It's a cute achievement, but the only reason Microsoft could set the Atari Ms Pac Man record is because no one else cared enough about that game to write an AI for it.
"Doina Precup, an associate professor of computer science at McGill University in Montreal said that’s a significant achievement among AI researchers, who have been using various videogames to test their systems but have found Ms. Pac-Man among the most difficult to crack.
But Precup said she was impressed not just with what the researchers achieved but with how they achieved it. To get the high score, the team divided the large problem of mastering Ms. Pac-Man into small pieces, which they then distributed among AI agents."
Ms PacMan is difficult for current general-purpose models. Nonetheless, it's accurate to say that other groups haven't focused on developing a model or method specifically tailored to the game. The result in the linked article is from a model engineered expressly for Ms PacMan, which makes comparison with previous work from the deep RL community a bit misleading.
https://blogs.microsoft.com/next/2017/06/14/divide-conquer-m...
"Doina Precup, an associate professor of computer science at McGill University in Montreal said that’s a significant achievement among AI researchers, who have been using various videogames to test their systems but have found Ms. Pac-Man among the most difficult to crack.
But Precup said she was impressed not just with what the researchers achieved but with how they achieved it. To get the high score, the team divided the large problem of mastering Ms. Pac-Man into small pieces, which they then distributed among AI agents."