tech
Google DeepMind is worried about what happens when millions of agents start to interact
The firm is calling for more scientists to study the risks of multi-agent systems.
TL;DR
- Google DeepMind is funding research into the risks of millions of interacting AI agents.
- This mass-market arrival of autonomous agents creates a new class of risk.
- Google DeepMind, Schmidt Sciences, ARIA, Cooperative AI, and Google.org are contributing $10 million to fund research in this area.
- The goal is to encourage academic research to look far into the future and study multi-agent system behavior.
- Potential risks include scams, prompt injections, and other cyberattacks, mirroring existing online issues but amplified.
- Researchers plan to use realistic simulations and sandboxes to study agent behavior.
- The complexity arises from numerous interactions, and AI agents may not always act rationally.
- Other AI firms like Anthropic are also issuing guidelines for AI agent deployment.