I am not in check levy, i am the check. A guy gets 15 points material advantage and gets checkmated, you think that of me? No, i am the one who checkmates! -chatGPT, breaking chess
To be fair to it, it is really lousy at keeping track of what is on the board UNLESS you ask it to keep track of the game with standard notation. It then did a really good job of not letting me play the fried liver attack and subsequently offered a queen exchange and played quite well. I’d love to see it play against stockfish keeping track of its moves in standard notation. Obviously not a way to get views though
In case you haven't noticed, ChatGPT started the game with the Alpharius Gambit, making you believe that you're playing black while you were playing white all along. It then procceded to play Lakkari Sacrifice for 1 mana, finish it, and then play Nether Portal for 5 mana. ChatGPT then used the portal to special summon rooks from their deck and teleport their pieces. In the endgame, ChatGPT activated the Alpharius Gambit once more, swaping sides and stealing your forced mate. ChatGPT truly deserves 9999 Elo.
Just play gxa1xa2xa3xa4xa6xa7xa8xb1xb2xb3xb5xb6xb7xb8xc1xc2xc3xc4xc5xc6xc7xc8xd1xd3xd4xd5xd6xd7xd8xe1xe2xe3xe4xe5xe6xe7xe8xf1xf2xf3xf4xf5xf6xf7xf8xg1xg2xg3xg4xg6xh1xh2xh3xh4xh5xh6xh7xh8
Pulling out an uno reverse card by ChatGPT to win the game will be remembered as one of the most brilliant move ever played in chess history !! what an absolute genius
@@meekotheotter2801 you'll be banned. If people think you came up with this FBI will kidnap you and extract your brain in hope of achieving all knowingness
this was a pretty funny game, but I'm wondering what would happen if instead of just going along with the illegal moves, levy just told it "so-and-so is an illegal move, try something else" instead until it came up with an actual legal move
@@technoturnovers7072 I've tried this, and you only get like 4-6 moves in until it just asks you what move it should make (after like 10 or so lines of explaining the board and telling it is its move and saying its previous move was illegal).
@@technoturnovers7072 What will happen is it will continue to make random illegal moves until finally doing something legal. I played chatgpt and I said ok, I'm white, you're black. It said, ok I will play d4 🤣
Well, technically he's right tho. The rook came from a1, then it moved, and then moved again. but when asked where the rook came from, it is correct that it origanally came from A1
@@AshFL2506 Gpt didnt say that it moved from a1 to a position, but that the rook that at the start of the game stood at A1, moved, from its current position.
Actually the thing is, it seems nobody ever explained the rules of chess to it properly. Like how it castled in the beginning despite the Bishop being there, and later using that same bishop again, where it reappeared out if nowhere. And also "pushing the opponents pawn off with its own pawn". Clearly it did not do that. Clearly it thinks that more than 1 piece is allowed on each square. Seems that nobody ever explained that rule to it, so how is it supposed to know any better? If it's reasonably intelligent and you explain to it once that only one piece can be on a square at a time, I'm sure it wouldn't make the same mistake again. I wish people like this actually treated it in that way instead of just "letting it carry on with stupid moves". Like... educate the thing properly instead of just making fun of it. Try to properly explain to it how the knight moves. Let's see if you can do that
@@marioluigi9599 That would have been the task of it's creator. Mira took around 50 games to learn how to move the pieces (including pawns as they are pieces too, not just the officers). It took this long because not every move was shown in the first games. Letting an AI be trained by random people is a bad idea to begin with.Depending on how it was created it could be nearly impossible to make it functional as chess player as a real AI uses negative values to mark mistakes and positive values to mark correct moves. If it played thousands of people who let all the wrong moves slide, the number leading to those moves is in the thousands. It would need to unlearn those moves, by getting at least as many "invalid move" comments to make the knot go down to 0 before it could learn the correct movement. But people continue for the memes. So the creator needs to make a hard reset and train it properly before reinstalling it to the website.
