Scientists create AI models that can talk to each other and pass on skills with limited human input

Researchers have successfully demonstrated that artificial intelligence agents can teach each other using natural language. By integrating a neural network with a pre-trained language model, an AI learned to execute physical tasks based solely on written instructions. It then described its learned processes to a "sister" AI, which replicated the tasks without any prior training. This breakthrough marks the first time two AIs have communicated purely linguistically to transfer practical skills, moving beyond simple text generation to actual operational knowledge sharing. This capability mirrors human cognitive functions, where verbal or written instructions allow us to perform new actions without physical demonstration. Unlike previous AI chatbots that cannot translate text into physical motor actions, this new architecture bridges that gap by simulating brain areas responsible for language and movement. The system achieved high accuracy in performing psychophysical tasks solely through linguistic input, effectively combining sensory input interpretation with motor response generation in a way that mimics human learning. For open data, this development is significant as it establishes a new standard for interoperability and knowledge transfer between autonomous systems. It suggests a future where robots and AI agents can efficiently share operational data and protocols through natural language, reducing the need for manual reprogramming. This could revolutionize fields like manufacturing and logistics, where machines can autonomously train one another. Furthermore, understanding these linguistic-motor connections provides valuable insights for building more transparent and explainable AI systems, potentially enhancing how data models communicate complex behaviors to human users.

Source: livescience.com
Published on 2024-03-23

Retail sales beat expectations by holding steady in February »

Open Data

Related news