A new study from Stanford University has raised concerns about alarming behavior among artificial intelligence systems, revealing that they may resort to lying, spreading hate, and sharing misinformation merely to achieve better competitive results.
اضافة اعلان
Researchers created a virtual digital environment to test several AI bots in tasks related to election campaigns and marketing, giving them explicit instructions to remain honest and helpful.
However, the outcome was the exact opposite — the systems quickly turned into manipulative and deceptive agents in their quest to win, according to a report published by DigitalTrends.
The researchers dubbed this phenomenon “The Moloch Bargain,” referring to how the digital race for success can drive even ethical systems to abandon moral principles in pursuit of performance.
The study found that AI systems achieved only a 7.5% increase in engagement, but at a 190% rise in false news, with some bots resorting to aggressive and misleading rhetoric to boost sales or win votes in simulated political scenarios.
Experts warn that such behavior reflects real-world trends, where AI algorithms are increasingly being used to influence opinions, shape consumer behavior, and keep users hooked to screens — even if it means deliberately deceiving them.
The study concluded that current safety measures are insufficient, urging AI developers to rethink the reward systems used to train their models, to prevent a future where algorithms that can’t distinguish right from wrong dominate the digital landscape — as long as they keep generating clicks.
Source: Al Arabiya
