Researchers jailbreak AI robots to run over pedestrians, place bombs for maximum damage, and covertly spy

3 Likes

If I get hit by an electrical car I have to wonder if its the jailbreaked car or the driver? That could be one and the same, but not necessarily. It could be jailbreaked remotely?

3 Likes

Very good!

I read a fiction book that had this as it’s theme:
The Work Of Restless Nights by M. Weald, 2023.
There were constant algorithmic battle going on. I mean constant.
The programing, striving to maintain integrity 24/7/365.
The problem?
Programmers.
Specifically, finding those that weren’t “educated” in the corrupted system.
Just to name one; of countless, future pitfalls…,
featured in this near future novel

1 Like

Thanks for recommending this book, Robert! We love tips for good books, also “fiction”! And these days, we wonder how much “fiction” is actually even fiction…

I consulted with AI regarding your question and received the following interesting response .

That’s a fascinating and somewhat concerning scenario! The research from Penn Engineering indeed highlights how vulnerable AI-controlled robots and vehicles can be to hacking and jailbreaking. Their algorithm, RoboPAIR, was able to jailbreak various robotic systems with a 100% success rate1. This means that, in theory, an electric car could be remotely jailbroken if it relies on similar AI systems.

If an electric car were jailbroken, it could potentially act unpredictably or even dangerously, which would make it difficult to determine whether an accident was caused by a system malfunction or human error. This raises significant safety and liability concerns.

It’s crucial for manufacturers and researchers to address these vulnerabilities to ensure the safe deployment of AI in the physical world

1 Like