I wonder if you were to train a single NN on multiple games (say maybe Mario, Commander Keen, Duke Nukem 2, Battle Toads, etc) if it would perform better agains

I wonder if you were to train a single NN on multiple games (say maybe Mario, Commander Keen, Duke Nukem 2, Battle Toads, etc) if it would perform better against any one game than say a NN trained against just the one same game? I know I have gotten better on any particular game when I have played more games of the genre. If that;s the case could we use this approach to make a General Game Playing AI.