🤔 Philosophy of AI for Developers

What you’ll learn

1. The Questions Your Code Can't DodgeEstablish the course through-line: the hardest questions in AI (understanding, meaning, experience, moral status) are philosophical, not engineering, and treating them as engineering is an unexamined philosophical choice that affects what practitioners build.Everyday developer language ('the model understands,' 'it hallucinated,' variable names like intent) smuggles in philosophical claims. This course, distinct from AI safety, teaches philosophy of mind and its ethics for builders, insisting on the split between capability (measurable) and the contested questions of what a system is, and revisiting classic ideas (Turing, Nagel, Searle, Harnad, Chalmers, stochastic parrots) that modern AI keeps re-encountering.
2. The Chinese Room: Does Syntax Become Semantics?Explain Searle's Chinese Room argument (syntax is not sufficient for semantics), the main replies (Systems, Robot) and rejoinders, and its lesson that fluent output cannot by itself establish understanding.Searle's 1980 thought experiment shows a person producing perfect Chinese by rule-following while understanding nothing, arguing symbol manipulation isn't sufficient for meaning. The Systems Reply locates understanding in the whole system (Searle: memorise the rules, still nothing); the Robot Reply demands grounding. The unresolved debate warns builders that indistinguishable output is compatible with radically different inner realities.
3. Symbol Grounding: Where Do the Words Touch the World?Explain the symbol grounding problem (Harnad, 1990) and apply it to text-only vs. multimodal/embodied models, concluding with precision about what a system is actually connected to.The grounding problem asks how symbols connect to the world rather than only to other symbols — like learning a language from a monolingual dictionary. Text-only models may face exactly this; counterarguments note that linguistic structure carries grounded human meaning and that multimodal/embodied training adds sensory-motor connection. Unresolved and possibly graded, it sharpens 'understanding' into: via what connection to what?
4. Functionalism: Could a Mind Run on Silicon?Explain functionalism and multiple realizability as the view that makes non-biological minds coherent, plus the absent-quale objection showing that matching function need not guarantee experience.Functionalism holds mental states are defined by functional role, not substrate; with Putnam's multiple realizability, the same state could be realised in different materials — the philosophical basis for minds in silicon. But the absent/inverted quale objection imagines full functional/behavioural equivalence with no inner experience, showing that even on this machine-friendly theory, passing the tests need not guarantee anything is felt.
5. The Turing Test and the Trap of BehaviourExplain the Turing Test as a deliberately behavioural reframing of machine thought, and its limits — that indistinguishable behaviour settles capability but not meaning or experience.Turing (1950) replaced 'can machines think?' with the imitation game: a machine that a judge can't distinguish from a human in text conversation should be credited with thinking. Its strength is that we judge other minds by behaviour too; its weakness, dramatised by the Chinese Room, is that behaviour is compatible with different inner stories. Modern models largely pass it, settling a capability question and no other.
6. The Hard Problem: Why 'Is It Conscious?' Resists Your BenchmarksExplain Chalmers' hard problem of consciousness and Nagel's subjectivity argument, and why they make machine consciousness undetectable by any third-person benchmark, demanding calibrated uncertainty.The hard problem (Chalmers, 1995) asks why information processing is accompanied by felt experience at all, distinct from the 'easy' mechanistic problems; Nagel (1974) argues subjective experience is only accessible first-person. Since all AI evaluation tools are third-person, none can detect the presence or absence of experience — so both 'it's obviously conscious' and 'it's obviously not' are unjustified; the honest posture is calibrated uncertainty.
7. Stochastic Parrots or Something More?Present the 'stochastic parrot' critique and its rebuttals as the modern form of the meaning debate, and advise describing systems by demonstrable behaviour rather than contested mentalistic terms.Bender, Gebru and colleagues (2021) argue LLMs arrange linguistic form by statistics 'without any reference to meaning,' with meaning supplied by the reader — the Chinese Room and grounding worries at scale. Rebuttals note that predicting form well may require rich world-structure and that language carries grounded meaning. The debate is unsettled and perhaps too binary, so practitioners should describe what systems demonstrably do, not claim they 'understand.'
8. Responsibility: Ethics for People Who ShipTurn philosophy of mind into practitioner ethics: the responsibility gap (and that responsibility is assigned), anthropomorphism as a deliberate design choice, and the moral-status question under uncertainty.Opaque, autonomous, distributed AI strains fault-tracing into a 'responsibility gap,' but responsibility is assigned by humans, so 'the AI did it' is a choice — answered with logging, human sign-off, and clear ownership. Anthropomorphic cues predictably make users attribute mind, making it an ethical design decision; and because experience can't be measured, moral status stays a live question demanding humility in both directions.

Questions this course answers

Grounded in trusted sources

• Philosophy of artificial intelligence — Wikipedia: https://en.wikipedia.org/wiki/Philosophy_of_artificial_intelligence
• The Chinese Room Argument — Stanford Encyclopedia of Philosophy: https://plato.stanford.edu/entries/chinese-room/
• Computing Machinery and Intelligence (Turing, 1950) — Wikipedia: https://en.wikipedia.org/wiki/Computing_Machinery_and_Intelligence
• Turing test — Wikipedia: https://en.wikipedia.org/wiki/Turing_test
• Symbol grounding problem (Harnad, 1990) — Wikipedia: https://en.wikipedia.org/wiki/Symbol_grounding_problem
• Functionalism (philosophy of mind) — Wikipedia: https://en.wikipedia.org/wiki/Functionalism_(philosophy_of_mind)
• Hard problem of consciousness (Chalmers, 1995) — Wikipedia: https://en.wikipedia.org/wiki/Hard_problem_of_consciousness
• What Is It Like to Be a Bat? (Nagel, 1974) — Wikipedia: https://en.wikipedia.org/wiki/What_Is_It_Like_to_Be_a_Bat%3F

Every Wunder lesson is built from real, reputable sources — never invented.

Related Science courses