Reliable AI: Grounding to Mitigate Hallucinations

Artificial Intelligence (AI) is on course to transform society. Recently, attention has focused on generative models and neural networks trained on internet-scale text data. These models are powerful, with some researchers describing them as approaching human-level intelligence. But if an AI system answers a question—about a real-time situation, about available intelligence, or even about a single document—how can the user validate the credibility of the answer?

Primer’s Senior Director of Science tackles this pervasive problem with a tangible solution: grounding.