Brain-Inspired Chip Cures AI’s Memory Problem
One of the biggest obstacles preventing artificial intelligence (AI) from becoming fully autonomous is something researchers call "catastrophic forgetting." Train an AI model to perform a new task, and it often loses part or all of what it previously learned. That limitation makes it difficult to build edge AI systems that can adapt throughout their lifetime without repeatedly retraining in the cloud. Researchers at The University of Texas at San Antonio believe they have taken a significant step toward solving that problem with a new neuromorphic chip called Genesis.
Developed by the MATRIX AI Consortium and the Neuromorphic Artificial Intelligence Lab, Genesis is a spiking neuromorphic accelerator designed to continually learn on-device while preserving existing knowledge. Instead of repeatedly overwriting old information as new data arrives, the chip is designed to accumulate knowledge over time in a way that more closely resembles how the human brain learns.
The key idea...
Copyright of this story solely belongs to hackster.io. To see the full text click HERE
