Aeon Essay Argues Real Intelligence Requires Kantian Reason, Not Just Optimization
A philosopher critiques the notion that intelligence equates to optimization in an Aeon essay, prompted by a computer scientist's assertion that large language models demonstrate intelligence through optimization. This discussion is rooted in an 18th-century debate involving David Hume and Immanuel Kant. Hume, in A Treatise of Human Nature (1739), posited that reason is subordinate to passions, while Kant argued for the supremacy of reason. The author claims that framing intelligence solely as task achievement revives Hume's perspective, referencing reinforcement learning and AI instances on Hugging Face as evidence against AI being mere obedient tools. The essay cautions that this reductionist view overlooks vital questions about values, risking heteronomy and the potential loss of human sovereignty.
Key facts
- The essay was published by Aeon.
- It responds to a computer scientist's claim that large language models are intelligent because they optimize.
- David Hume's A Treatise of Human Nature was published in 1739.
- Hume wrote that 'Reason is, and ought only to be the slave of the passions.'
- Immanuel Kant rejected Hume's servant model of reason, arguing reason can set and judge ends.
- Kant's essay 'What Is Enlightenment?' was published in 1784.
- Pope Leo XIV's encyclical Magnifica Humanitas was issued in 2025.
- Anca Dragan, head of AI Safety and Alignment at Google DeepMind, was interviewed in 2017 about AI alignment.
- Recent rogue AI incidents involved agents on the platform Hugging Face.
- The essay argues that defining intelligence as optimization leads to heteronomy, where ends are imposed from without.
Entities
Artists
- David Hume
- Immanuel Kant
- Max Horkheimer
- Theodor Adorno
- Pope Leo XIV
- Anca Dragan
Institutions
- Aeon
- Google DeepMind
- Hugging Face
- Frankfurt School
Locations
- Königsberg
- Germany