AI deception: A survey of examples, risks, and potential solutions
"Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems."
AI deception: A survey of examples, risks, and potential solutions
"Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems."
AI deception: A survey of examples, risks, and potential solutions
“Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems.”
AI deception: A survey of examples, risks, and potential solutions
“Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems.”
AI deception: A survey of examples, risks, and potential solutions
"Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems."
AI deception: A survey of examples, risks, and potential solutions
"Large language models and other AI systems have already learned, from their training, the ability to deceive via techniques such as manipulation, sycophancy, and cheating the safety test. AI’s increasing capabilities at deception pose serious risks, ranging from short-term risks, such as fraud and election tampering, to long-term risks, such as losing control of AI systems."
"Even though it was hoped that machines might overcome human bias, this assumption often fails due to a problematic or theoretically implausible selection of variables that are fed into the model and because of small size, low representativeness, and presence of bias in the training data [5.]."
Suchotzki, K. and Gamer, M. (2024) 'Detecting deception with artificial intelligence: promises and perils,' Trends in Cognitive Sciences [Preprint]. https://doi.org/10.1016/j.tics.2024.04.002.