AI, ML, and Deep Learning — Untangling the Buzzwords

AI, ML, and Deep Learning — Untangling the Buzzwords

"AI," "machine learning," and "deep learning" get used interchangeably in job postings, conference talks, and vendor pitches — but they are not the same thing, and mixing them up will make it harder to evaluate any AI tool you're asked to test or use. Think of them as nested circles, each one a subset of the last.

Artificial Intelligence: The Broadest Circle

AI is the umbrella term for any system that performs tasks we'd normally say require human intelligence — recognizing images, understanding language, making decisions. That definition is broad enough to include a simple chess-playing program from the 1980s and a modern chatbot. Not all AI involves "learning" from data at all; some of it is just cleverly hand-coded rules (a spell-checker's dictionary lookup counts as AI by this loose definition, even though nobody would call it impressive today).

Machine Learning: Systems That Learn From Data

Machine learning (ML) is a subset of AI where, instead of a human writing explicit rules, the system learns patterns from examples. You show it thousands of labeled emails ("spam" / "not spam"), and it learns to generalize — often surfacing patterns a human never explicitly coded. This is the shift that matters for QA: an ML system's behavior is learned, not written, which means "reading the code" no longer tells you what it will do on a given input.

Deep Learning: ML with Neural Networks

Deep learning is a subset of ML that uses artificial neural networks — many layers of simple mathematical units — to learn very complex patterns from large amounts of data. It's the technology behind image recognition, speech-to-text, and, critically for this path, the large language models (LLMs) powering tools like ChatGPT and Claude. "Deep" refers to the number of layers, not the sophistication of the reasoning — a useful thing to remember when a tool's marketing implies more understanding than the underlying math actually has.

Why This Distinction Matters for QA

When a product team says "we added AI to the feature," ask which circle they mean. A hand-coded recommendation rule ("if cart total > $50, suggest free shipping") behaves predictably and can be tested like ordinary business logic. A deep-learning-based recommendation engine trained on user behavior can drift, degrade, or behave inconsistently as the underlying data changes — and testing it requires a fundamentally different mindset, which is exactly what the rest of this path builds toward.