Superhuman AI Bots Are Surprisingly Vulnerable to Exploits

image via New Scientist

July 11, 2024, 10:00 AM

• •Researchers have revealed weaknesses in one of the most successful AI systems — a bot that plays the board game Go and can beat the world’s best human players.
• •The study raises questions about whether more general AI systems will suffer from vulnerabilities that could compromise their safety and reliability, and even their claim to be ‘superhuman’.
• •The researchers used adversarial bots to test three ways of defending Go AIs against such attacks.

A new study has found that even the best AI systems can be defeated by adversarial attacks. This means that AI systems are not as superhuman as they are often made out to be, and that humans may still have an edge over them in some areas.

Read original article