Signaltheorie der Intelligenz

Abstract  

Where does intelligence in artificial or natural systems come from? This work demonstrates that intelligence does not arise from the complexity of biological structures or the size of artificial models, but from an arithmetically necessary architecture of signal processing. An intelligent system requires only four elements: a linear projection, a linear reconstruction via the transpose, a non-linearity, and a recursive connection. This minimal architecture inevitably generates signal completion, i.e. the supplementation of missing or noisy signals. This gives rise to internal models, categorical states and a cyclical alternation between elementary and complex forms — the dynamic core of thought. Modern AI architectures such as Transformers and State Space Models merely extend this structure with a third linear mapping for pattern recognition, but follow the same principle. The theory thus explains the functioning of both natural brains and artificial systems, demonstrating that intelligence is an inevitable property of recursive, signal-completing systems.

Monografie von Dr. rer. nat. Andreas Heinrich Malczan