AI That Thinks Just Like Us: Boffins Unveil Artificial Intelligence – Bestgamingpro

Researchers at Fujitsu and the MIT Center for Brains, Minds, and Machines (CBMM) have made a "crucial step" in enhancing AI model image recognition accuracy. The scientists, as reported in a new study presented at NeurIPS 2021, have developed an approach for computing that mimics the human brain to create AI capable of detecting information that isn't present in its training data (also known as out-of-distribution data or ODD). The nature of the data, as well as its size and characteristics, have a major impact on how AI technologies may be used. Despite their current application in a variety of domains (e.g., medical x-rays analysis), existing models are highly sensitive to context. The importance of AI capable of recognizing ODD is that accuracy is preserved in conditions that are not ideal, such as when the perspective or light level differs from those used to train the model.