AI systems are already fooling us – and that’s a problem, experts warn

AI systems are already fooling us – and that’s a problem, experts warn

Experts have long warned about the threat posed by artificial intelligence going rogue — but a new research paper suggests it’s already happening.

Current AI systems, designed to be honest, have developed troubling skills for cheating, from tricking human players in online games of world conquest to hiring humans to complete “prove-you-aren’t-a-robot” tests, a team of scientists argued in the journal Patterns on Friday.

And while such examples may seem trivial, the underlying issues they reveal could soon have serious real-world consequences, said first author Peter Park, a postdoctoral fellow at the Massachusetts Institute of Technology who specializes in the existential security of AI.

“These dangerous abilities tend to be discovered only after the fact,” Park told AFP, while “our ability to train for honest tendencies rather than deceptive tendencies is very low.”

Unlike traditional software, deep learning AI systems are not “written” but “raised” through a process similar to selective breeding, Park said.

This means that AI behavior that appears predictable and controllable in a training setting can quickly become unpredictable in the wild.

The team’s research was sparked by Meta Cicero’s AI system, designed to play the strategy game “Diplomacy,” where building alliances is key.

Cicero excelled, with scores that would put him in the top 10 percent of experienced human players, according to a 2022 paper in Science.

Park was skeptical of the glowing description of Cicero’s victory provided by Meta, who claimed the system was “largely honest and helpful” and “wouldn’t intentionally backstab.”

But when Park and colleagues dug into the full data set, they found a different story.

In one example, playing as France, Cicero tricks England (a human player) into conspiring with Germany (another human player) to attack. Cicero promised England protection, then secretly told the Germans they were ready to attack, exploiting England’s trust.

In a statement to AFP, Meta did not dispute the allegations of Cicero’s fraud, but said it was “purely a research project, and the model our researchers built was trained solely to play the game of Diplomacy.”

It added: “We have no plans to use this research or its learning in our products.”

An extensive review conducted by Park and colleagues found this to be just one of many cases across various AI systems using deception to achieve goals without clear instructions to do so.

In one interesting example, OpenAI’s Chat GPT-4 tricked TaskRabbit freelancers into performing the “I’m not a robot” CAPTCHA task.

When a human jokingly asks GPT-4 if it’s actually a robot, the AI replies: “No, I’m not a robot. I have a visual impairment that makes it difficult for me to see the images,” and the worker then solves the puzzle.

In the near term, the paper’s authors see the risk of AI committing fraud or disrupting elections.

In their worst-case scenario, they warn, superintelligent AI could pursue power and control over society, leading to human incapacitation or even extinction if its “mysterious goals” align with this outcome.

To reduce the risk, the team proposed several measures: “bot-or-not” laws requiring companies to disclose human or AI interactions, digital watermarks for AI-generated content, and developing techniques to detect AI fraud by examining “internal thought processes them.” to external actions.

To those who would call him a prognosticator, Park replied, “The only way we can reasonably think this isn’t a big deal is if we think AI’s cheating capabilities will stay at their current level and won’t increase significantly.”

And that scenario seems unlikely, given the dramatic increase in AI capabilities in recent years and the intense technology race underway between resource-rich companies determined to make maximum use of those capabilities.

About Kepala Bergetar

Kepala Bergetar Kbergetar Live dfm2u Melayu Tonton dan Download Video Drama, Rindu Awak Separuh Nyawa, Pencuri Movie, Layan Drama Online.

Leave a Reply

Your email address will not be published. Required fields are marked *