How A.I. Learning Could Help Us Find Out How We Learn

Could we use machine learning as an example for our own brains learn?


“Machine Learning & Artificial Intelligence” by mikemacmarketing is licensed under CC BY 2.0

By: Cavanagh Keelin, Journalist

For the past few decades, scientists have been on a search to understand how we learn, and how machines learn. Recently, some scientists have started to believe that the two may be related. Namely, how modern methods of machine learning could finally dig up the last bit of the treasure that is how our brains learn. If proven, this information could allow us to find optimal methods of teaching, help solve problems like addiction, and aid us in creating better A.I.

The big question that researchers have been looking for an answer to is how the brain identifies what it needs to change to complete a task better. With A.I, one of the most popular systems of this credit assignment is called error backpropagation (or backprop for short) where if the actual result is different from the anticipated result, data about what went wrong is passed  backwards through layers of the neural system, modifying the system respectively. After backprop started many advancements in A.I image recognition, researchers revisited the idea of the brain doing something related.

It is still up to debate on whether error backpropagation is really the/a system that the brain uses to identify errors, but the answer to this long studied question is getting closer by the week.

Related Stories: