For writing the next column, I am doing a simple simulation experiment.

"Destructive experiment of deeply learned neural network".

On the other hand, I can say,

"Destruction work from the human side to AI technology"

However, what I am doing is easy as stupid.

-----

Actually,

"How robust a neural network has against external attacks or self-destruction"

or

"How much can I stop a neural network if I break it.

I wanted to know the result approximately.

I tried searching for papers about the above trial, but I could not find it.

Actually, I also try it by myself.

-----

I plan to disclose the results as the final round of "nursing care" of this month's "work way reform" series.

This column will also be "extremely uncomfortable" content.