After reading a research paper posted by the University of California, San Diego, there is a significant risk of a program using machine learning or AI. The program can “forget” what the programmer is trying to teach. A researcher from UCSD decided to take a more human approach and treat the program as if it was a human brain.
A human is reported to sleep 7–13 hours per 24 hours to build a rational memory. A rational memory is defined as one that can recall random associations between memories, people, events or objects.
Artificial Intelligence or Machine Learning contains neural networks that contain an architecture very similar to a brain and even though the 2 can be classified as quite similar, computer scientists continue to make the mistake of treating it to learn like a mindless robot. However, it is not mindless and cannot learn sequentially as one would hope. One needs to teach it to learn as a human instead of a numbered list of events. Or else, it can overwrite its past information with new information, also considered a phenomenon of ‘catastrophic forgetting.
After UCSD put their theory to the test, they found that they were right, and that “these networks could learn continuously like humans … Understanding how human brain processes information during sleep can help to augment memory in human subjects. Augmenting sleep rythms can lead to better memory”.
Maybe we should try to treat AI to be more like humans as we are not so different after all…
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1010628
