Why AI Chess Bots Are Virtually Unbeatable (ft. GothamChess) | WIRED
“I got checkmated in 34 moves.” Levy Rozman a.k.a. GothamChess plays chess against Stockfish 16, the strongest chess computer in the world, and analyzes the way it thinks in order to apply it to his own gameplay. With help from computer chess software engineer Gary Linscott, these chess pros identify why Stockfish is virtually unbeatable by a human, from opening move to endgame.
Watch more GothamChess here:
Director: Lisandro Perez-Rey
Director of Photography: Francis Bernal
Editor: Paul Isakson
Talent: Gary Linscott; Levy Rozman
Line Producer: Joseph Buscemi
Associate Producer: Paul Gulyas; Brandon White
Production Manager: D. Eric Martinez
Production Coordinator: Fernando Davila
Camera Operator: Brittany Berger
Gaffer: Mar Alfonso
Sound Mixer: Michael Guggino
Production Assistant: Albie Smith
Post Production Supervisor: Alexa Deutsch
Post Production Coordinator: Ian Bryant
Supervising Editor: Doug Larsen
Assistant Editor: Andy Morell
Still haven’t subscribed to WIRED on YouTube? ►►
Listen to the Get WIRED podcast ►►
Want more WIRED? Get the magazine ►►
Follow WIRED:
Instagram ►►
Twitter ►►
Facebook ►►
Tik Tok ►►
Get more incredible stories on science and tech with our daily newsletter:
Also, check out the free WIRED channel on Roku, Apple TV, Amazon Fire TV, and Android TV.
ABOUT WIRED
WIRED is where tomorrow is realized. Through thought-provoking stories and videos, WIRED explores the future of business, innovation, and culture.
Gotham!!
levi surely becoming walter white
This felt a little too.. produced. Like they weren't even actually talking to each other, just reading scripts. Its fine but videos like this are usually better when they feel more authentic. Interesting subject though. I did still really enjoy it. Levy has a way of making chess fun and understandable for people like me who enjoy it but are still terrible at it.
Please tell me he's going to upload the video of him playing Stockfish onto his channel 🤞
Okay this was Levi asking 10 variations of the same questions and the dude giving the same answer
Stockfish has past chess games database, it blazes through all the options 50 moves ahead, calculates which has the highest probability of winning and plays that
I love this vid
The way top engines just sit and shuffle pieces seemingly without principle because they found some sequence related to something very deep requiring tempi…
The best chess players (human) regularly think at least 5-10 moves ahead. There are famous games where the players thought 10, even 20 moves ahead. The first or second Kasparov Vs. The World game was famous because most people voted for a move in the endgame that according to Kasparov was losing. No one could see it. There were countless arguments over it (not the least of which the game was plagued by hackers). Kasparov then shut everyone up by publishing his analysis. It was about 20 moves ahead, and each move had multiple variations. And it turned out he was right. The move that was voted in was losing. Everyone was blown away by his ability to think that far ahead.
And nowadays, just look at Hikaru. There's even a comedic video out that shows how far he thinks ahead within split seconds in very complicated positions. That was one of few instances that I considered the word, "bruh!" appropriate, LOL.
Now, compare that to Stockfish, which calculates 50-60 moves ahead in every position within seconds.
does stockfish without its book vs itself play the perfect is boring game of chess?
That’s was some great questions asked.
The display of the game can't be more worse
Stop putting this joke on your channel. Put a proper chess player on FFS.
Wired and Gotham Chess is the new Hikaru and Gotham Chess.
Request for the video title to be altered, it’s not virtually impossible, it’s literally impossible
What is the highest possible rating a perfect computer with, say, unlimited processing power can achieve?
Most amount of opened option per move is the goal for the first 2 move . Knights has a lot of weight in that regard
WIRED here helping bridge the gap between entertainment and science. More Youtubers etc. interviewing the experts please 🙂
Based on the ELO difference, Levy has 1 in 1 trillion chance to beat Stockfish. But I highly doubt it.
wow third time 😲😲😲
As a chess engine author, this was fun to watch.
What does it have to do with AI.
AI and computer program / algorithms are not synonyms.
Wow, Levy without screaming, yelling and cussing 😮 then I realize he isn't on his channel, it's like he's a different person 😂
Levy looks professional
Would love to see AI machines playing football, motoGP, F1, etc.
So what happens if you play Stockfish vs Stockfish? Is it 50/50 between each. Is it the player that goes first gets an advantage? Would they just play the exact same game every time as they would choose the best move which would be the same every game they played?
the beauty of this video is that it is entertaining and contains new information for both people who dont play chess at all and people who are really good at chess.
really interesting how the AI is designed to 'think'.
thanks wired, thanks levy, thanks… stockfish i guess!? 😅
So best chance to beat stockfish would be to close the position and try to not lose any pieces and be 3499 elo
OK but shouldn't someone stronger do this kind of evaluation ?
This guy should make a YouTube channel. What a lad.
By delving deeper into Stockfish's and overall chess engine development history, you could make like two more high quality videos.
Is it more than brute force or does it think creatively? This video suggests not.
I love how Levy is asking all these questions like he didn't already know most of the answers
It helps people improve, yes… but never to the point that we can actually beat AI… which is kind of pointless. I am saying this as a low elo chess player and a computer science student.
It’s curious to note that they both kept using the word “think” and kept personifying the software as having a human like presence. Unsure if it’s intentional or just a quirk of ours. Either way it furthers the mythology of the software.
Unsure if there’s anything to be had about discussing software playing a game of chess. It’s supposed ELO rating is irrelevant and besides the point. Whether it’s 3000+ or 10 million. Allegeding it here again is building the mythology of the software. Chess is a human game. If it’s just about “winning”, the players have already lost.
It’s an uneven match up for the player vs software. In a finite game with specified moves, it’s bringing a stick to the tank fight. As the dev concedes, things slip when the ability to search the database is removed. Can a librarian help find more info, at a faster speed than a search engine? No. But that’s not the point of a librarian. Help the librarian do what a librarian is to do, and use the software for what’s it’s for. Let’s not confuse and compare the two.
High leavel reinforcement learning
This was a really satisfying and entertaining video. Thanks!
In other words – to beat Stockfish, you have to find a way to play an absolutely terrible move early on that it will rule out of its analysis, only to beat it just a few moves later.
Let it play against Dr. Strange.
Stockfish will never make blunder like human beings make under pressure
Me: Nobody can read the future
Stockfish:
As impressive as stockfish is, what its doing is fundamentally cheating. It has a database with billions of options in a game meant for humans where that’s not possible. A human playing stockfish is the same as a human playing a human using a chess engine.
So basically stockfish just cheats by using a table base
This video contradicts itself. It claims that games are fed into a neutal network, but then goes on to explain about Alpha-Beta-Pruning. Stockfish has nothing to do with Neural Networks..
Thank you drug addict Daniel Radcliffe
Don't compare yourself with stockfish 😅, basically they are just big cheaters with access to all the knowledge 😂
How did stockfish lose 106 times to alphazero then?
Stockfish – aka 'The Chuck Norris of Chess Engines'.
Well chess in kinda way has limited amount of moves even though its a big number if some computer has access to all of these moves then chess is pretty much solved
You are gem of the world ❤
Just to clarify. Stockfish is not Chess AI. It is chess with conventional engine based. It calculates all possible position, and trim any bad possibility.
On the other hand, Chess AI use neural network (Alpha Zero, leela). You train the AI to have "chess brain" (checkpoint) and this checkpoint will play chess.
Thats why neural network chess engine plays a bit different than stockfish. They play a bit more human (but brilliant human).