Posted inBusiness

Meta’s new Few-Shot Learner AI system to address harmful online content

Meta’s AI unit creates Few-Shot Learner system to address harmful content on its platforms.

Few-Shot Learner
Few-Shot Learner AI system addresses the evolution of harmful content.

Meta’s AI unit has announced the creation of a new AI system to quickly take action against new or evolving types of harmful content on their platforms.

Meta’s AI unit have announced the creation of a Few-Shot Learner, an Artificial Intelligence (AI) technology that adapts to take action against new or evolving types of harmful content within weeks instead of months.

Faster, with fewer examples

The new AI system uses a method called few-shot learning, in which models start with a general understanding of many different topics and then use much fewer, even zero, labeled examples to learn new tasks. The Few-Shot Learner can be used in more than 100 languages and learns from different kinds of data, including images and text. The new technology will help augment existing methods of addressing harmful content.

The new system works across three different scenarios:

·       Zero-shot: Policy descriptions with no examples

·       Few-shot with demonstration:Policy descriptions with a small set of examples, less than 50

·       Low-shot with fine-tuning: Developers can fine-tune the FSL base model with a low number of training examples

Successful trials

The Few-Shot Learner has been tested on recent events, including content that gives misleading or sensationalized information discouraging COVID-19 vaccinations. In another example, the new AI system improved an existing classifier that flags content that comes close to inciting violence (for example, “Does that guy need all of his teeth?”).

The traditional to screening for harmful content may have missed these types of inflammatory posts as there not many labeled examples that use “DNA changer” to create vaccine hesitancy or references to teeth to imply violence.

In combination with existing classifiers along with efforts to reduce harmful content, ongoing improvements in Meta’s technology and changes made to reduce problematic content in News Feed, FSL has helped reduce the prevalence of other harmful content like hate speech.

For more information on Meta’s AI Few-Shot Learner, click here.

Social media company Facebook was rebranded to Meta in late October.