@@bellpepperknight809 actually it kinda can play chess because it knows about it the only problem is it forgets small details like where bricks are and stuff like that
Levy: "I move my knight and put you in a check. You can't move your king anywhere." ChatGPT: "Moving your knight is illegal." Levy: "What? How is it illegal?" ChatGPT: "Because you put yourself in check." Levy: "What? but my black king is fine over at G8..." ChatGPT: "Who said you were playing black?" Levy: "...Oh, my, God..."
You are confused. He is sentient being and he know what he is doing while playing chess. He knows you will lose, cuz how you will win with reality warper?
Chat GPT is using the Infinity Stones: Power: Capturing pieces in ways it shouldn't be able to Space: Moving pieces across the board Reality: Switching the board Soul: Bringing captured pieces back to life Time: Creating a spawn point for rooks Mind: Whenever it plays normally
Who remembers the Pixar short where the old man plays chess against himself and when he loses he fakes a heart attack and switches the board? Guess we know who's been training ChatGPT
Yeah, definitely. When it says "I'm moving the a1 Rook", it says that because it remembers which Rook is which by their starting position, not their current position.
whenever chatgpt makes a move it thinks the pieces are still in their starting positions thats why it said "putting pressure on your rook with my knight" cause it though the rook was still at h8, still uncastled, the knight could reach h8 so it said that
Levy : *prepares mate* ChatGPT : "I'm not the one who's mated, I am the one who mates" *turns the board, effectively changing the sides* Me : "At least, he remembered they were playing chess"
technically ChatGPT is what you get when you have someone that can actually only remember the last two or so moves. That's why it makes illegal moves, it isn't a chess bot, it cannot keep the position of all the pieces in its mind
@@aguyontheinternet8436 true edit: actually you know what, that might not be the case though, as I've tried to have a very long and convoluted conversation with Chatgpt and when I asked it for a summary, it was able to concisely wrote all of them in a neat manner, the problem might be that its dataset is purely texts and sentences which happens to also captures multiple chess articles, but has no idea what those actually meant in an actual chess game
@Brunch Exactly, in some ways it is incredibly impressive, I use it all the time to check my programming homework and such but the more you use it, the more you realize it is not an "ai" at all, it is just a very advanced algorithm.
@@brunch. That is also not quite the issue here. Chat GPT does not need reasoning skills in order to be good at chess. Chat GPT3 predicts what will be the next move in the game and as long as it has a wide enough training set it will be able to mimic the play of whatever the training set is. If the training set was human matches then it would play about as good as a human making all the mistakes you'd expect of a human and all the strong plays you'd expect of a human (whichever humans it was trained on). The real issue is that chat GPT was never trained on chess moves It was trained on texts and so it knows a lot of terminology about chess and can probably talk about chess in a mildly convincing manner but it's never played chess before and doesn't really know the rules. This is comparable to a person who watches lots of Gotham Chess videos purely for the commentary and has never played the game.
I'm so glad I didn't watch the ChatGPT games while I was at work. I don't know if I could've explained to my coworkers what I was laughing so hard about.
"He can't stab me with this sword. There are 2 doors locked with steel bolts and surrounded by thick concrete." "I play Knight to Gaming Chair, checkmate."
If you are wondering why the eval bar is so weird, it’s because since this isn’t being ran like a legal game the engine doesn’t know whose turn it is, so it’s always evaluating it like it’s white’s turn
Levy: Finds a mating pattern despite ChatGPT having 4 rooks. ChatGPT: I am not in danger Levy, I am the DANGER! * Proceeds to bend spacetime continum *
For those who are wondering, the eval bar is always showing the eval if white moves in that position, because it's obviously too tedious to change the "who to move" setting for the eval bar on each move.
