The end goal is something capable of human-level intelligence in most, if not all, fields, and from there, it's not that difficult to continue scaling it up across speed / scale (and perhaps intelligence.)

See Dario Amodei's Machines of Loving Grace: https://darioamodei.com/machines-of-loving-grace

It doesn't take much imagination to think about how this could end poorly if it goes wrong or is controlled by the wrong party.