Gradient Descent: Walking Downhill Blindfolded

In the last post, we built a way to measure exactly how wrong a guess is, using the cost function. Given any w and b, we can now compute a single number that tells us how badly that line fits our data