You may be familiar with the tech news aggregator TechMeme, a site that pulls together clusters of breaking stories in the hi-tech world - like Google News on steroids. (If you're still looking for a primer, Jack profiled it - along with sister site Memeorandum - earlier this week).
Yesterday its creator, Gabe Rivera, shocked some followers by announcing that he was going to hire an editor for the site to help improve its coverage. "Guess what?" he wrote, "Automated news doesn't quite work". The reaction was fairly mixed, but some true believers weren't happy (TechCrunch's Michael Arrington called it a slippery slope that undermines objectivity, for example).
While Rivera was evasive when I asked him whether more direct human input would feed back into improving the algorithm, the question remains: since algorithms are designed by people, can any of them really be classed as objective anyway?
Fortunately I had the good fortune to be speaking to Jon Kleinberg, a computer science professor at Cornell and one of the world's leading experts on internet algorithms today - so I asked him about the relationship between man and machine.
Here's what he had to say:
"One interesting thing about designing algos in this world is that the world reacts to what you do," he told me. "Once people know they're being evaluated by an algorithm, they always try to reverse-engineer it. There's always this feedback loop which makes things very interesting."
"At the simplest level, it is worth pointing out that when you read Google News there isn't a human editor deciding what to show… it's the result of an algorithm. But the distinction blurs around the edges in many ways; the simplest being that these algorithms are designed by human beings who can change them and do change them regularly - and change them with an eye towards achieving certain, desired results."
"You see lots of systems where there's a kind of hybrid human and algorithmic process that's producing what you see," he said. "There are cases where humans and computers have complimentary strengths and can work together."
No comments:
Post a Comment