Amy Ingram, the artificial intelligence personal assistant from startup X.ai, sounds remarkably like a real person. The company designed her to take on the mundane tasks of scheduling meetings and e-mailing about appointments. If a bot had access to your calendar and was cc-ed on correspondence, why couldn’t it do the work for you? After she made her debut in 2014, users praised her “humanlike tone” and “eloquent manners.” “Actually better than a human for this task,” a beta tester tweeted. But what most people don't realize about this artificial intelligence is that it isn't totally artificial: Behind almost every e-mail is an actual human—someone like 24-year-old Willie Calvin.
Calvin, who worked as an AI trainer for X.ai before he said he quit in October, was part of the reason Amy never tripped up, sending the sort of blind response that reveals she’s a bot. The company advertises Amy as an AI personal assistant who can “magically schedule meetings,” and its software does scan e-mails and can usually guess that “tomorrow" means Tuesday. But the system isn’t yet ready to take the next step on its own. Multiple former AI trainers said that as recently as a few months ago, trainers looked over parts of almost all incoming e-mails — to evaluate what Amy guessed the user was saying— before Amy generated an auto response. A company spokeswoman said the service still has trainers verify “the vast majority” of information in e-mails so the system can improve.