I think Stanford doesn't really know how to do real math. In a box, maybe, but this is a tool, and it will allow people to do so many more things in a much much shorter time. This will definitely help advance technology much much faster. On the flip side, like any tool, it can and will become very dangerous. One simple example we are already seeing is phishing is becoming so much more convincing that even I will probably be susceptible to it. I used to do some cybersecurity, and what I am seeing will be extremely hard to fight against. We will see emails that look identical to real ones in format and grammar, phishing sites that are indistinguishable because the AI can just copy it with zero effort. Phone calls can now sound almost identical to a target. I'm not sure if we will be fast enough in the cat and mouse game before the bad guys win.