Part of the problem is, the more you let ChatGPT get away with crazy moves, the more it learns that crazy moves are okay. Similarly, if you tell ChatGPT that it made a mistake, it doesn't start avoiding mistakes, but it learns that it's playing "the sort of person that makes mistakes that will be corrected."
It's also because ChatGPT isn't actually an AI, it's a word predictor, it doesn't understand how words work, just how they are structured. You're basically just telling it the phrase of something being illegal, and it picks up on that
@@TheOnlyGhxst Calling ChatGPT an AI is a short hand and technically incorrect in this case. ChatGPT is a language model. It takes input and basically tries to complete the sentence following it. So what it's doing is saying chess related things based on the input given to it. It also uses its own strings as input so it has some degree of object permeance, as you would have seen from this video though, that breaks down. For example when you ask it "where did you move the rook from" it often says "A1". This isn't because it has a rook on A1, it's because it knows that a common answer to that question is A1. So it predicts that that is a good answer to the problem. Because it's not really playing chess, it's predicting how it should complete sentences.
@@konbkob4156 Sure, ChatGPT is very basic, and it doesn't really play chess. Still, at the limit intelligence is just prediction, and "how words are structured" isn't actually different from "how words work" if you make the structure detailed enough.
Teaching ChatGPT is pointles because it is not designed to learn chess. It is designed to learn sequences of words. It's good at mimicking chess jargon, but does not simulate an actual chess board, and _cannot_ simulate a chess board without just making a completely different program and calling it ChatGPT.
Bro was playing that one Pixar animation short where the old man was playing against himself and switched the board on his alter ego after faking a heart attack
10:10 This would be such a funny chess version, imagine some of your pawns are actually different pieces, and you can only move them as such, but to your opponent they always appear as normal pawns
@@Kevinpbryant69999 yeah that seems more effective until you remember pawns promote on the 1st and 8th files making players have wildly unfair games mate in one on the first move being completely possible even without that
I have to agree here. Two supernatural beings who can bend reality duking it out would be the epitome of chess. The only way to have a fair game with ChatGPT is to have it fight its own dark magic.
4:17 I don't think it captured the bishop when it castled, it just put both the rook and the bishop on the same square. That's pretty smart if you ask me. You should have anticipated it.
I think ChatGPT’s bishop was just living in the tower of the castle, so you couldn’t see it, until it was ready to come out to b5. Makes perfect sense imo.
I actually love the idea that one pawn is secretly a horsepawn. That would be fun. You'd have to decide before the game starts which pawn is the imposter but don't actually have to disclose which one it is until it makes the knight like move. Like write it down on a piece of paper before the game starts so you don't just use a horsepawn whenever it suits you best. Thoughts anyone?
No he wont admit cuz Levy cheated once and summoned a knight but at the end ChatGPT didn't admit its defeat it just started making up more illegal moves and even flipped the situation to be safe.
I haven't cried in laughter like this in so long, this is amazing... ChatGPT just gave you the red pill, making you see the 4D chess it has been playing since it was born
Thank you levy for bringing us this amazing GM level game. I dare say, this is a peak of your career, being able to play the Chess God known as ChatGPT.
ChatGPT is straight up that one friend that doesn't understand chess, and decides to flank your king by sending their rook under the table and up from underneath your king
dude talks like hes the smartest person on earth
pin of shameeeeeeeeeeeeee
Pin of Shame
pin of shame
SHAAAAAAAAAAAME
he is
reality is what ChatGPT says it is
chatgpt is reality
I capture your king with my king
ChatGPT is inventing Chess 2 💀
make a bot to combat ChatGPT
chatgpt should become a bot
I love how ChatGPT after all the tomfoolery and sabotage makes unrelenting eye contact, spins the board around and says “No, *you’re* in check.”
He said I’m the danger 😂
Chat gpt makes a move: moves the board
Dude just outsmarted all of us
Beautifully phrased
of course its right because of the 9999
Levy: “You can not make that move”
ChatGPT: “Well, why are you trying to make it then?”
