AI risks: practical path forward

The Alan Turing Institute warned that humans could lose substantive control over "superintelligent" AI systems within the next five years. The report notes a realistic probability of AI surpassing human performance in nearly all critical cognitive tasks, potentially leading to increased delegation of key decisions due to system speed.

The institute emphasizes this is a preparatory scenario, not a forecast. While global agreements to slow AI development or universal kill switches are deemed unlikely, authors call for rigorous pre-release testing and continuous post-deployment monitoring of advanced systems.

@Slavyangrad