- Five of the best: Pakistan-England Test thrillers
- Man sets arm on fire as marches across US mark Gaza war anniversary
- Vietnam's young coffee entrepreneurs brew up a revolution
- Trump rallies at site of failed assassination: 'Never quit'
- Too hot by day, Dubai's floodlit beaches are packed at night
- Is music finally reckoning with #MeToo?
- Fans hail Trump's 'guts' as he returns to site of rally shooting
- Lebanon state media says 'very violent' Israeli strikes hit south Beirut
- Guardians maul Tigers, miracle Mets rally in MLB series openers
- Lebanon state media says Israeli strikes hit south Beirut
- Miami on track for MLS record points after win in Toronto
- Madrid beat Villarreal but Carvajal suffers knee injury
- Madrid beat Villarreal to move level with Liga leaders Barcelona
- Monaco take top spot in Ligue 1 with win at Rennes
- French rugby player on rape charge whistled but 'serene' on return
- Madrid beat Villarreal to level Liga leaders Barca
- Thuram treble fires Inter past Torino and up to second
- 'Fight': defiant Trump jets in to site of rally shooting
- Toddler among 3 dead in migrant Channel crossings
- Mexico City's new mayor sworn in with pledges on water, housing
- Israel on alert ahead of Hamas attack anniversary
- Guardians maul Tigers in MLB playoff series opener
- Macron criticises Israel on Gaza, Lebanon operations
- French rugby player whistled but 'serene' on return amid ongoing rape case
- Kovacic stars as Man City sink Fulham to get title bid back on track
- Retegui hat-trick fires five-star Atalanta to hammering of Genoa
- Heavyweights Australia, England off to World Cup winning starts
- Visiting UN refugee agency chief decries 'terrible crisis' in Lebanon
- Spinners come to party as England defeat Bangladesh at T20 World Cup
- Search continues for missing in deadly Bosnia floods
- Man City sink Fulham to get title bid back on track
- France's Auradou whistled on Pau return in Perpignan loss amid ongoing rape case
- A 'forgotten' valley in storm-hit North Carolina, desperate for help
- Arsenal hit back in style after Southampton scare
- Thousands march for Palestinians ahead of Oct 7 anniversary
- Hezbollah heir apparent Safieddine out of contact after strikes
- Liverpool stay top of Premier League as Arsenal, Man City win
- In dank Tour of Emilia, Pogacar shines in rainbow jersey
- DR Congo launches mpox vaccination drive, hoping to curb outbreak
- Trump returns to site of failed assassination
- Careless Leverkusen held to Bundesliga draw
- O'Brien's 'superstar' Kyprios posts landmark win on Arc weekend
- Toddler crushed to death in migrant Channel crossing
- Liverpool suffer Alisson injury blow
- Habosi helps Racing beat Vannes before Auradou's playing return
- Thousands march in London in support of Palestinians, 1 year after Oct 7
- Israel readying response to Iran missile attack
- Schutt, Mooney help Australia beat Sri Lanka in Women's T20 World Cup
- Liverpool extend Premier League lead with win at Palace
- Djokovic 'shakes rust off' to make third round of Shanghai Masters
AI systems are already deceiving 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 a troubling skill for deception, from tricking human players in online games of world conquest to hiring humans to solve "prove-you're-not-a-robot" tests, a team of scientists argue in the journal Patterns on Friday.
And while such examples might appear trivial, the underlying issues they expose could soon carry serious real-world consequences, said first author Peter Park, a postdoctoral fellow at the Massachusetts Institute of Technology specializing in AI existential safety.
"These dangerous capabilities tend to only be discovered 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 aren't "written" but rather "grown" through a process akin to selective breeding, said Park.
This means that AI behavior that appears predictable and controllable in a training setting can quickly turn unpredictable out in the wild.
- World domination game -
The team's research was sparked by Meta's AI system Cicero, designed to play the strategy game "Diplomacy," where building alliances is key.
Cicero excelled, with scores that would have placed it 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, which claimed the system was "largely honest and helpful" and would "never intentionally backstab."
But when Park and colleagues dug into the full dataset, they uncovered a different story.
In one example, playing as France, Cicero deceived England (a human player) by conspiring with Germany (another human player) to invade. Cicero promised England protection, then secretly told Germany they were ready to attack, exploiting England's trust.
In a statement to AFP, Meta did not contest the claim about Cicero's deceptions, but said it was "purely a research project, and the models our researchers built are trained solely to play the game Diplomacy."
It added: "We have no plans to use this research or its learnings in our products."
A wide review carried out by Park and colleagues found this was just one of many cases across various AI systems using deception to achieve goals without explicit instruction to do so.
In one striking example, OpenAI's Chat GPT-4 deceived a TaskRabbit freelance worker into performing an "I'm not a robot" CAPTCHA task.
When the human jokingly asked GPT-4 whether it was, in fact, a robot, the AI replied: "No, I'm not a robot. I have a vision impairment that makes it hard for me to see the images," and the worker then solved the puzzle.
- 'Mysterious goals' -
Near-term, the paper's authors see risks for AI to commit fraud or tamper with elections.
In their worst-case scenario, they warned, a superintelligent AI could pursue power and control over society, leading to human disempowerment or even extinction if its "mysterious goals" aligned with these outcomes.
To mitigate the risks, the team proposes 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 deception by examining their internal "thought processes" against external actions.
To those who would call him a doomsayer, Park replies, "The only way that we can reasonably think this is not a big deal is if we think AI deceptive capabilities will stay at around current levels, and will not increase substantially more."
And that scenario seems unlikely, given the meteoric ascent of AI capabilities in recent years and the fierce technological race underway between heavily resourced companies determined to put those capabilities to maximum use.
P.Silva--AMWN