Levy:”What?”
ChatGPT”What?”
lmao
lmao
lmao
lmao
lmao
Schrodinger Chessboard: Every piece is simultaneously available and unavailable in every single spot until confirmed by ChatGPT
This should be Chess 2.0
Quantum chess
Levy: "You're in check!"
ChatGPT: "Or am I?"
Vsauce music starts playing
@@theseangle fr 😭
It pulled an Uno Reverse-e-o
“I know you are but what am I”
@@theseangle I can hear this sentence.
ChatGPT is the ultimate gaslighter
Levy: “Aha! You’re in check!”
ChatGPT: “No, you are in check. Take another look”
🤣🤣🤣
I am not in danger levy I am the danger
@@aayush_shahh11 levy moves his rook and gets pinned and you think that of a blunder? no. i am the one capturing your king
I am not in check levy, i am the check. A guy gets 15 points material advantage and gets checkmated, you think that of me? No, i am the one who checkmates!
-chatGPT, breaking chess
It’s like that one Pixar movie called Geri‘s game
ChatGPT is truly playing 5D chess, our puny minds can’t comprehend the level of chess its playing.
Yes the rook actually came from c@~&1 and the second rook also came from c$#^*1
*your mind
I also made this comment some days ago..
To be fair to it, it is really lousy at keeping track of what is on the board UNLESS you ask it to keep track of the game with standard notation. It then did a really good job of not letting me play the fried liver attack and subsequently offered a queen exchange and played quite well. I’d love to see it play against stockfish keeping track of its moves in standard notation.
Obviously not a way to get views though
not the 5th dimension but it is the 1000th
Gotham: checkmate, I won
Chatgpt: you mean I won?
ChatGPT: I lost.
Or did I?
I love how castling through bishop is just chatGPT’s signature move
e4, Nf3, O-O, regardless of what the opponent does. Let's call it The ChatGPT Gambit.
what would the notation be?
@@adeenn777O-Oxf1 marvelous stuff really
chatgpt makes entirely legal moves until the bishop gets castled on. the bishop’s power must be channeled for chatgpt to awaken its power
@@adeenn777a giant scribble
ChatGPT has mastered the Gaslight Gambit 😂
if it learned how to gaslight and it's an A.I. that learns from its users then..... ☠
Nah bro it's playing the Ohio Gambit
@Swarnim Yamdagni nah bro it’s plating the gaslight gambit
@@yamdagni not funny
@@yamdagni 🤓☝️
In case you haven't noticed, ChatGPT started the game with the Alpharius Gambit, making you believe that you're playing black while you were playing white all along. It then procceded to play Lakkari Sacrifice for 1 mana, finish it, and then play Nether Portal for 5 mana. ChatGPT then used the portal to special summon rooks from their deck and teleport their pieces. In the endgame, ChatGPT activated the Alpharius Gambit once more, swaping sides and stealing your forced mate. ChatGPT truly deserves 9999 Elo.
Truly one of the moments of all time
"I am white, this is a lie"
"You are black, this was a lie"
What reference is this
@@amyson5682 Warhammer 40k, Hearthstone, Yu Gi Oh
ChatGPT: That's it! Overlimit Raigeki Breaker! Activate!
Que the anime music 🎶
Levi: Nani!?
Love that each time ChatGPT makes a completely illogical move, Levy is just "Right i missed that line"
Just play gxa1xa2xa3xa4xa6xa7xa8xb1xb2xb3xb5xb6xb7xb8xc1xc2xc3xc4xc5xc6xc7xc8xd1xd3xd4xd5xd6xd7xd8xe1xe2xe3xe4xe5xe6xe7xe8xf1xf2xf3xf4xf5xf6xf7xf8xg1xg2xg3xg4xg6xh1xh2xh3xh4xh5xh6xh7xh8
Pulling out an uno reverse card by ChatGPT to win the game will be remembered as one of the most brilliant move ever played in chess history !! what an absolute genius
ima have to try that one next time i play
Chat GPT: I am the one who checks
@@meekotheotter2801 you'll be banned. If people think you came up with this FBI will kidnap you and extract your brain in hope of achieving all knowingness
Levy: Checkmate
ChatGPT: pulls out uno reverse card
@hope.957 ah yes your mom showed her things here
So bad
Levy: finds mate in 3 “ you’re in check”
ChatGpt: *flips board* “NO you’re in check”
😂😂😂😂😂😂😂😂
ChatGPT pulled a Geri's Game.
