Informed AI is a new way of thinking about AI incorporating large data sets, establishing ground truth (through intelligent tagging) and leveraging real-world production data. All too often AI models are constructed with smaller data sets that have been improperly tagged which can introduce bias. By using informed AI, Jumio’s identity verification processes are faster, more accurate and contain less built-in bias.
Read How Jumio Reduces Demographic Bias
Informed AI is built right into Jumio’s online identity verification solutions so you can deliver a better, smarter, faster online identity verification experience.
Jumio’s AI is informed by hundreds of millions of domain specific data points.
Our data sets also have considerable depth in terms of the ID document types and subtypes — from UK driver’s licenses to Bahamian national ID cards to everything in between. This enables our AI models to support many different types of ID (more than 3,500) with high confidence and verification precision.
It’s not easy to establish ground truth in online identity.
Jumio’s AI algorithms are based on datasets tagged by the best verification experts in the industry. We've invested in human review (agents who tag our verifications to keep our algorithms honest) which allows our AI to get smarter, faster.
Jumio’s AI models have been developed from 10 years of real-world production data — not from academic data-sets. This helps eliminate some of the bias built into off-the-shelf data sets that many start-ups have used to kickstart their AI modeling.
Real data means better, more informed, bias-free AI which leads to faster, more accurate verifications.