Skip to Content
MIT Technology Review

Human labor is propping up some companies’ fake AI software

There isn’t always a computer program behind “AI” services—sometimes it’s just plain old “I.”

Some background: Effective artificial-intelligence programs can require gobs of data, time, and money to produce. That means a lot of up-front investment before achieving a minimum viable product.

The cheap solution? Humans. Some companies have decided to forgo algorithms altogether and use people to power programs that are advertised as AI.

For example: Last year, the expense reporting app Expensify posted images of receipts on Amazon’s Mechanical Turk, and paid gig workers to manually write up the data contained on the images (the company claimed the process was done using its own SmartScan software). Other companies have enlisted people to pretend to be chatbots.

Why it matters: While it can give businesses a way to test out a new idea before committing engineering resources, it’s not a great way to build trust with customers. Alison Darcy, founder of a mental health support chatbot, told the Guardian, “There’s already major fear around AI and it’s not really helping the conversation when there’s a lack of transparency.”