Catastrophic forgetting (or catastrophic interference) is a long-standing problem with neural networks, the theoretical underpinnings for deep learning and other machine learning techniques. Here’s the problem: say you teach an AI app to recognize an apple, and then you show it a pear. It will forget how to recognize an apple.
Heretofore, the only solution to this catastrophic forgetting problem is to repeatedly show the application both an apple and a pear (as well as everything else you want it to recognize) – a time-consuming, data-intensive task.
Neurala has solved the catastrophic forgetting problem via careful analysis and modeling of human learning. The result: machine learning that requires far less compute power and data than traditional approaches, making it possible to put AI on small, inexpensive hardware platforms.
First out of the gate: AI-driven drones (both for consumer and professional use), toys, and interest from vehicle manufacturers because of the potential to support self-driving cars. The big win for Neurala, however, promises to be the Internet of Things (IoT), as it can drive AI into the edge devices that make up the IoT.
Copyright © Intellyx LLC. Intellyx publishes the Agile Digital Transformation Roadmap poster, advises companies on their digital transformation initiatives, and helps vendors communicate their agility stories. As of the time of writing, none of the organizations mentioned in this article are Intellyx customers. To be considered for a Brain Candy article, email us at firstname.lastname@example.org.