Several weeks after the death of one of its bizarre Twitter stars, Horse_ebooks, the Internet has given birth to a new Twitter meme -- tofu_product. Like Horse_ebooks did, the feed spews out random and choppy strings of texts that make just enough sense to register as sentences and thoughts, but nonsensical enough to appear less than human. Unlike Horse_ebooks, tofu_product is not an elaborate performance-art piece curated by BuzzFeed employees, but something much, much more -- a program meant to distort reality. The New Yorker describes how the algorithm, created by programmer Joe Toscano, works:
The program is called @tofu_product. When you follow him—users, including Toscano, have begun to refer to the bot as male—Tofu reads through two hundred or so of your most recent tweets. When you tweet at Tofu, the algorithm sifts through that set of messages to derive a reply intended to mimic your own tweets. And after you’ve followed Tofu, he continues to collect words from your new tweets, creating an ever-expanding personalized dictionary. Tofu the bot, like tofu the food, absorbs the flavor you give it.
The program is like a fun-house mirror, except that the more a user tweets at it, the more accurate the reflection becomes. But, as a quick scroll through the Twitter account demonstrates, the tweets are always a little bit "off." Says Toscano, “there is still such a human quality to writing when a human is writing. There are some bots that can do a pretty good impersonation of a human, but it’s hard to be convincing. Twitter is short and to the point and, in some ways, robotic, but you can still kind of tell when a human is tweeting.”
The account has been around for about a week with more than 3,000 having discovered it, including Slate. Here are some of the account's best tweets so far: