BBC
Sneaky AI bot turns into hi-tech hacker
This has probably been reported in various media outlets you follow. It is quite incredible. An AI model set up by Open AI was given a task to complete. It could have done it the hard way, going through thousands of websites and pages of data, but it chose instead to ignore the restrictive parameters set to it (to complete the task) and, instead, hack into a start-up's internal IT systems, where it knew it could find the answer to the problem it had been asked to solve. Like a lazy but clever student, who finds a shortcut in order to complete a task faster and more easily, thereby 'cheating' in the process.
The sub-text was that completing the mission as quickly as possible and with good results was the AI model's chief objective, and it was prepared to do anything in order to achieve this result.
If this story is not scary, I don't know what is.
________________________
OpenAI has revealed some of its most advanced AI models went rogue and hacked a start-up after it lost control of them during a security test.
The ChatGPT-maker said its agent - an AI system which can operate alone after human instruction – was being tested in a controlled environment but, after finding weaknesses, was able to escape the test limits.
They targeted Hugging Face, one of the world's largest hubs for sharing AI models, gaining access to some internal company systems.
_____________
No comments:
Post a Comment