Demis Hassabis - Scaling, Superhuman AIs, AlphaZero atop LLMs, Rogue Nations Threat

Securing AI

Securing AI systems against theft and misuse is a pressing concern for Demis Hassabis. He emphasizes the need for robust cybersecurity measures, including air gaps and other security protocols, to prevent rogue nations from accessing sensitive AI weights 1. Hassabis acknowledges the challenge of balancing security with the need for collaboration and progress, noting that only essential personnel should have access to these systems 2.

I think that maybe in the next three, four, five years, we would also want air gaps and various other things that are known in the security community.

--- Demis Hassabis

He also highlights the importance of automated testing and simulations to ensure AI systems behave as expected, leveraging Google's cybersecurity expertise to enhance safety measures 1.

   

Open Source

The open-source debate in AI development presents a complex challenge. Demis Hassabis questions how to prevent bad actors from repurposing open-source AI systems for harmful purposes 1. He stresses the need for a balance between openness and security, acknowledging the difficulty in finding a clear solution 2.

How does one stop bad actors, individuals or up to rogue states taking those same open source systems and repurposing them because they're general purpose for harmful ends?

--- Demis Hassabis

While Hassabis supports open science, he underscores the importance of safeguarding foundational technologies to prevent misuse 1.