SECAI Core · phase 1 of 15
Deep learning
A refinement of machine learning that enables a machine to develop strategies for solving a task given a labeled dataset and without further explicit instructions.
The Explain card
- Plain English
- Deep learning is machine learning using many-layered neural networks. Given a labeled dataset, the network works out its own useful features and strategies without explicit instructions.
- Example
- A phishing detector is shown raw screenshots of login pages labeled "real" or "fake". Instead of engineers choosing features, the network learns by itself which visual cues betray a fake.
- Why it matters
- Because the network invents its own features, nobody can fully read its reasoning. That opacity hides bias, backdoors and strange failure modes, and it is what makes adversarial examples possible.
- Hook
- More layers, more power, less visibility into why.
Where it sits in the deck
Phase 1: What AI Is: Core Concepts and Paradigms
You cannot secure, govern, or attack something you cannot define — establish what AI actually is before any other concept can land.