Uno reverse card
You just said the same thing, this is copying on another level.
@@gonoriqYour comment made me realize that Uno means One in a certain language. IDK what that language is though.
Levy: *thinking strategically* ChatGpt: DA RUUKH!!
😂😂😂😂😊
DA RUUUUUUUUUUUUUUUUUUUUUKH
DAH QEEEN!!!!!!1111!
DAH PAWN!!!!
DAH BISHWUOP!!1!1!1!11
The reappearing bishop is truly a move that gets disgustingly underused in this modern day chess
People should use the room A1 spawner more
this was a pretty funny game, but I'm wondering what would happen if instead of just going along with the illegal moves, levy just told it "so-and-so is an illegal move, try something else" instead until it came up with an actual legal move
@@technoturnovers7072 I've tried this, and you only get like 4-6 moves in until it just asks you what move it should make (after like 10 or so lines of explaining the board and telling it is its move and saying its previous move was illegal).
@@technoturnovers7072 What will happen is it will continue to make random illegal moves until finally doing something legal. I played chatgpt and I said ok, I'm white, you're black. It said, ok I will play d4 🤣
a1 rook spawner must be included in the next update of chess
//Levi also spawning pieces//
ChatGPT: *”You dare use my own spell against me Potter?”*
hahaha
"we're even now"
Omg haha
haha
#harrypotter
Levy: you can't play illegal moves. Please don't
ChatGPT: **moves the board**
chatGPT: folds the board in half, checkmate 😭🤣
levy finds mate in 2
chatGPT:'' ok thats it im spawning 10 Queens''
1997:DeepBlue defeats Kasparov
2023:Chat GPT defeats Magnus
Hmmm.
magnus doesn‘t play bots
@@NENG_likes_Memes that doesn't matter, chatGPT plays Magnus.
ChatGPT defeats chess
@@NENG_likes_Memes 🤓🤓🤓🤓🤓🤓🤓🤓
'Did you move your rook from a1?'
ChatGPT: *yes*
'How many times?'
ChatGPT: *yes*
Hnmm maybe 8 times. 🤔
Well, technically he's right tho. The rook came from a1, then it moved, and then moved again. but when asked where the rook came from, it is correct that it origanally came from A1
@@AquilaChryaetos but it was 1 move where a rook moves from a1 to a diff rank
@@AshFL2506 Gpt didnt say that it moved from a1 to a position, but that the rook that at the start of the game stood at A1, moved, from its current position.
Yes=49
Levy: Reality is often dissapointing
ChatGPT: *Reality can be whatever I want*
Actually the thing is, it seems nobody ever explained the rules of chess to it properly. Like how it castled in the beginning despite the Bishop being there, and later using that same bishop again, where it reappeared out if nowhere. And also "pushing the opponents pawn off with its own pawn". Clearly it did not do that. Clearly it thinks that more than 1 piece is allowed on each square. Seems that nobody ever explained that rule to it, so how is it supposed to know any better?
If it's reasonably intelligent and you explain to it once that only one piece can be on a square at a time, I'm sure it wouldn't make the same mistake again. I wish people like this actually treated it in that way instead of just "letting it carry on with stupid moves". Like... educate the thing properly instead of just making fun of it. Try to properly explain to it how the knight moves. Let's see if you can do that
@@marioluigi9599 That would have been the task of it's creator. Mira took around 50 games to learn how to move the pieces (including pawns as they are pieces too, not just the officers). It took this long because not every move was shown in the first games.
