The Latest Lesson Chess has for Knowledge Work is About AI

Since the pandemic began, almost 13 million people joined chess.com. This tidal wave of interest was spurred mostly by the surplus of free time the world had on its hands and also partially by the chess-themed Netflix show The Queen’s Gambit, named after a famous opening sequence of moves in the game. And while Chess’s popularity may have only recently been rejuvenated, it has been a ubiquitous metaphor for war and business throughout the ages.

As explained in the below video, Chess may have originally created to represent a reenactment of battle. Unsurprisingly, popular Chess principles like “Every move should have a purpose,” “Know your opponent,” and “Make strategic sacrifices” have frequently been reflected in military and corporate strategy for centuries. George Washington is documented playing chess in uniform and chess strategies are visible in the patience demonstrated by Japanese car manufacturers in the 1970s.

One of chess’s latest parallels to industry, and one that extends beyond gameplay, is centered on the emergence of Artificial Intelligence. While AI’s incredible advancements have taken the game of chess by storm, it has failed to “solve” the game and render human intervention in a chess match entirely redundant. This is a testament to the astounding complexity of the game, and also a critical lesson for businesses, militaries, and all types of enterprises to take note of. Looking forward, AI will be a powerful complement to, and not a replacement for, human thinking.

Recognizing the power of AI through playing chess

Like the aforementioned millions of others, I dusted off my chess skills at the start of the pandemic last year. My grandfather had taught me the rules and some basics when I was nine years old, but since then I had hardly played at all. So when my roommate and I instated a weekly Friday night chess match into our quarantine rituals, I quickly realized how astonishingly bad I was compared to my friend.

The first night we played, I completely embarrassed myself, losing within the first handful of moves. While I handled the loss with outward grace, I privately struggled a lot with this outcome. I wanted to be more prepared for the next Friday night, so I created an account on chess.com and proceeded to establish a mild addiction to the game. For those who are unfamiliar, chess.com has an elegant and robust interface that makes it simple to get into the weeds of the game and learn, whether you are playing against the computer or another player somewhere out there in the world.

Screenshot from a match against the computer on chess.com

Chess.com’s Coaching Mode allows you dissect every single move of a match, and produces a “grade” for each move you make. Then, you can go back and compare the move you made with the “optimal” move calculated by the computer’s algorithm and learn from your mistakes. Playing this way is an incredible way to learn the game, and fast too. Through this type of regular practice, I quickly became a much more formidable opponent for my roommate. By the time it was the summer, about three months after we started this routine, I was beginning to win with some level of consistency.

And now more than twelve months later, as an #ISYS8621 student, I’m realizing that this learning curve that I climbed my way through with tremendous levels of effort and concentration, is something that a deep learning algorithm can probably accomplish in a few minutes if it is fed with sufficient data.

How AI is enhancing, not replacing, human thought in chess strategy and beyond

AlphaZero is one of the most advanced chess playing artificial intelligence systems in existence. It taught itself chess starting with no knowledge beyond the rules, and developed its own chess strategies by playing millions of games against itself. In the process, it has unlocked new and fresh ways to think about the game that haven’t been previously considered.

For Matthew Sadler and Natasha Regan,  a couple of very advanced players who co-contributed an article to Wired, having AlphaZero at their side felt like “having a human chess genius on tap, who never got tired and never asked for coffee.” The article continued to explain:

AlphaZero’s strength and originality truly surprised us. Chess is full of superhuman expert systems, yet AlphaZero discovered an uncharted space in which its self-taught insights were both startling and valuable.

And yet, it remains impossible for any machine to fully “master” the game of chess and render human thought redundant. As the very entertaining video above explains, with 10^120 possible gameplay configurations in a chess game, there is yet to be a computer built with the depth to take all of those potential moves into consideration. The video also makes an illuminating connection to the number of atoms in the universe, which is a mere 10^75 in comparison.

With its limitations, AI is not the ultimate authority on the game, but it is a one-of-a-kind training partner that provides an enormous leg up for any player who is tapping into it. And industries driven by similar types of knowledge work would benefit from viewing AI through the same lens.

As professor of AI at The University of South Wales Toby Walsh explains, robots can help humans play to our strengths.

“Amplifying intelligence is a great way for AI to help us rather than replace us,” he explains. “So computers should bring to the table their memory and their brute force ability to search through many possibilities. And we should bring to the table our creativity, our emotional intelligence, and our ability to cope with new and unplanned situations.”

