Adversarial AI Attacks Highlight Fundamental Security Issues


Researchers from the Massachusetts Institute of Technology (MIT), the University of California at Berkeley, and FAR AI discovered that a professional-level Go AI could be easily defeated with moves that tricked the machine into believing the game was over.

While a professional or novice Go player could be defeated by a Go AI by using a logical set of movies, an adversarial attack could easily defeat the computer by making choices that no rational player would typically make.

Also Read: Cybersecurity: An ounce of Prevention Could be worth a pound of Cure

According to Adam Gleave, a doctoral candidate in artificial intelligence at the University of California, Berkeley, these attacks show that while AI technology may operate at superhuman levels and even be extensively tested in real-life scenarios, it is still susceptible to unanticipated inputs.

Read More: Adversarial AI Attacks Highlight Fundamental Security Issues

For more such updates follow us on Google News ITsecuritywire News. Please subscribe to our Newsletter for more updates.