Facebook has announced some advances in AI that help detect hate speech and enforce community standards on the platform.
Facebook reported that the prevalence of hate speech declined for the third consecutive quarter in the second quarter of 2021, reflecting improvements in the detection of hate speech on Facebook’s News Feed.
With new techniques and tools such as Facebook’s Reinforcement Integrity Optimizer (RIO), the social media powerhouse has been able to curb the amount of hate speech and other unwanted content Facebook users see.
The classic approach to training AI uses a fixed set of data to train a model that is then used to decide on new content. RIO instead cites an AI model to learn directly from millions of current content, while constantly evaluating, learning, and adapting to changes that better ensure Facebook’s security over time.
For more information, read the original story in ZDNet.