17 comments

  1. Great post!! Thanks for explaining why AI cannot fully “master” the game of chess. It reminds me of my bad experience in the computer algorithms class back in my college years ago. I still remember one day professor told us why computer cannot solve some problems as the number of data set increases, the computational power needed is increased in exponential, and we call these types of problems an np-problem https://en.wikipedia.org/wiki/NP-hardness, if using big-O notation, it would be O(x^n). https://en.wikipedia.org/wiki/Big_O_notation

    One of my school projects years ago was to create a robot in an Othello game. What we basically did was to create a list of possible paths, using a suitable tree, toward whether to win or lose the game for every move a human player made and then can calculate the success rate of placing another chess piece leading to winning the game. It is not an O(x^n) game so we are able to find all possible paths and place the chess piece that maximizes the possibility of winning the game.

    https://www.researchgate.net/figure/An-example-of-a-suitable-tree-for-an-Othello-end-game-position-This-game-tree-has-a_fig2_312618160

  2. Andrae Allen · ·

    Hey Scott, overall, this was an excellent read. It’s late, so I’m just gonna blurt out some thoughts.

    At my previous place of employment, I started an interoffice chess group. Since we all traveled, it made more sense for us to an app. At the time, the most convenient app was called “Chess with Friends”. Playing chess this way, kept the group connected even when we were separated by thousands of miles.

    I recall Prof Kane saying in class that the best Chess teams were composed of both Human and AI members. This combination makes me wonder, if the Ai has memorized 5+ levels of every possible maneuver, isn’t that enough to beat any human? I imagine this being like that scene from The Matrix when Neo meets the Architect and the Architect has a wall of screens containing a simulation of every reaction Neo could make.

    I think it’s a great story how you transformed defeat into desire. All too often, when people meet resistance, they turn around and run the other way, like screw this, I don’t want to play anymore. You chose to back up, gather momentum and run through this resistance.

    Admittedly my favorite section of the post was at the end of the first TED video (A brief history of Chess). At 5 minutes and 20 seconds, the narrator explains that the treadmill was invented as a torture device. I just want to say I ****ing knew it! And yes, The Queens Gambit was a great series.

  3. ritellryan · ·

    Really interesting piece, and I really liked the second video describing how the algorithm works. Andrae’s comment about it knowing every maneuver highlights the point really well that computers are the ultimate training partner and show us that you also need to vary what you do as well. A computer might provide the most optimal solution for what it expects the opponent to do, but the capacity of the human to identify and understand nuance is what elevates the thinking. To continue on the war metaphor, Washington was unpredictable with his tactics for the time period and it was the reason the US won the war.

  4. therealerindee · ·

    First off, @scottysiegs you continue to amaze me with your willpower and dedication to things. The fact that you didn’t just settle with losing to your roommate while slowly picking up moves is quite impressive to me, and obviously what I would have done. That being said, I do think the fact that you leveraged the ability for chess.com to teach you as you played to get better was awesome. There are lots of apps doing this in a vast array of activities. I play a weekly Friday night virtual Texas Hold Em game and after a winner is crowned, you can go back and look at the analysis of all your hands versus the other players at the table. This is not only a great way to learn, but a lovely peak into the human psyche as more hands than you think are total bluffs. Anyways, I totally agree with you that AI is a fantastic way to amplify human intelligence rather than replace human intelligence because the algorithms are the specifics way faster and more accurately than we are and we can use those outputs to influence our decisions whether it be in chess or a much larger business decision.

  5. I connected well with your blog because board games was a hobby my roommate and I picked up during the early days of the pandemic as well. It’s a great way to fill some time while also being competitive. We played a lot of cribbage and this makes me wonder if AI could be used in games like that to teach and enhance gameplay. One of the reasons why I have not had a strong desire to learn or play chess has been because it seems complicated to learn. But with the knowledge from your blog I may be inclined to check out chess.com to see if its something that I could pick up quickly like you did.

  6. abigailholler1 · ·

    This is really interesting – I’m not a chess player myself, but I do really enjoy card games like spider solitaire and the likes. I think what is unique about using AI in this way is how we as humans can learn how to be more successful with various games through a more systematic approach. For example, when I play spider solitaire, if I can’t crack a specific hand, there is a playback where I can watch the computer show me the quickest solution. Allowing me to watch the solution helps me understand the option set at every turn, and encourages me to think a couple moves ahead. If AI can help gamers think more systematically and a couple moves ahead, it will eliminate any moves that would have been played based on emotion or without a thought through strategy. Also, I think I may need to take up chess after watching The Queen’s Gambit, perhaps a hobby for our next quarantine (kidding!!).

  7. olivia_levy8 · ·

    Ah yes, the Queen’s Gambit chess craze, I definitely tried my hand at a game of chess after watching and was disappointed by my lack of skill and ability to visualize games on the ceiling. Great post, initially when I read the title I thought this was going to be about the AI bias that incorrectly flagged chess comments on YouTube as harmful and inappropriate as they contained the words “Black”, “White”, and “Attack”. This is just another interesting intersection between Chess and AI.

