-4 has introduced capabilities to robots, enabling them to learn from natural language instruction.

AI-powered robots, the researchers added, "have been primarily focused on facilitating basic communication between life and robots within a computer, utilizing large language models (LLMs) to interpret and pretend life-like responses. Direct control is [now] feasible by mapping the linguistic expressions of human actions onto the robot's body through program code," they said.

Instead of having to code the coordination of several joints, the researchers suggest that LLMs free them up from iterative labor, and instead provide verbal instructions describing the movements they desire. Once the team issues a prompt, the LLM is instructed to create Python code that runs the Android engine.

An article detailing the development, "From Text to Motion: Grounding GPT-4 in a Humanoid Robot 'Alter3'," appears in the journal arXiv.

To see the Alter3 play air guitar, watch the accompanying video that appears courtesy of the University of Tokyo.