Researchers from Loughborough University, MIT, and Yale have introduced the concept of ‘Collective AI.’ In a perspective paper published in Nature Machine Intelligence, they propose Shared Experience Lifelong Learning (ShELL) as a framework for creating decentralized AI systems composed of multiple independent agents, or “Collective AI.”
These individual AI units continuously learn and share knowledge throughout their lifetimes, challenging centralized monolithic architectures and resembling the hive mind concept.
How might collective AI work?
If developed, collective AI could mirror the abilities of sci-fi concepts like Star Trek’s “The Borg,” “The Get” from Mass Effect, or “The Replicators” from Stargate SG-1. ShELL systems enable agents to learn from their own experiences and shared knowledge, resulting in faster learning, improved performance, and greater flexibility akin to that of biological organisms.
Dr. Andrea Soltoggio of Loughborough University, the lead researcher, envisions a network of AI units capable of instant knowledge sharing and continuous adaptation to new data, enabling rapid responses to novel situations, challenges, or threats. This decentralized AI approach draws parallels to the human immune system, where multiple components collaborate to mount a coordinated defense against threats.
Potential real-world applications of ShELL include space exploration, personalized medicine, cybersecurity, disaster response, and multi-agent sensing for tasks like search and rescue operations.
Despite promising uses, researchers are cautious of risks such as the rapid dissemination of incorrect or unethical knowledge between units, suggesting measures to promote autonomy and balance cooperation.
Building collective AI involves mechanisms like lifelong machine learning, federated learning, multi-agent systems, and edge computing, drawing inspiration from bio-inspired AI architectures and recent developments.
Interest in decentralized AI is growing, with industry developments like the resignation of Stability AI CEO Emad Mostaque and the funding of startup Sakana indicating a shift towards decentralized projects diffusing AI power from Big Tech.
