Connectionism
A foundational philosophy of AI: intelligence emerges from networks of simple, interconnected units (neurons), not from explicit rules.
Think of it like
Instead of following a recipe (symbolic AI), you learn by adjusting many tiny weights based on mistakes — like a student learning by doing, not by instruction manual.
Example
Neural networks (including today's Transformers) are connectionist: they learn by backpropagation, adjusting weights until the system produces correct outputs.
How it actually works
Connectionism emerged in the 1980s as a counter to symbolic AI. The idea: intelligence doesn't arise from explicit knowledge and rules, but from learned patterns in networks of artificial neurons. Early connectionist systems were limited (small networks, slow training), but as computation scaled, they dramatically outperformed symbolic approaches. Today's deep learning and LLMs are thoroughly connectionist — they learn numerical representations, not symbols. The term "connectionism" is less common now (we just say "neural networks" or "deep learning"), but it represents a major philosophical shift: from top-down knowledge engineering to bottom-up learning.
For product teams
A winning philosophy: learning from data proved vastly more scalable and general than hand-crafted rules.
For engineers
An AI approach based on networks of artificial neurons with weighted connections trained via learning algorithms like backpropagation.
Related
- Symbolic AI — Overcame symbolic AI limitations.
- Neural network — Foundation of modern neural networks.
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