Photo by Marek Piwnicki on Pexels

MIT researchers have unveiled a groundbreaking technique enabling large language models to learn incrementally and retain existing knowledge. This breakthrough addresses the persistent challenge of “catastrophic forgetting,” where AI models lose previously learned information when trained on new data. The new system enables continuous improvement and adaptation, paving the way for more versatile and intelligent AI applications that can constantly evolve and acquire new skills throughout their lifespan.