Rationalism grounds intelligence in discrete, compositional symbols and explicit rules for transforming them. In the classical physical-symbol-system view, reasoning must be specified in logic, making structure explicit but risking brittleness when messy details or grounding are underspecified. section 7.6
Its central dispute with connectionism concerns where reasoning structure comes from: rationalists prescribe symbolic programs, while empiricists learn representations from data and inductive biases. Critics emphasize compositionality; a connectionist reply treats hard rules as higher-level descriptions of softer neural constraints. section 7.6
From philosophical dispute to benchmark
The Malik–Hinton wager illustrates a pragmatic test of competing assumptions: rather than argue about neural intelligence, Malik demanded performance on PASCAL, while Hinton answered with ImageNet and AlexNet. Standardized benchmarks, datasets, and measured baselines became the practical way to assess the debate. section 7.6
