What exactly is an algorithm?
What exactly is an algorithm?
JON STROUD: They're everywhere. Apparently. Algorithms are in the social media apps we use every day, in search engines, even in dating apps. But I'll be honest I've got absolutely no idea what they actually are or how they work or who makes them. Should we be worried about them? Can they think for themselves? Rather than be baffled by science, I'm going to go find out. WHAT ON EARTH IS AN ALGORITHM? Where does every good piece of research start? A search engine. So what is an algorithm? Can I even spell it? Oh, I think so. OK. A process or set of rules to be followed in calculations or other problem solving operations, especially by a computer. Sounds like a riddle. After an hour of searching about the internet it all felt a little overwhelming. So many different definitions without any clarity in what they actually do. During my search I came across professor Victoria Nash from the Oxford Internet Institute. I called to pick her brains. Speaking to Victoria has helped me understand what algorithms are but how can the same thing that helps me bake a cake, also give me the best results from a search engine? No idea. I'm taking a trip to Oxford to visit one of Victoria's colleagues, Dr Bernie Hogan. Bernie, great to see you. BERNIE HOGAN, Oxford Internet Institute: Jon, nice to see you. JON STROUD: Where on earth are we? BERNIE HOGAN: So this is our university's data centre. You know, it's pretty big. It's really noisy. JON STROUD: Really noisy. BERNIE HOGAN: There's a lot of computation happening here. Each one of these belong to different departments. They're doing different kinds of calculations. JON STROUD: So there's thousands of algorithms going on in all these massive boxes like now? BERNIE HOGAN: Thousands? Try billions. JON STROUD: Billions? BERNIE HOGAN: Billions of algorithms. JON STROUD: Shall we go somewhere a little bit quieter and talk in a little bit more detail and try and get a better understanding of how all of this works? BERNIE HOGAN: Sure. Let's do it. JON STROUD: It's weird to think how much of our lives go on in nondescript server rooms all over the world. But what exactly are these billions of algorithms doing? BERNIE HOGAN: So the reason that we use a list of instructions is because we have a lot of data and we have to deal with that data. Now data could be anything. Data could be a list of towns in the UK and how I get from one town to another, or it could be a number of tweets. Which tweet is going to show up at the top of the list? JON STROUD: Right. BERNIE HOGAN: Algorithms calculate based on a bunch of features, the sort of things that will put something at the top of the list and then something at the bottom of the list. JON STROUD: So if it's that simple, should we be scared of algorithms? BERNIE HOGAN: Well the trick with algorithms that we perhaps should be a bit concerned about is what happens in the black box. JON STROUD: So is that like, when you search for something you don't know what their algorithm is doing because we can't see it. BERNIE HOGAN: Well a classic case of this is people talk about searching for prices for flights, and depending on which day you search on you might get a different flight, where you search from. And so this can mean difference of hundreds of pounds. That's an example where an algorithm is not transparent and perhaps should be. JON STROUD: Can algorithms think for themselves? BERNIE HOGAN: Well we wouldn't necessarily think of a computer as thinking, but we know that algorithms can learn. They can learn from other algorithms and algorithms can create their own instructions now. But the basis of it is still the same. Data goes in, goes through instructions, result comes out. JON STROUD: I'm beginning to get it, but I've still never actually seen an algorithm. I don't even know what they look like. So I'm heading to one of the UK's leading coding schools to see for myself what goes into making one. JON STROUD: So this is code, right? ISABEL MACCABEE, Northcoders: Yeah. JON STROUD: So what's the difference between an algorithm and code? ISABEL MACCABEE: Coding is algorithms that a computer can run the instructions for you. So we have to do it in a language which the computer can actually understand. So we've written this in Scratch and it's really nice to use, really intuitive, and you can just drag and drop these blocks and we've got these instructions for the drone to follow. JON STROUD: It's going to do a flip? ISABEL MACCABEE: It's going to do a flip. I hope it's going to do a flip. JON STROUD: It's time for a challenge - a drone race. Izzy's algorithm versus me. ISABEL MACCABEE: So the course is through the hoop, do a flip, come back round, land and again. JON STROUD: So technically, because yours is programmed by an algorithm, you should be able to do exactly the same thing three times without a problem. ISABEL MACCABEE: That's the plan. You ready? JON STROUD: Challenge accepted. ISABEL MACCABEE: Three, two, one. JON STROUD: I don't know what I'm doing. ISABEL MACCABEE: Take off. Yes. JON STROUD: OK. Right. ISABEL MACCABEE: So we're going set the speed. Fly up. And then it's going fly left. JON STROUD: You're already ahead but I think mine's more reliable. ISABEL MACCABEE: Hopefully it's going go forwards. Go forwards, there we go. Nice. JON STROUD: No, how did it do that? Yes. ISABEL MACCABEE: Full turn. JON STROUD: I'm going to make the time back in speed. Speed. ISABEL MACCABEE: And then forwards. Jon Stroud: No! JON STROUD: Go, go, go, go, go, go, go, go, go. ISABEL MACCABEE: Yes, oh, oh, oh! JON STROUD: Oh. Woah, woah. Sorry cameraman! Isabel Maccabee: I can just go and make a cup of tea. I'm just going to leave it. JON STROUD: No! [LAUGHS] ISABEL MACCABEE: And down. Oh look how calm. JON STROUD: Have you done your three laps? ISABEL MACCABEE: I did do the three laps, yeah. JON STROUD: So you've won? ISABEL MACCABEE: Yes. So I think some of the big benefits of having algorithms versus humans is that you don't have the human error that, no offence, I think you had. You don't have the human error in the same way. The computer goes through the instructions and that's all they know how to do. The person writing the code could have written an error and that's where problems can arise but the computer doesn't make mistakes it just does what it's supposed to do. JON STROUD: A computer might only do what it's supposed to do, but what are the ethical considerations around algorithms making so many decisions for us? One of the public concerns is that computers are taking over the world, robots are going to take over the world, algorithms are going to take all of our jobs. Is that going to be the case? DR ALLISON GARDNER, WOMEN LEADING IN AI: Taking our jobs, it's possible. But also deskilling humans if we become too dependent upon them and too trusting of them, it can deskill us as well. But on the flip side of it, they can be hugely beneficial and useful speeding up decision-making, making whole processes efficient, maybe spotting things that we might not have spotted ourselves. So we mustn't be frightened of them, we just must use them in the correct manner. JON STROUD: What have we learnt then? Algorithms are actually remarkably simple. Just like Bernie said - data goes in, follows a list of instructions and a result comes out. In some parts of the world, algorithms are now used in the criminal justice system, in social care, in credit checks - they're prolific machines making decisions that directly affect our human lives, not just the adverts that you see on the internet or the people you match on dating apps. The question for society isn't the algorithm, it's who controls the algorithm and where the data comes from that goes into them.
