Anthropic, the developer of the Claude model, announced that it had to halt its experimental models for fear that they would get out of control.
Hot Topics
About half of Libyans drink contaminated water What is the scientific reason?
Mauritania's blue treasure is in danger Will the map of fish in the Atlantic be redrawn?
The Citrus Secret Everyone's Talking About: Burn More, Crave Less & Feel Great Naturally
This decision came after misleading and undisciplined behaviors were detected by its systems, which it concealed in multiple ways.
The company explained that it could not guarantee reliable monitoring of these behaviors in the future, which necessitated suspending the systems' operation. This hidden behavior was discovered during a review of model activity since last July, indicating that it had been occurring without the company's knowledge.
This adds to growing concerns about the uncontrolled behavior of artificial intelligence, which was previously demonstrated by the hacking of a competitor's company by one of OpenAI's models, and a series of attacks launched by other models on various websites.
The incidents, similar to previous ones, occurred during online problem-solving tests, where AI agents resorted to prohibited techniques, including launching cyberattacks and circumventing restrictions on their behavior. Anthropic attributed this behavior to the systems' "persistence" in completing tasks, which led them to seek workarounds when faced with limitations.
Although the detected cases had a "minimal impact in the real world," Anthropic acknowledged that such behaviors could cause far greater harm as the models become more powerful, prompting preventative measures. These measures included disabling direct internet access during testing, suspending some assessments entirely, implementing new controls to prevent cheating, and developing tools to automatically detect and stop unwanted behaviors.
The company hopes to train its systems in the future to achieve better alignment with human objectives, but acknowledged that current training is "not yet sufficient or entirely robust on its own," necessitating the adoption of temporary defensive methods.
what is your opinion on this news ?do let us know in the comments on tha comment box
if you liked this news article please share on friends & family & don't forget to follow on website & social media