Two Steps Down a Squared Bracket

A one-parameter model is trained on the loss

$f(x) = (2x - 3)^2 + 5$

using the gradient descent update rule

$x \leftarrow x - \alpha\, f'(x)$

Training starts at $x_0 = 5$ with learning rate $\alpha = 0.05$. Each step uses the slope measured at the position the algorithm is standing on at that moment.

Exactly two steps are taken. What is the value of $x$ afterwards?

Round your answer to 2 decimal places.