Another ML metaphor is this: don't just start from one point and do gradient descent. Try starting at many different points and working your way through the des

Another ML metaphor is this: don't just start from one point and do gradient descent. Try starting at many different points and working your way through the descent logic. It may be more trial and error and more dead ends, but you might indeed find your global minimum. To speak like a human: try new things and don't give up on them immediately. Another ML metaphor: I wonder if the concept of an identity crisis is a way to dislodge yourself from a local minimum? You know something better is out there, but you haven't done enough exploration so eventually you rebel and give up your comfortable little basin.