Mathematician David Sumpter finds that, contrary to what many of us believe, algorithms aren’t killing civilisation – and they can be a force for good
I am standing in a small group at a party, taking a short break from my time locked in my office, coding algorithms, fitting statistical models, writing about my results and interviewing engineers and scientists over Skype.
There is small talk. I listen politely. Then someone brings it up – the stories may vary, but the theme is the same.
“The danger with Facebook is that it controls what people see,” he says. “I read about one study they did where they only showed people negative posts and they got depressed. It can control our emotions and these negative vibes spread like a virus.” I remain quiet and let someone else speak.
“Yes. They profit from selling their data. It’s very likely that Trump won the US election because of this company from Oxford or somewhere that downloaded everyone’s profile,” she says.
“You see, the problem is that their algorithms have been trained on fake news,” says the next one.
“It’s the same with Google. They built some computer to understand language and it started saying it hated Muslims.”
“I know,” the first man returns to the conversation, “and this is just the start. The scientists reckon that in 20 years’ time there will be a computer with human-like intelligence that will decide to turn us into paper clips. I read it in Elon Musk’s book.”
Debunking the myths
It is then that I explode. “First of all,” I say, “the Facebook study showed that people who were bombarded with negative news had a propensity to write one more negative word per month, a minuscule but statistically significant effect.
“Secondly, the company, called Cambridge Analytica, had nothing to do with Trump’s election success. Their CEO has made a series of unverifiable claims about what they can do.
“Thirdly, yes, computers trained on our language do make sexist and racist associations but that’s because our society is implicitly prejudiced. Most Google searches give you, and the rest of the population, the blandest, most accurate results imaginable. The biggest problem with the service is that it is overwhelmed with meaningless links trying to get you to go to Amazon and buy more crap.
“And finally, there have been some interesting developments in neural networks recently, but the question of whether we can create a general AI remains wide open.”
Everyone looks at me. I hate myself. I hate the boring idiot that I have become, who has read all the scientific papers, who has to spoil everything with details and caveats.
I know that I have misjudged the situation. I am being pedantic and petty. Beyond the small inaccuracies in the details, the people I am talking to have genuine worries about how society is changing. This is why they are talking about Facebook, Cambridge Analytica and artificial intelligence.
Algorithms are not as scary as we think
Directly after my visit to Google over a year ago now while carrying out research for my book Outnumbered, looking into how algorithms are affecting our online lives, I had felt the same way they do. I had felt afraid about the future that mathematicians and computer scientists were creating for us.
After a year of examining the evidence, I now understand that algorithms are not scary in the way I had thought they might be. It is sad that algorithms haven’t solved the problems our society has with sexism and racism, but they haven’t made them worse, either. They have highlighted issues about bias that we all need to work harder to solve.
It is frightening that SCL group can start a company like Cambridge Analytica that markets itself as targeting personalities, when it doesn’t have the tools or the data to do so, but that’s global capitalism for you. Either accept it, or make sure you attack the system and not its lies and deceptions.
Read more
It’s official: Facebook makes us stressed but we can’t live without it
Now Google wants to be your friend after the Facebook scandal
How deleting your Facebook account could cause some problems
Fake news, troll bots and Tinder
It’s depressing that Facebook contains so much fake news and Twitter is full of troll bots, but it’s nice to know that hardly anyone is listening to them.
It’s worrying that success online is only partially related to talent, but it is a consolation when we fail.
It’s a bit of a bummer that you need to be good-looking to hook up on Tinder, but that’s not really a big change for you, is it?
Understanding algorithms allows us to understand future scenarios a little bit better. If you can understand how algorithms work today, then it is easier to judge which scenarios are realistic and which are not.
The biggest risk I had found from algorithms was when we failed to think rationally about their effects, when we got carried away in science fiction scenarios.
Parents’ fears
I am conscious now that my journey, as I stand here after my failed attempt to respond to what was just a bit of “have a go at Facebook” banter and “what does it all mean” artificial intelligence speculation, has made me a not very interesting person to talk to.
Luckily, I am married. My wife starts talking about Pokémon Go, the conversation moves on, and I tune out, thinking about our 14-year-old daughter, Elise. She went off the other week to meet a friend from a chat group on Snapchat. My wife Lovisa and I had been worried at first – might this online “friend” turn out to be a 40-year-old paedophile?
Our concerns were unjustified. Elise met up with a normal 13-year-old with bright blue hair.
This summer she wants to go to visit another friend she has made online who lives in Poland. They often chat on Skype while they do their homework together. Elise’s parents will have to think long and hard about whether this will be allowed or not.
My son, Henry, and I had just returned from a trip to Newcastle. Through Twitter, I had got in contact with another dad, Ryan, who like me trains his son’s football team. I’d arranged to take 32 12-year-old boys from Sweden, to play a series of matches against his team.
Read more
The private information Facebook’s Mark Zuckerberg does not want to share
Tobacco and cigars are being sold on Facebook through an advertising loophole
Facebook knows how often I text my wife and get a curry – and it’s creepy
The upside of algorithms
Ryan invited me to talk about Soccermatics at his work, the Department of Work and Pensions. There, I found myself in a small room packed with data scientists with the same enthusiasm for visualising and understanding data.
After my talk, Ryan showed me what he and his colleagues were doing. They want to connect local councils in the UK who face similar challenges, such as a big local employer shutting down, so they can build on shared experiences. Ryan’s approach used cutting-edge statistics, but focused on helping his co-workers who could, in turn, help the people in local communities.
“We want to allow decision-makers to play with the data,” he told me, “to discover solutions themselves.” Ryan’s view was that we needed to combine algorithmic and human intelligence to solve our problems. Algorithms alone were not enough, but they were useful.
Our culture has produced forms of mathematical “artificial” intelligence for solving problems over thousands of years: from the early geometry of the Babylonians, through the development of calculus, through the hand calculator allowing us to perform faster arithmetic, to the modern computer and the connected society. We are getting smarter with the help of mathematical models, and models are improving because we develop them.
Bringing people together
On Sunday morning, a bunch of lads from Gateshead played football against a bunch of lads from Uppsala. It was a sunny day and a good game with teamwork and fair play, and happy parents cheering the boys on. Afterwards we all watched fireworks together at the local rugby club. All because an algorithm had suggested to Ryan and me that we follow each other on Twitter.
There are risks with the way we interact online. A lot of it isn’t very pretty, but there are such incredible possibilities to be had from working together with algorithms. And for now at least, it is us who control the algorithms and not them that are controlling us.
This is an edited excerpt from ‘Outnumbered: From Facebook and Google to Fake News and Filter-bubbles – The Algorithms That Control Our Lives’ by David Sumpter (£16.99, Bloomsbury Sigma)
