Drupal is a free and open source content management platform used for all sorts of websites, from personal blogs to NASA’s. During his opening keynote at DrupalCon Seattle 2019, Drupal Founder Dries Buytaert shared that 1 in 30 websites today are built on Drupal’s back-end framework.
Each year, the nonprofit Drupal Association hosts DrupalCon North America—an annual gathering of developers, UX designers, content managers, and more who use Drupal each day. I was pleased to speak at this year’s conference as part of the “Builder” track on a subject that influences all: artificial intelligence (AI) and the datasets used to train these products.
Today’s software teams integrate AI into a wider range of products than ever before. If you use cloud collaboration software such as Slack or Google Drive, you use the AI that’s built these products as well.
But AI’s benefits come with a warning: If the datasets used to train AI algorithms aren’t large or diverse enough, they risk perpetuating bias that will affect end users.
This isn’t an abstract concept: Judges in more than 12 U.S. states have used a machine learning (ML) algorithm called COMPAS to predict defendants’ likelihood of recidivism.
COMPAS’ results impact factors like the lengths of prison sentences and whether defendants are released on parole—even though the algorithm has incorrectly predicted that black defendants are more likely to recommit crimes.
Still think machine bias can’t hurt you? Consider what would happen if a speech recognition API system isn’t trained on a dataset that includes a wide range of accents and inflections. If that API is part of an autonomous car—and that car can’t recognize voice commands from a wide range of users—the end results could be deadly.
Now, the good news: Although machine bias is an unavoidable problem, it is not unmanageable. If you’re part of a product team tasked with building datasets used to train ML algorithms, you can start taking several steps to reduce the risk of bias.
Learn how by watching my DrupalCon presentation below. Then, let us know your thoughts in the comments and read more about how AI will impact small businesses across industries.