A trio of researchers from the University of California, working with a colleague from the Institute of Computer Science of the Czech Academy of Sciences, has found that it is possible to prevent catastrophic forgetting in AI systems by having such systems mimic human REM sleep.