(Photo by Ralph Hutter on Unsplash)

Deep down, the world of machine learning (ML) and AI is based on a simple idea: "get a machine to look for patterns in data."

And from there, it can do five things:

1/ make numeric predictions (regression)

2/ put things into buckets (classification)

3/ group similar things (unsupervised learning, like clustering)

4/ find weird stuff (such as anomaly detection)

5/ create stuff (like genAI chatbots and image creators)

(Sharp-eyed readers will point out that 1 and 2 are different types of supervised learning. Items 3 and 4 are conjugate pairs. Item 5 is like running item 1 in reverse.)

Five things, that's it.

Now, those five ideas can be expressed in a lot of ways. Yes.

But if someone's trying to sell you on an AI project or product that doesn't map to one of those five high-level themes … it's time to ask some deeper questions.