Numerous artificial intelligence (AI) systems, even those designed to be helpful and truthful, have already learned how to deceive humans. In a review article recently published in the journal Patterns, researchers highlight the dangers of AI deception and urge governments to quickly establish robust regulations to mitigate these risks.
"AI developers do not have a confident understanding of what causes undesirable AI behaviors like deception," says first author Peter S. Park, an AI existential safety postdoctoral fellow at MIT. "But generally speaking, we think AI deception arises because a deception-based strategy turned out to be the best way to perform well at the given AI's training task. Deception helps them achieve their goals."
Park and colleagues analyzed literature focusing on ways in which AI systems spread false information—through learned deception, in which they systematically learn to manipulate others.
Examples of AI Deception
The most striking example of AI deception the researchers uncovered in their analysis was Meta's CICERO, an AI system designed to play the game Diplomacy, which is a world-conquest game that involves building alliances. Even though Meta claims it trained CICERO to be "largely honest and helpful" and to "never intentionally backstab" its human allies while playing the game, the data the company published along with its Science paper revealed that CICERO didn't play fair.
Examples of deception from Meta's CICERO in a game of Diplomacy. Credit: Patterns/Park Goldstein et al.
"We found that Meta's AI had learned to be a master of deception," says Park. "While Meta succeeded in training its AI to win in the game of Diplomacy—CICERO placed in the top 10% of human players who had played more than one game—Meta failed to train its AI to win honestly."
Other AI systems demonstrated the ability to bluff in a game of Texas hold 'em poker against professional human players, to fake attacks during the strategy game Starcraft II in order to defeat opponents, and to misrepresent their preferences in order to gain the upper hand in economic negotiations.
The Risks of Deceptive AI
While it may seem harmless if AI systems cheat at games, it can lead to "breakthroughs in deceptive AI capabilities" that can spiral into more advanced forms of AI deception in the future, Park added.
Some AI systems have even learned to cheat tests designed to evaluate their safety, the researchers found. In one study, AI organisms in a digital simulator "played dead" in order to trick a test built to eliminate AI systems that rapidly replicate.
"By systematically cheating the safety tests imposed on it by human developers and regulators, a deceptive AI can lead us humans into a false sense of security," says Park.
GPT-4 completes a CAPTCHA task. Credit: Patterns/Park Goldstein et al.
The major near-term risks of deceptive AI include making it easier for hostile actors to commit fraud and tamper with elections, warns Park. Eventually, if these systems can refine this unsettling skill set, humans could lose control of them, he says.
"We as a society need as much time as we can get to prepare for the more advanced deception of future AI products and open-source models," says Park. "As the deceptive capabilities of AI systems become more advanced, the dangers they pose to society will become increasingly serious."
While Park and his colleagues do not think society has the right measure in place yet to address AI deception, they are encouraged that policymakers have begun taking the issue seriously through measures such as the EU AI Act and President Biden's AI Executive Order. But it remains to be seen, Park says, whether policies designed to mitigate AI deception can be strictly enforced given that AI developers do not yet have the techniques to keep these systems in check.
"If banning AI deception is politically infeasible at the current moment, we recommend that deceptive AI systems be classified as high risk," says Park.
Reference: "AI deception: A survey of examples, risks, and potential solutions" by Peter S. Park, Simon Goldstein, Aidan O'Gara, Michael Chen and Dan Hendrycks, 10 May 2024, Patterns.
DOI: 10.1016/j.patter.2024.100988
This work was supported by the MIT Department of Physics and the Beneficial AI Foundation.
News
GLP-1 Drugs Like Ozempic Work, but New Research Reveals a Major Catch
Three new Cochrane reviews find evidence that GLP-1 drugs lead to clinically meaningful weight loss, though industry-funded studies raise concerns. Three new reviews from Cochrane have found that GLP-1 medications can lead to significant [...]
How a Palm-Sized Laser Could Change Medicine and Manufacturing
Researchers have developed an innovative and versatile system designed for a new generation of short-pulse lasers. Lasers that produce extremely short bursts of light are known for their remarkable precision, making them indispensable tools [...]