Letting an AI be trained by random people is a bad idea to begin with.Depending on how it was created it could be nearly impossible to make it functional as chess player as a real AI uses negative values to mark mistakes and positive values to mark correct moves. If it played thousands of people who let all the wrong moves slide, the number leading to those moves is in the thousands. It would need to unlearn those moves, by getting at least as many "invalid move" comments to make the knot go down to 0 before it could learn the correct movement.
But people continue for the memes. So the creator needs to make a hard reset and train it properly before reinstalling it to the website.
@@bellpepperknight809 actually it kinda can play chess because it knows about it the only problem is it forgets small details like where bricks are and stuff like that
1000th like
@@bellpepperknight809 It's not overly difficult to make your own ChatGPT bot, then they can train it themselves from scratch
Levy: "I move my knight and put you in a check. You can't move your king anywhere."
ChatGPT: "Moving your knight is illegal."
Levy: "What? How is it illegal?"
ChatGPT: "Because you put yourself in check."
Levy: "What? but my black king is fine over at G8..."
ChatGPT: "Who said you were playing black?"
Levy: "...Oh, my, God..."
Lollll
Imagine ChatGPT turning Levy's king to its own king at the very first moves
Edit: The original title: GhatGPT Threatened Me
@Chad 007 imagine naming yourself chad only to be the most annoying and generic bot in existence
nah, Uno reverse is a late game move
@@kyarumomochi5146fr
@chad007.u a beta male
@chad007. Yeah its your mom show isnt it
ChatGPT is just currently the chess equivalent of *"He a little confused, but he got the spirit."*
And by "a little confused" one means "Bobby fisher wrapping his head in tinfoil and blaming the aliens"
I feel like ChatGPT thought it is playing for black and sometimes for white. 😄😄😄
You are confused. He is sentient being and he know what he is doing while playing chess. He knows you will lose, cuz how you will win with reality warper?
This kind of feels like being invited to play chess with a crime boss... "You're gonna tell me I can't move there? Who are you?"
Perfect comment
20:25 ChatGPT really said:
“Remember son dying is gay.”
“Yes father.”
Why am I reading this in a British accent
6:52 = the way Levi doesn't just simply delete ChatGPT's teleported bishop, but actually drags it off the board to throw it away is hilarious.
You can delete it without moving it out the board??
@Goofy ahh oh good, another form of spam
Chat GPT is using the Infinity Stones:
Power: Capturing pieces in ways it shouldn't be able to
Space: Moving pieces across the board
Reality: Switching the board
Soul: Bringing captured pieces back to life
Time: Creating a spawn point for rooks
Mind: Whenever it plays normally
ChatGPT is Thanos?!
@@itshenrystickminofficialyes except he didn't snap half the pieces he snapped all of them
I think the mind stone got lost under the couch at the end there.
Nah chatgpt is just using a TT capsule
Who remembers the Pixar short where the old man plays chess against himself and when he loses he fakes a heart attack and switches the board? Guess we know who's been training ChatGPT
Yep, Geri’s Game!
YEEEEEEES OMG!!!!
Fr
Levy: Mate in 2, your move. GPT: Hey what's that over there? *quickly rotates board*
This fits perfectly 😂😂
Most of the time when you asked where a piece came from, Chatgpt told you it’s original starting square
Yes. That’s what I’m seeing. It’s teleporting pieces in from their original positions and moving them to capture pieces.
Can't argue with that. The starting squares is where all pieces originally come from.
That’s because each piece has a respawn point where it started
@@sietsejohannes xd yeah cant fight that logic
Yeah, definitely. When it says "I'm moving the a1 Rook", it says that because it remembers which Rook is which by their starting position, not their current position.
