Scale Alone Does not Improve Mechanistic Interpretability in Vision Models
This highlights a critical gap in AI safety and understanding for researchers and practitioners, showing that scaling alone does not enhance interpretability, which is incremental in revealing limitations of current approaches.
The study investigated whether scaling up vision models improves mechanistic interpretability, finding no scaling effect and that state-of-the-art models are not easier to interpret than older ones like GoogLeNet, with some even being less interpretable.
In light of the recent widespread adoption of AI systems, understanding the internal information processing of neural networks has become increasingly critical. Most recently, machine vision has seen remarkable progress by scaling neural networks to unprecedented levels in dataset and model size. We here ask whether this extraordinary increase in scale also positively impacts the field of mechanistic interpretability. In other words, has our understanding of the inner workings of scaled neural networks improved as well? We use a psychophysical paradigm to quantify one form of mechanistic interpretability for a diverse suite of nine models and find no scaling effect for interpretability - neither for model nor dataset size. Specifically, none of the investigated state-of-the-art models are easier to interpret than the GoogLeNet model from almost a decade ago. Latest-generation vision models appear even less interpretable than older architectures, hinting at a regression rather than improvement, with modern models sacrificing interpretability for accuracy. These results highlight the need for models explicitly designed to be mechanistically interpretable and the need for more helpful interpretability methods to increase our understanding of networks at an atomic level. We release a dataset containing more than 130'000 human responses from our psychophysical evaluation of 767 units across nine models. This dataset facilitates research on automated instead of human-based interpretability evaluations, which can ultimately be leveraged to directly optimize the mechanistic interpretability of models.