Less Than One - A radical new technique lets AI learn with practically no data

I came across with this very interesting new idea about training AI with a substantial small data set.

Since we humans sometimes don't even need to see one example of something to recognize it, why should AI be fed with tons of data to identify something?

Here's the 'Less Than One'-Shot Learning: Learning N Classes From M<N Samples original paper.

The A radical new technique lets AI learn with practically no data article explains it all.

./M6