    It is comforting to know, that at least for now AI is not going to replace us to the full extent, as chess will hold them back from this feat. I think that the coaching application is super helpful and once again underlines the point that the answer isn’t just humans or just AI but a combination and collaboration between humans and AI that will best suit society moving forward. Great post!

  8. Great post. I do think that human partners inject a certain degree of unpredictably/ creativity that the AI currently have difficulty dealing with.

  9. williammooremba · ·

    Awesome post Scott. I have known how to play chess and played off and on for a decent while. I got really interested in Chess last year when one my favorite streamers played in PogChamps 2, a chess.com sponsored Twitch tournament, around September of last year. I just checked my chess.com account and I have played over 500 rapid games since I made my account around the same time. One thing that is interesting to me about AlphaZero is how its moves seem differ from traditional human moves. I had watched a video analyzing AlphaZero’s attacking style when playing against the traditional chess engine Stockfish 8. From my understanding, Stockfish 8 seemed to follow a lot of traditional chess strategy particularly with trying to keep pawns on the board. AlphaZero on the other hand seemed happy to sacrifice pawns for long-term positioning many moves ahead. To me it seems the AI based AlphaZero is somewhat decoupled from traditional human based chess theory. I think the concept of AI not fully thinking the way humans could have very interesting long-term impact on its ability to work on problems.
    AlphaZero’s Attacking Chess video by IM Anna Rudolf: https://www.youtube.com/watch?v=nPexHaFL1uo
    Article about the AlphaZero Stockfish match:
    https://www.chess.com/news/view/updated-alphazero-crushes-stockfish-in-new-1-000-game-match

  10. changliu0601 · ·

    Interesting Post!The first time i know AI is applied into board game is the game between AlphaGo and Go Champion.In 2016 the South Korean Go champion decided to retire from professional play because after the debut of AI in Go games, he realized AI cannot be defeated.But he is the only human to ever beat the AI.Such a sad story.

  11. alexcarey94 · ·

    Cool post- I also loved queen’s gambit so it is interesting to see after watching the show how tech can really improve the game playing but ultimately not master the game due to the unpredictability of the human mind. I feel that this similar thing is applicable to other areas of tech such as customer service. AI can not always replace because you can not always predict the types of questions people will ask or the best approach if a question is different than something typical.

  12. lisahersh · ·

    Excellent post., Scott! I had a similar experience at the beginning of the pandemic too, but with poker. I had never played and my husband is an avid lover of poker (and a sore loser). Using an app where I played against the computer and/or real people I was really able to improve and now occasionally win some games. I see so many potential training applications through this kind of technology in the future too. I remember having to do business simulations in a few MBA classes and could see AI as helping to make the patterns and consequences of specific actions more concrete.

  13. sayoyamusa · ·

    I was impressed with how tactful you utilized the technology in your life. Thank you for sharing the perfect example of leveraging technology to enhance human intelligence! Your blog is also a great reminder that machines and humans have different strengths for sure.
    Regarding this topic, I’ve found an interesting AI chess program called Maia that deliberately makes mistakes. “It aims to predict human moves, even the wrong ones, instead of being the grandmaster of chess and making every move right,” according to the article. I’m not sure if this AI can eventually work or not, but it will be interesting to see how machines will imitate even human flaws.
    Here is the link: https://www.analyticsinsight.net/an-artificial-intelligence-program-that-makes-mistakes-yes-it-exists/

  14. shaneriley88 · ·

    This is awesome! This reminded me of using ship simulators in college for navigational training. Albeit we weren’t using AI I still benefited greatly from “battling” a computer program and a instructor that was able to add human inputs. Fascinating post. I got into the Catan game a bit with my friends during COVID not sure i’m ready to play a bot yet.. I’m still working on my Sheep strategy.

  15. Jie Zhao · ·

    Amazing post, Scott! I’m impressed with how you continue to connect everyday lives to emerging technology. While I’m not a chess player myself, I enjoyed the second video explaining the fundamentals of programming behind a computer player! This is a great example showcasing what AI is good at – things that are structured and can be optimized with a great memory. I love how you ended with the quote, and agree that the best product will be a combination of leveraging what both machines and humans are good at.

  16. courtneymba · ·

    Awesome post! What an amazing quarantine ritual! I’ve always wanted to learn chess, and that is so cool about learning from the AI within Chess.com. Chess is the classic example in my mind of demonstrating the benefits of a “man/woman + machine” approach. Also I love how you touched on how AlphaZero unlocked new and fresh ways to think about the game and how chess experts personify the tool lol!

  17. Divya Jha · ·

    I really enjoyed reading your post! Great analogy, taking us through your chess.com journey – I haven’t played chess since I was a child and this makes me want to check out Coach Mode to get some practice. I liked how you tied it to Alpha Zero, I also like the idea of looking at AI as a training partner and not a replacement to human intelligence.

%d bloggers like this: