Silicon Valley-based AI startup Hive has announced its plans to pay more than 600,000 people to help it label photos for its image recognition AI system. The workers, who are being paid pennies per image, will help the company create an AI the firm's CEO Kevin Guo claimed would be "whole orders of magnitude better than Google".
AI models are notoriously hard to train and can take a significant amount of human labor performing very mundane tasks such as labeling or classifying images. Instead of trying to work around this issue, Hive has decided to embrace it and gamified the act of training an AI system with Hive Work, a smartphone app and website people can sign up to and get paid to assist Hive train its AI.
Hive trains systems for businesses such as NASCAR, whereby it uses AI-powered machine vision to recognize and track the screen time of corporate logos. In order to do this and compete with tech giants such as Google, Hive incentivizing the general public to sign up to Hive Work and help it categorize, 'box bound' (whereby you draw a box around an image) and text transcribe images.
While the pay is not significant per image, Guo claimed it could add up to "tens of dollars" relatively quickly. And as it is just a smartphone app, people can "play" it whenever they are free and slowly accumulate money – something Guo says few other apps offer.
"What's the alternative? Playing Candy Crush Saga and losing money," Guo asked.