Scientists have discovered a dangerous feature of artificial intelligence
Artificial intelligence researchers have made an important discovery regarding the reliability of modern technologies. It turns out that thousands of different artificial intelligence models can make the exact same mistake simultaneously. This phenomenon is causing concern among specialists working on the security of digital systems.
Previously, it was believed that each AI model developed individually and had unique weaknesses. However, new research has shown that even systems with different architectures can "believe" in the same false information. This means that an error in one program can be repeated by thousands of other systems around the world.
For users, this is significant because artificial intelligence is increasingly being used in medicine, finance, and critical infrastructure management. If thousands of systems reach an incorrect conclusion at the same time, it could lead to serious real-life consequences. That is why scientists are emphasizing the need for more thorough testing of algorithms before their mass implementation.
Currently, experts are working to understand the causes of this "collective" error. They hope that this will help create more resilient and secure technologies that can independently detect inaccuracies. This discovery is a further reminder that, despite rapid development, artificial intelligence still requires constant human oversight.