My wife’s a teacher so she’s had months of guff about education from people who know nothing about it.

As I write this there’s a load of guff going around about algorithms by people who know nothing about them.

An algorithm is just a way of making a decision.

If you’re trying to decide whether to take an umbrella with you, your algorithm might be “gather data (look at the sky) and if it’s dark grey take the umbrella”. Simple right?

But that algorithm is based on a piece of knowledge - dark skies normally mean rain coming soon. Asking someone who has only ever lived in Death Valley to write your umbrella algorithm might look very different.

However now we have a sea change in algorithms. Machine learning has changed everything - instead of asking a human being to use their experience to make the decisions, we give a machine tons of data and it spots trends and analyses that humans can’t see. And then uses those to make the decisions.

Which is amazing.

However the decisions that come from this are only as good the data that the system is trained on. If you give the umbrella algorithm nothing but Death Valley weather data it’s not likely to be able to predict umbrella usage in Manchester.

So algorithms aren’t the problem. We use them every day.

The real problem is keeping control of who writes them, who trains them. And no one is looking at that.

The Art and Science of Ruby

The algorithm made me do it