Presumably the reason for changing the rules is to limit the number of draws, because wins and losses are more exciting. As he mentions, high level, full length chess has a rather disappointing percent of games ending in a draw. However, he doesn't even show that alphazero produces less draws in this variant (even the example game is a draw through repetition!). It may be interesting because it's different, but I don't really think Kramnik successfully argues that it's any better.