When Christopher Steiner, the 35-year-old cofounder of Aisle50, a Y Combinator startup offering online grocery deals, set out to write the book Automate This: How Algorithms Came to Rule Our World, (out tomorrow), he’d planned to focus solely on Wall Street. "There were a ton of good stories and then the Flash Crash happened. There was a lot to tell," says Steiner. "But at some point I thought, ‘Do people really care about the 13 different electronic training networks that were going on in the 1990’s?’" Instead, the former technology journalist expanded his research to explore how the power of algorithms has spread far beyond Wall Street and now touches all of us—starting with today’s young innovators.
Fast Company: You say that those who design algorithms are "the preeminent entrepreneurs of this generation."
Christopher Steiner: If you look at who has the biggest opportunity in society right now, who’s the most upwardly mobile and could just build something out of nothing, it’s developers. It’s people who are able to write code, not just any code—you’re average developer’s going to make a nice salary—but if that person innovates with code that "solves a problem" the opportunities are huge. This is why places like Y Combinator and all these other seed funding enterprises are booming. If you have the skills, the startup costs are basically zero. It’s your own time. This is not rocket science to use that old cliché; it’s a skill you get by putting time into the medium. It’s not something you need to learn at MIT.
So what does this suggest for those entering universities and the job market these days?
I’m not Peter Thiel; I’m not going to tell you not to go to college, but there are certainly many interesting options for those who are willing to take a bit of a beating and learn quantitative concepts. Frankly, once you build up a little base of knowledge, it’s very easy to get a couple of jobs. The people who are in the sweet spot of the job market right now have two to three years experience in the newest coding languages.
You title one of the later chapters of your book "Wall Street Versus Silicon Valley" Explain.
It’s incredible; people don’t realize how many software engineers Wall Street takes off the market. And in the past, when Silicon Valley companies went head to head with Wall Street firms, it was very hard to compete for the best engineers because the salary packages were so dissimilar, including the bonuses. And there was a prestige in working for a company like Goldman Sachs. So, I’ll just say, luckily for the economy, some of that prestige has worn off. And I think that’s better all in all because the utility that someone with that kind of skill brings to the economy when they go to a place like Morgan is minimal—or even negative in the worst cases. Whereas if they go to a startup, they’re actually building the economy. They’re building GDP by affecting the most dynamic and growing segments of our economy. At Wall Street, they’re just moving money around.
It wasn’t always like that.
Yes, there were good things that happened within the algorithmic trading industry 10, fifteen years ago that changed the game for all of us. Like letting you and me trade stocks from our house for six bucks, which gave the normal guy liquidity, essentially democratizing the markets. But we’ve long since passed that type of utility. Now we are just moving around decimal points and in fact, building up the risk profile for everyone else. This is what we saw with the Flash Crash and at Knight Capital a few weeks back.
What went wrong?
It’s a software problem. Anytime you have so many layers of software—algorithms, really—nobody knows how one new layer will affect the other layers. That’s why good programmers and algorithm writers will create tests as they work. Google, for example, tests their algorithms a hundred million times before they ever hit the market. But at Knight, they wrote the program and sent it out. Apparently there were no tests because their algorithm just went bananas from the moment it was turned on. The danger with Wall Street, is that the whole thing’s so focused now on speed that there’s no time to write tests. This stuff literally happens every two weeks. Usually it’s not a $440 million loss, but there’s just so much risk built into the market right now that doesn’t have to be there.
And algorithms are doing a lot more than automating stock trades.
Most people don’t know that there are algorithms that decide how customer service calls get routed or how customer service requests will be treated. When people call these big companies like their health insurer or telecom company, they’re actually being categorized, sliced, diced and parsed by a bot. It’s incredible to think that the words someone chooses on a given morning will forever change how that company treats him or her. These algorithms don’t just affect people involved in computer science.
I was intrigued by your discussion of Jon Kleinberg, a Cornell computer science professor who devised an algorithm to identify the influencers in a given organization.
He was the guy who came up with the original method that Google eventually used to create their PageRank algorithm. His newest algorithm ranks people and their place in society by how they affect others through language. For example, if, in any given group there’s one guy who influences the others more strongly than anyone else, he tends to be the leader. This can be measured quantitatively. The schematic of how this works looks just like the schematic of how web pages are ranked. Whoever is linked and has more power over all of these trusted sites is who ends up at the top of the Google rankings. Same for people.
Much of this is still in the experimental phase but how might this play out?
The interesting thing is that at some point you’ll be able to drop this into applications like Facebook or Twitter and rank people by their command over others. Suddenly, the things we kind of know, like that this person is the leader of the pack, will be out there for everyone to see. That’s a little bizarre. That would make people uncomfortable, but the ability is there for these short lines of code—maybe 1,000 lines or so—to decide who the alpha dog is. Of course, people will learn how to game it. At some point, employees will write emails a certain way to game the algorithm and then the algorithm will change because they know they’re getting gamed, and it’ll change again. It’ll be this constant war of attrition of blogs talking about how to defeat your company’s algorithm-ranking system, and this is totally going to happen.
[Image: Flickr user Entrer dans le rêve]