Abstract
I propose a model of systematic understanding, suitable for machine learning systems. On this account, an agent understands a property of a target system when it contains an adequate internal model that tracks real regularities, is coupled to the target by stable bridge principles, and supports reliable prediction. I argue that contemporary deep learning systems often can and do achieve such understanding. However they generally fall short of the ideal of scientific understanding: the understanding is symbolically misaligned with the target system, not explicitly reductive, and only weakly unifying. I label this the Fractured Understanding Hypothesis.
Key Points
- A systematic model of understanding: Understanding a property of a target system requires an adequate internal model that (1) tracks real regularities, (2) is coupled to the target via stable bridge principles, and (3) supports reliable prediction.
- Deep learning qualifies: Contemporary deep learning systems often can and do achieve understanding in this sense.
- The Fractured Understanding Hypothesis: Despite achieving understanding, deep learning systems fall short of the ideal of scientific understanding because their understanding is:
- symbolically misaligned with the target system,
- not explicitly reductive,
- only weakly unifying.