Chapter 14: Why should the updated value of AC be AC-(A0 * Cdelta)? (Page 367)

[loongluangkhot]

In Chapter 14: Backpropagation, page 366-367, the author set up a scenario where the change in network error and change in AC can be described by the equation $y=(A_0 \times C\delta)x$, where y = change in network error, and x = change in AC (Figure 14-13).

If we update $AC$ to $AC - (A_0 \times C\delta)$, i.e. $x=- (A_0 \times C\delta)$, the expected change in network error will be as follow: $y=(A_0 \times C\delta)x=(A_0 \times C\delta)[-(A_0 \times C\delta)]=-(A_0 \times C\delta)^2$

I don't understand why we are trying to change the error by $-(A_0 \times C\delta)^2$?

P.S. Not sure if this is the right forum to ask this. If it's not, would appreciate suggestion of where I can clarify this.