Luke: What is AI and machine learning? 

Mike:

Hello, I’m Mike Tipping. I’m a consultant in machine learning and AI.

I’ve be working in the field for about 25 years now both in industry and also in academia; I’m also a visiting professor of mathematics at the University of Bath. And I’ve worked in everything from fraud detection to AI for video games in my Microsoft days.

So, people often ask me what AI is, and there is a kind of a jokey answer that it’s a way of programming computers to make the same mistakes that humans make. But ultimately it is really about designing computer algorithms that behave in potentially a way that we would view as being intelligent, that replicates certain human capabilities, that traditionally to now computers just haven’t been able to do.

A lot of the successes we’re seeing in AI, at the moment, are being driven by something called machine learning. Which is this idea of rather writing down the algorithm that we want to run to process our data, to detect our disease or to make our financial estimate or something, rather to hand code that and scratch our heads and work it out, we take these big datasets and we apply something what are called learning algorithms to them.

And then in principle we learn these models from the data without having to actually understand how they work. And that can, of course, can be quite a disadvantage if you don’t know how they work. There are issues of transparency, but at the same time it’s enabling us to achieve quite a lot that we’ve not been able to do before.