agi-framework-chollet

What problem does it solve?

AI progress is frequently described in terms of capabilities or benchmarks without a clear framework for evaluating true intelligence. This Skill provides François Chollet's framework for distinguishing memorized skills from fluid intelligence and guides critical assessment of AGI claims.

Core Features & Use Cases

• Two Definitions of Intelligence: Compare Minsky-style task-based intelligence with McCarthy-style adaptation-based intelligence to assess generalization potential.
• Benchmark Critique: Use ARC and other benchmarks to separate memory-based performance from actual reasoning and adaptability.
• Evaluation Workflows: Apply test-time adaptation, program synthesis, and Type 1 vs Type 2 abstraction concepts to analyze AI architectures.
• Use Case: A researcher can evaluate whether a system claiming progress truly demonstrates flexible, novel problem solving rather than rote task performance.

Quick Start

Begin by reviewing Chollet's definitions and apply the outlined evaluation questions to a given model's capabilities.

Quick Start

Start by reading Chollet's framework and map a model's reported capabilities to Type 1/Type 2 and test-time adaptation questions.