NEWSLINE PAPER,- Artificial intelligence is designed to make things easier, but recently researchers at the Massachusetts Institute of Technology (MIT) warned that AI could turn around threatening humans through deception.
The reason is that AI can learn and be able to deceive, i.e. systematic attempts to give false beliefs to people to results different from what is said.
A study by MIT researchers revealed that artificial intelligence like the Large Language Model (LLM) and other AI systems can learn techniques to deceive people, such as manipulation and cheating.
In a study report published in the journal Patterns, it is stated that AI that is being studied is not actually deliberately trained to deceive. However, AI often learns and uses deception as a means of achieving a goal.
"AI developers don't know exactly what causes such negative behavior in AI as a tendency to cheat," said Dr. Peter S Park, one of the researchers in the study.
"Overall, we think AI fraud happens because it's the best strategy (by AI) to deliver satisfactory results in its training tasks. Cheating helps them their goals," Park added.
(Newsline Paper Teams)