Multispin Physics of AI Tipping Points and Hallucinations
This addresses the problem of AI transparency and risk quantification for users and developers, with potential legal implications, though it appears incremental in applying physics models to AI behavior.
The paper tackles the problem of generative AI outputs mysteriously tipping from correct to misleading, which reportedly caused $67 billion in losses and several deaths in 2024, by establishing a mathematical mapping to a multispin thermal system to reveal a hidden tipping instability at the scale of the AI's basic Attention head.
Output from generative AI such as ChatGPT, can be repetitive and biased. But more worrying is that this output can mysteriously tip mid-response from good (correct) to bad (misleading or wrong) without the user noticing. In 2024 alone, this reportedly caused $67 billion in losses and several deaths. Establishing a mathematical mapping to a multispin thermal system, we reveal a hidden tipping instability at the scale of the AI's 'atom' (basic Attention head). We derive a simple but essentially exact formula for this tipping point which shows directly the impact of a user's prompt choice and the AI's training bias. We then show how the output tipping can get amplified by the AI's multilayer architecture. As well as helping improve AI transparency, explainability and performance, our results open a path to quantifying users' AI risk and legal liabilities.