3 ms·
The interesting part of this story for me is that computer proficiency on language tasks is improving so rapidly that a benchmark introduced just last year has
by skybrian 7y ago
The interesting part of this story for me is that computer proficiency on language tasks is improving so rapidly that a benchmark introduced just last year has become too easy. It seems like there is little reason to think this research will hit a wall any time soon? Will we have computers acing SAT's and writing college essays?
It hardly matters which tech giant paid for it this time, since they are publishing the results and the others will build from there. Next year it might be Google's turn again?
Here is the part I find scary: consider that they're publishing the models and the code, so anyone could use it. What might malware authors or nation-state actors build? (Sometimes working together.)