Are there games between top computers which uncovered never-seen-before ideas?

Question

Are there games between top computers which uncovered never-seen-before ideas?

And if so, how can humans notice?

Addendum: I understand the position of James Christopher in his answer, but my question goes a little beyond that. It is possible for the computer to find a winning move without understanding the idea behind it, and thus to win the game. If afterwards a human notices that the move discloses a new theme or idea never before seen, that would qualify as a positive answer.

Addenddendum: We have a linguistic problem here. Let me clarify: The computer does not need to have a new idea or even understand what it is doing. But it can find a move that wins the game, and later, when people analyze what happened, they realize "oh! this move wins because of such-and-such!" The computer does not need to know that it forked the king and queen, it just evaluates the material gain and decides that the knight move is best. We call it a fork and add a section in the basic tactics chapter.

Answer 1

My answer is - Yes. A simple example is the use of endgame tablebase. Many positions, which were thought to be drawn, now have a winning line. IMO this is a new idea. One day computers might uncover that castling is bad - it is a new idea, doesn't matter if it was achieved by brute force.

ps: even though computers work on fixed algorithms, they can uncover new ideas. Same algorithm can return different results when run for longer duration and/or better hardware. In the end, chess is deterministic and finite.

Answer 2

I am surprised that a more up-to-date answer is not added here yet. Endgame tables is one area where major improvements happened (see Nishanth's answer). But, I think the opening is the main area revolutionized by the AI wave. By the way, these are not the only areas of remarkable improvement. In fact, some engine games are now featured in the list of games that revolutionized chess (e.g. see the article 7 Games That Transformed Chess). According to the Authors of Game Changer, the book on AlphaZero, the engine gives more importance to piece mobility and king safety, compared to non-AI engines or even most humans. Studies on what factors these advanced engines give more importance to could revolutionize chess theory as a whole. After all, ascertaining the importance of chess themes in a way different from the traditional wisdom is exactly how many movements in chess theory such as hypermodern school were born.

Opening is one area where AI wave made remarkable improvements. AI-based engines have introduced several opening ideas that challenge conventional wisdom, and some that use the known ideas in novel ways.

Quote from a chess.com article:

Before the advent of engines, opening theory was largely developed through human trial and error over decades of play. Grandmasters would experiment with different lines and ideas, and over time, certain lines would rise to prominence while others were discarded.

Today, engines have accelerated this process exponentially. Top players use engines to test out new ideas, find improvements in well-known lines, and discover new theoretical novelties that could catch their opponents off guard.

Specific examples for this include those mentioned in the following quote from the article How Chess Engines Have Shaped Grandmaster Play?

Chess engines have revolutionized the study of openings, enabling grandmasters to delve deeper into established lines and discover entirely new variations. For instance, the Ruy Lopez, a classic opening, has seen an explosion of new ideas thanks to engine analysis. Grandmasters now use engines to refine and innovate openings, as well as to uncover hidden resources. An example of this can be seen in the World Chess Championship match between Magnus Carlsen and Fabiano Caruana in 2018. In Game 1, Carlsen’s novelty with 6.d3, discovered with engine assistance, caught Caruana off guard and set the tone for the entire match.

A quote from another chess.com article.

AlphaZero's style was deeply counterintuitive for a human and a bit closer to the Romantics or Tal than most chess players would probably have expected. "AlphaZero appeared to target the opponent's king directly from the opening moves and sacrificed material freely to achieve its goals," wrote GM Matthew Sadler and WIM Natasha Regan in their book on AlphaZero, Game Changer.

A specific example is featured in the article Stockfish On The Classics: Kasparov's Gambit.

For more specific examples and details of what we learned from AI engines, see videos in the Learn from the best series in chess.com. Some compilations from these series: Leela, Stockfish. The videos in the stockfish compilation specifically feature opening ideas.

Answer 3

Computers are based on algorithms written by humans, which are based on knowledge and concepts known to humans. They do not work based on abstract chess ideas. Computers are not known for creativity, so I would say that the answer to your question is no.

They will certainly find moves in a position that GMs have not thought of before, but that is not the same as a new chess idea.