AI can now detect signs of violent intent before violence occurs. That may save lives, but it also raises an unsettling question: when does prevention become prediction, and prediction become judgment?
There was a story in the news recently about a 36-year-old farmer who was arrested after allegedly discussing plans with ChatGPT to murder his eight-year-old son and other people.
OpenAI, the company behind ChatGPT, detected what it described as a credible threat of harm to others in the conversations and reported it to the FBI in the United States. The FBI then passed the information to Brazilian authorities.
According to reports, when police searched the suspect’s home, they found bottles containing various substances and lengths of rope. He was also allegedly researching poisonous substances and asking questions in different ways in an apparent attempt to bypass the AI platform’s safety restrictions.
This is, of course, a good thing. If the intervention helped save the life of an eight-year-old boy, and perhaps others, that is something to be welcomed. But as I went through the details of the story, I couldn’t escape the feeling that it reminded me of something.
It did not take long to find it: Minority Report, the film based on Philip K. Dick’s 1956 novella of the same name. The film, starring #Tom_Cruise is set in 2054.
At the heart of the story is a pilot “Precrime” system operating in Washington, D.C., in which murders are predicted before they happen. Police use these predictions to identify and arrest would-be killers before they can commit their crimes. Those arrested are imprisoned for crimes they have not yet committed.
The central plot, of course, revolves around the system identifying one of the very people responsible for running the Precrime program as a future murderer. He is then forced to fight the system he once trusted in order to prove that the future is not necessarily predetermined.
What matters here is not that ChatGPT has suddenly become the Precrime system of Minority Report. It has not. In the Brazilian case, there were reportedly conversations, searches and physical evidence that authorities could investigate. But the parallel is still worth considering.
Science-fiction writers imagined systems capable of anticipating human behaviour decades before such questions became technologically plausible. More importantly, they warned us about placing too much trust in systems that make judgments about what people might do rather than what they have actually done.
Today, artificial intelligence can identify patterns in enormous amounts of information and flag behaviour that appears dangerous. In some circumstances, that capability could save lives. But the same principle raises difficult questions: How certain must a threat be before a private company reports someone to the authorities? Who decides where disturbing fantasy ends and genuine intent begins? How much context can an automated system truly understand? And what safeguards exist when it gets that judgment wrong?
There is also a broader question. What happens when such systems are used not only to prevent an imminent act of violence, but by individuals, companies or governments pursuing other objectives? A technology designed to identify genuine threats could, under different rules or in different hands, become a powerful instrument of surveillance, profiling or control.
In the case of the Brazilian boy, the story appears to have had a happy ending. But Minority Report was never really a warning against preventing murder. Its warning was about something more fundamental: the danger of confusing prediction with certainty.
As artificial intelligence becomes increasingly capable of observing, interpreting and anticipating human behaviour, that distinction may become one of the most important ethical questions we face.
The future has not happened yet. We should be very careful about treating it as evidence.
No comments:
Post a Comment