A new study is painting a dire picture of the future of AI, saying AI models are at risk of ‘Model Autophagy Disorder’ (MAD), the digital equivalent to mad cow disease.

Companies are racing to train AIs on as much data as they can feed them, gobbling up text, images, videos, and more. The practice has drawn criticism, however, raising questions of ownership and copyright. The controversy has led some companies to evaluate synthetic data as an alternative to real data, but Rice University is warning of a major downside.

“The problems arise when this synthetic data training is, inevitably, repeated, forming a kind of a feedback loop ⎯ what we call an autophagous or ‘self-consuming’ loop,” said Richard Baraniuk, Rice’s C. Sidney Burrus Professor of Electrical and Computer Engineering. “Our group has worked extensively on such feedback loops, and the bad news is that even after a few generations of such training, the new models can become irreparably corrupted. This has been termed ‘model collapse’ by some ⎯ most recently by colleagues in the field in the context of large language models (LLMs). We, however, find the term ‘Model Autophagy Disorder’ (MAD) more apt, by analogy to mad cow disease.”

The report goes on to highlight the etymology of the MAD label.

Mad cow disease is a fatal neurodegenerative illness that affects cows and has a human equivalent caused by consuming infected meat. A major outbreak in the 1980-90s brought attention to the fact that mad cow disease proliferated as a result of the practice of feeding cows the processed leftovers of their slaughtered peers ⎯ hence the term “autophagy,” from the Greek auto-, which means “self,”’ and phagy ⎯ “to eat.”

To test the phenomenon, scientist used synthetic data in AI training loops to see what the result would be. The results were both revealing and disturbing.

Progressive iterations of the loops revealed that, over time and in the absence of sufficient fresh real data, the models would generate increasingly warped outputs lacking either quality, diversity or both. In other words, the more fresh data, the healthier the AI.

Side-by-side comparisons of image datasets resulting from successive generations of a model paint an eerie picture of potential AI futures. Datasets consisting of human faces become increasingly streaked with gridlike scars ⎯ what the authors call “generative artifacts” ⎯ or look more and more like the same person. Datasets consisting of numbers morph into indecipherable scribbles.

The scientist ultimately concluded that AI models must consume a diet of real data to avoid MADness.

“Our theoretical and empirical analyses have enabled us to extrapolate what might happen as generative models become ubiquitous and train future models in self-consuming loops,” Baraniuk said. “Some ramifications are clear: without enough fresh real data, future generative models are doomed to MADness.”

The study, in all its detail, is a revealing look at the future of AI models and could have profound implications on where AI will fit in. Proponents who claim AI will continue to revolutionize industries and become ever more capable, as well as critics who fear AI will wipe out entire industries and job descriptions, may both be in for a surprise.

Unless AI researchers can solve the MAD dilemma, the human element will never full be replaced, since there will always be a need for fresh data that AI can continue to be trained on.