GOFAI
A slang term for pre-neural-network AI — symbolic, logic-based, hand-crafted knowledge — now mostly a punchline about how hard we got it wrong.
Think of it like
Like the Model T versus modern cars: it worked, it was impressive for its time, but you'd never go back.
Example
In the 1960s, researchers believed GOFAI would soon reach human-level reasoning; by the 1990s, researchers had switched to probabilistic and connectionist methods, and "GOFAI" became a term for the naive old ways.
How it actually works
The acronym was popularized by researcher John Haugeland in his 1985 book. By that time, the "good old-fashioned" approach — explicit rules, logic, symbolic knowledge — was already failing at real-world tasks. It couldn't handle uncertainty, required immense manual labor to encode expertise, and broke on edge cases. The rise of machine learning and neural networks displaced it almost entirely. Today GOFAI is mainly historical; the term itself is a bit affectionate and a bit mocking.
For product teams
Instructive failure: the assumption that intelligence could be built top-down through logic turned out to be far harder than scaling up learning from data.
For engineers
Umbrella term for pre-learning-era approaches: rule engines, theorem provers, knowledge graphs, logic programming.
Related
- Symbolic AI — A synonym for classical symbolic AI.
- Machine learning — Displaced by learning-based approaches.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome