Have you tried playing chess with chatgpt? It suggested illegal moves eventually, with seemingly reasonable "motives". (Not suggesting beating chess as a goalpost) A tool is good if it follows intention. But a tool out of control can be messy.

I use ChatGPT daily for programming help (love it!) but the people saying that machines like this will outsmart us imminently make me wonder if they've even tried it or they're just going by headlines. It goes from genius to full-retard at the snap of a finger, quite frequently so, and as many have already pointed out, appears totally confident in its mistakes. Some are saying this last 10% will just be a matter of polish, but my impression is that, like the old software adage goes, the last 10% will be 90% of the work.

To me it's shocking how far it's come in such a short time, that I worry if the last 10% will happen quickly.

>Some are saying this last 10% will just be a matter of polish, but my impression is that, like the old software adage goes, the last 10% will be 90% of the work.

Even if that last 10% takes three years, that's still a terrifying prospect and a breakneck pace.

wasnt it super slow?

yes