Sign In

Communications of the ACM

ACM TechNews

AI Is Learning From Humans. Many Humans.


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
iMerit employees in Kolkata, India, label images used to teach artificial intelligence systems.

Artificial intelligence systems must learn from troves of data labelled by humans before they can be used for their intended purposes.

Credit: Rebecca Conway/The New York Times

While artificial intelligence (AI) systems are viewed by most people in the technology field as the future of the industry, tech executives rarely discuss the labor-intensive process that goes into creating these new programs.

AI systems must learn from human workers labeling troves of data before they can effectively be used for their intended purposes.

Tens of thousands of workers spread around the world, many of them independent contractors working from home, annotate data through crowdsourcing services like Amazon Mechanical Turk.

The market for data labeling, which accounts for about 80% of the time spent building AI systems, passed $500 million in 2018, and will reach $1.2 billion by 2023, according to the research firm Cognilytica.

While AI researchers hope they can create system that can learn from smaller amounts of data, human labor will be essential to the industry for the foreseeable future.

From The New York Times
View Full Article

 

Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA


 

No entries found