These four words get used interchangeably and it causes endless confusion. Here is the accurate mental model — they are nested circles, not synonyms.
Artificial Intelligence (any machine mimicking "intelligent" behaviour)
└─ Machine Learning (systems that LEARN patterns from data, not hand-coded rules)
└─ Deep Learning (ML using many-layered neural networks)
└─ Generative AI (deep models that CREATE — text, images, code, audio)The one idea that changed everything
Traditional software: a human writes the rules. if spam_words > 5: mark_as_spam(). Machine learning flips this — you show the machine thousands of labelled examples (spam / not-spam) and it learns the rule itself. You provide data + answers; it produces the program.
Where each lives in 2026
- Classic ML — fraud detection, credit scoring, recommendations, demand forecasting. Runs most of the money in tech.
- Deep learning — image recognition, speech-to-text, self-driving perception.
- Generative AI / LLMs — ChatGPT, Claude, Copilot, Midjourney. Text, code and images from a prompt.
What AI is NOT
It is not magic and not "conscious". An LLM is a very sophisticated next-token predictor trained on huge text. Understanding that removes both the fear and the hype. We build up to exactly how it works in the LLM track.
Your path: Foundations → a little Python → Machine Learning → Deep Learning → LLMs → shipping apps. You do not need heavy math to start. Begin with How Machines Actually Learn.