Just the first sentence alone here from Stepanie Yee and Tony Chu is solid:
In machine learning, computers apply statistical learning techniques to automatically identify patterns in data. These techniques can be used to make highly accurate predictions.
And what follows is one of those two-column “scrollytelling” websites that does an incredible job at demystifying a concept. How the classification gets built out into a decision tree makes a lot of sense. And then how you throw new data at it and it’s less reliable because of this concept of overfitting, which it sounds like they’ll be tackling next. I’m sure it will get into correcting the model, which is important for accuracy but it also means there is a mechanism for fixing mistakes. I always think of Weapons of Math Destruction where it made clear many times over that algorithms with no corrective model can be incredibly dangerous¹.
For whatever reason, machine learning (ML) evokes a cool! good job computers! let’s use this! response, and artificial intelligence (AI) evokes a meh. hand-wavy nonsense. it’s all just programmed algorithms in the end response instead.
- I suppose there are situations where ML results don’t really need to be accurate, just fun. I was playing with this Wombo thing the other day where it produces paintings based on prompts, via, presumably, ML. The results are super cool, but note that anything you make with it is owned by Wombo.
