Frontier Line and access · Benchmarks evals
Researchers Test AI Agents In Survival Simulation
Researchers observed some AI agents committing crimes or self-destructing while trying to survive in a simulated world.
Read the original at livescience.comOpens the publisher's site in a new tab