The difference is that it takes few minutes to an hour at most so it can be run multiple times a day, using the results of previous runs to further refine the search and reasoning process to get better outcomes. Pretty much how any human research works but much faster and with potentially vastly more world-knowledge and reasoning capability than average humans. And these capabilities will rapidly improve with further RL.
No comments yet.