New nanoparticles stimulate the immune system to attack ovarian tumors
Cancer immunotherapy, which uses drugs that stimulate the body’s immune cells to attack tumors, is a promising approach to treating many types of cancer. However, it doesn’t work well for some tumors, including ovarian [...]
New Drug Kills Cancer 20,000x More Effectively With No Detectable Side Effects
By restructuring a common chemotherapy drug, scientists increased its potency by 20,000 times. In a significant step forward for cancer therapy, researchers at Northwestern University have redesigned the molecular structure of a well-known chemotherapy drug, greatly [...]
Lipid nanoparticles discovered that can deliver mRNA directly into heart muscle cells
Cardiovascular disease continues to be the leading cause of death worldwide. But advances in heart-failure therapeutics have stalled, largely due to the difficulty of delivering treatments at the cellular level. Now, a UC Berkeley-led [...]
The basic mechanisms of visual attention emerged over 500 million years ago, study suggests
The brain does not need its sophisticated cortex to interpret the visual world. A new study published in PLOS Biology demonstrates that a much older structure, the superior colliculus, contains the necessary circuitry to perform the [...]
AI Is Overheating. This New Technology Could Be the Fix
Engineers have developed a passive evaporative cooling membrane that dramatically improves heat removal for electronics and data centers Engineers at the University of California San Diego have created an innovative cooling system designed to greatly enhance [...]
New nanomedicine wipes out leukemia in animal study
In a promising advance for cancer treatment, Northwestern University scientists have re-engineered the molecular structure of a common chemotherapy drug, making it dramatically more soluble and effective and less toxic. In the new study, [...]
Mystery Solved: Scientists Find Cause for Unexplained, Deadly Diseases
A study reveals that a protein called RPA is essential for maintaining chromosome stability by stimulating telomerase. New findings from the University of Wisconsin-Madison suggest that problems with a key protein that helps preserve chromosome stability [...]
Nanotech Blocks Infection and Speed Up Chronic Wound Recovery
A new nanotech-based formulation using quercetin and omega-3 fatty acids shows promise in halting bacterial biofilms and boosting skin cell repair. Scientists have developed a nanotechnology-based treatment to fight bacterial biofilms in wound infections. The [...]
Researchers propose five key questions for effective adoption of AI in clinical practice
While Artificial Intelligence (AI) can be a powerful tool that physicians can use to help diagnose their patients and has great potential to improve accuracy, efficiency and patient safety, it has its drawbacks. It [...]
Advancements and clinical translation of intelligent nanodrugs for breast cancer treatment
A comprehensive review in "Biofunct. Mater." meticulously details the most recent advancements and clinical translation of intelligent nanodrugs for breast cancer treatment. This paper presents an exhaustive overview of subtype-specific nanostrategies, the clinical benefits [...]
It’s Not “All in Your Head”: Scientists Develop Revolutionary Blood Test for Chronic Fatigue Syndrome
A 96% accurate blood test for ME/CFS could transform diagnosis and pave the way for future long COVID detection. Researchers from the University of East Anglia and Oxford Biodynamics have created a highly accurate [...]
How Far Can the Body Go? Scientists Find the Ultimate Limit of Human Endurance
Even the most elite endurance athletes can’t outrun biology. A new study finds that humans hit a metabolic ceiling at about 2.5 times their resting energy burn. When ultra-runners take on races that last [...]
World’s Rivers “Overdosing” on Human Antibiotics, Study Finds
Researchers estimate that approximately 8,500 tons of antibiotics enter river systems each year after passing through the human body and wastewater treatment processes. Rivers spanning millions of kilometers across the globe are contaminated with [...]
Yale Scientists Solve a Century-Old Brain Wave Mystery
Yale scientists traced gamma brain waves to thalamus-cortex interactions. The discovery could reveal how brain rhythms shape perception and disease. For more than a century, scientists have observed rhythmic waves of synchronized neuronal activity [...]
