AI time bombs could sneak cyberattacks past watchful eyes
Malicious code hidden inside neural networks could hijack things like image recognition algorithms long after people start using them.
The situation: Image recognition AIs can be tricked quite easily, which raises the specter of, say, a cyberattack convincing a self-driving car to ignore a stop sign. But what if malware could be woven into algorithms so that they were, in effect, programmed to mess up?
The fear: A new paper shows how certain neural networks could be tainted by sneaking in malicious code. The nefarious program then sits there, waiting for a trigger that activates it to hijack the system and force it to start falsely predicting or classifying data.
Why it matters: The US government already worries that hardware built in other countries could have back doors that allow foreign agents to spy on or take control of computerized systems. High-tech paranoia? Maybe. But this latest work suggests that even AI isn’t immune to digital cloak-and-dagger tactics.
Deep Dive
Artificial intelligence
Large language models can do jaw-dropping things. But nobody knows exactly why.
And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.
Google DeepMind’s new generative model makes Super Mario–like games from scratch
Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.
What’s next for generative video
OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.
Stay connected
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.