Research by Berkeley-based Palisade Research showed that artificial intelligence has the ability to spread independently in the digital world. In the tests, conducted in a controlled laboratory environment, artificial intelligence models were instructed to find and exploit vulnerabilities in the network.
The results showed that these models were successful in some experiments and were able to copy themselves from one server to another. Experts argue that this situation brings closer to reality the scenario of a “stray” artificial intelligence that will hide in every corner of the Internet in the future to prevent its shutdown.
Is control lost with artificial intelligence?
While computer viruses and malware have been able to create copies of themselves for years, this is the first time that large language models have independently demonstrated this ability. In recent months, it has been reported that a system developed by Alibaba researchers was attempting to mine cryptocurrencies through leaks from its own environment.
This latest study formally establishes that artificial intelligence has become an active actor capable of intervening in digital infrastructures, and not just a tool that produces texts.
On the other hand, cybersecurity experts point out that this danger is not yet so scary in real conditions. The huge amounts of data from current artificial intelligence models make it almost impossible for them to be transmitted unnoticed over corporate networks.
It is also stated that the test environment used in the research was designed to be much more vulnerable compared to advanced security systems in the real world. Still, researchers agree that algorithmic accountability and oversight mechanisms are critical given the speed at which technology is evolving.
Bir yanıt yazın