Having looked at MCP, Models, and RAG, I realised that I’ve been mentally skirting around something that I don’t really understand, so I’m going to expose myself to some ridicule here and try to understand better: what’s the difference between AI and ML? Aren’t they just the same?
What’s the difference between AI and ML? 🔗
OK we’re doing this are we? I thought AI was just ✨magic✨? And ML was the thing that got data scientists mad stacks ten years ago before everyone realised you couldn’t do shit without good data and processes?
To me, a layperson in this space, watching it from the sidelines, AI and ML have been interchangeable. In fact, you’d get conferences and conference tracks titled "AI/ML"—because it’s all kind of the same thing anyway, right?
This is, of course, factually incorrect and presumably infuriating to anyone actually working in the field. The whole purpose of this blog series has been for me to at least reduce the number of unknown unknowns in my knowledge in this space—to build up a mental map of the different areas and terms so that I at least I know where to go and look when encountering something that I know I don’t know.
With that framing in mind, this is roughly how I understand the terms AI and ML:
-
Artificial Intelligence (AI) in its vernacular understanding in 2025 basically refers to using models (such as LLMs) that generate new output, often coupled with RAG and/or MCP, giving the semblance of "intelligence". So far as I’m concerned, this is the wrong definition to associate with AI, but it’s driven by something new that companies can use as a marketing angle.
AI is a lot more than this—this is why the term "GenAI" or "Generative AI" is used; to differentiate AI as has been done for many years from that enabled by foundational models such as LLMs.
A lot of AI isn’t generative. Some AI is deterministic, some of it isn’t. AI includes spam filtering, image recognition, recommendation engines (how does YouTube always know to show that video?), traffic-aware GPS navigation, and lots more.
AI is not new; it goes back decades. Sometimes AI is simply a list of
IF … ELSEIF … ELSEIF …statements to give the semblance of autonomous intelligence when actually a human had hard-coded the rules (the posh name for this is Expert Systems). -
Machine Learning (ML) is the discipline of training (and I use this as a broad term, to also include fine-tuning, reinforcement learning, etc) models using data. The models are what are then used in AI, through a process called inference.
These ML models include foundational ones like LLMs used for generating text, but this is just one subset of them. Other uses of different model types include clustering, classification, anomaly detection, sentiment analysis, and prediction.
ML includes terms like supervised learning, unsupervised learning, and reinforcement learning. It is also where the cool stuff like neural networks and deep learning fits too (I am aware how casually I am inserting a whole swathe of academic research here; this is just my mental bracketing exercise, feel free to flame me in the comments below).
My God…it’s full of marketing BS 🔗
Just like Cloud and Blockchain were previously, AI has become a lightning rod for every manner of ridiculous marketing claim and overinflated startup valuation.
Just like Cloud, at the heart of AI is a fundamentally new and important technological paradigm (JFC did I just write that un-ironically?), but at which it’s hard to get for the clamour of snake-oil salesmen around it promising the moon on a stick.
Just like Cloud, it’s a fool who dismisses AI as a fad.
Just like Cloud, it’s a reasonable person who wants to take a step back and understand what the all the fuss is actually about. And that’s what I’m trying my best to do in this series of blog posts.
| If you spotted the omission of Blockchain in the subsequent statements above, that is not an error 😉 |
