Instagram’s CEO Wants to Clean Up the Internet—But Is That a Good @&#$ing Idea? | WIRED

Sometimes using a machine can feel like the purest way to solve a problem. Humans are boiling stews of biases and contradictions, and computers don’t have emotions. But machines are only as good as the rules built into them. Earlier this year, the chief scientist of the text analytics company Luminoso, Rob Speer, built an algorithm based on word embeddings to try to understand the sentiment of text posts. He applied the algorithm to restaurant reviews and found, oddly, that Mexican restaurants all seemed to do poorly. Stumped, he dug into the data. People like Mexican food, but the system suggested they didn’t. Ultimately, he found the culprit: “The reason was that the system had learned the word ‘Mexican’ from reading the Web,” he wrote. And on the internet, the word Mexican is often associated with the word illegal, which, to the algorithm, meant something bad.