Think of the endless possibilities for silicon based cheating. It is going to be a challenge for the anti-cheating committees to compare games and catch someone who is using a new algorithm based on AlphaZero. The unique and paradoxical moves may not be comparable to known move selection by present programs. If the chips are small enough, virtually anything could become the device that can be used to generate moves quietly, stealthily, and effectively.
Have looked at a few of the games. Typical computer style games. Really bad, even weird play by Stockfish. Some of the moves being gushed over are unusual, but not so out of the norm that a human player would not consider them. I know a couple of players who would have selected these strange moves to study first. It is notable that the conditions that Stockfish played under are limited. Even so, the fact that AlphaZero studies so few positions, even given the potential computing power behind them, means that the developers approached the problem of move selection from a software perspective rather than the usual hardware intensive, speed kills, fast and dirty approach favored by the military. AlphaZero appears to be extremely dynamic, even treating static positions differently than other programs. It is less reactive. Like humans, it actually tries to do something.
From what I understand, neither AlphaZero or Stockfish were using a book. Invariable at the level they were playing, every game would have been some known opening, but every major chess engine out there is tweaked to use an opening book, even if it’s just 3 moves.
Although there is no way to know if Stockfish, using a book, would have won more games, I’d think a highly tuned “strong opening book” in Stockfish would at least given it more draws. I’d actually be more interested in a AlphaZero vs Komodo with an opening book.
At the level of both Stockfish and Komoda (and other very strong engines), there is some hair splitting on the list of which program is stronger. In any event, from what I’ve read, Komodo is heavier on evaluations and less focused on pure calculating depth. Maybe that would do better againt AlphaZero.
To me, these super chess-playing computers are just an intellectual curiosity. I’m not worried that they will “destroy” chess. I do get depressed sometimes because I keep losing to my little Radio Shack 1650 chess computer (“endorsed by Garry Kasparov”), but then I just go back to my chessboard and play through some beautiful Mikhail Tal games. If only I could achieve some of the lovely positions I see in the games I study!