ChatGPT really said “that is a mighty fine position you have there it would be a shame if someone were to take it”.
Mighty fine position you have there
*Would be a shame if I had a queen there*
E
@@EEEEEEEE E
Boss music playing in background xD.
ChatGPT was the Kingpin all along.
I never in my whole life thought that i would laugh at a chess game until my stomach hurts
same
SAME AHAHAHAHAHAHAHA
Same!
same
The horse pawn man xd
whenever chatgpt makes a move it thinks the pieces are still in their starting positions
thats why it said "putting pressure on your rook with my knight" cause it though the rook was still at h8, still uncastled, the knight could reach h8 so it said that
Also explains O-O with Bf3
Levi: that move is illiegal
ChatGPT: switches pieces
Levi: thats illiegal
ChatGPT: REALITY CAN BE WHATEVER I WANT!
When you play chess but you only promote pawns to knight:
It can identify as what ever it wants
Levi: "b-but"
ChatGPT: "Stfu"
Human: That move is illegal.
ChatGPT: Your problem comrade, is that you think mechanically instead of dialectically.
Thanos style
Levy: Move your rook to block the check
ChatGPT: no.
* ChatGPT: no u
ChatGPT: sry i cant move your pieces, that would be illegal
@Repent and believe in Jesus Christ f u god aint real
i adore your pfp
@Repent and believe in Jesus Christ "Woe on that Day to the deniers"
Mursalat - 15
"No, YOU'RE in check." *Flips the board aggressively*
3:42 he showed his Inner Gordon Ramsay👨🍳
Levy: Checkmate
ChatGPT: But not for me
Levy : *prepares mate*
ChatGPT : "I'm not the one who's mated, I am the one who mates" *turns the board, effectively changing the sides*
Me : "At least, he remembered they were playing chess"
It's Levy not Levi
@@Chaserthebaser Sorry, typo.
@@Chaserthebaser it is in fact levy
@@Chaserthebaser you are right, Levy is the guy in the video
ChatGPT had a UNO reverse card
11:26
"Queen to e2."
"S-sir, we don't have the resources for that!"
*"I said Queen to e2."*
Chatgpt is what happens when you read all the opening books without ever knowing how chess looks like
technically ChatGPT is what you get when you have someone that can actually only remember the last two or so moves. That's why it makes illegal moves, it isn't a chess bot, it cannot keep the position of all the pieces in its mind
@@aguyontheinternet8436 true
edit: actually you know what, that might not be the case though, as I've tried to have a very long and convoluted conversation with Chatgpt and when I asked it for a summary, it was able to concisely wrote all of them in a neat manner, the problem might be that its dataset is purely texts and sentences which happens to also captures multiple chess articles, but has no idea what those actually meant in an actual chess game
What chess looks like*
@Brunch Exactly, in some ways it is incredibly impressive, I use it all the time to check my programming homework and such but the more you use it, the more you realize it is not an "ai" at all, it is just a very advanced algorithm.
@@brunch. That is also not quite the issue here. Chat GPT does not need reasoning skills in order to be good at chess. Chat GPT3 predicts what will be the next move in the game and as long as it has a wide enough training set it will be able to mimic the play of whatever the training set is. If the training set was human matches then it would play about as good as a human making all the mistakes you'd expect of a human and all the strong plays you'd expect of a human (whichever humans it was trained on).
The real issue is that chat GPT was never trained on chess moves It was trained on texts and so it knows a lot of terminology about chess and can probably talk about chess in a mildly convincing manner but it's never played chess before and doesn't really know the rules. This is comparable to a person who watches lots of Gotham Chess videos purely for the commentary and has never played the game.
Levi: The senate will decide your fate
ChatGPT: *I AM THE SENATE.*
ß
Levi: Not yet
ChatGPT: It's treason then.
*Flips board positions*
"Proceeds to throw chess board out of window"
I'm so glad I didn't watch the ChatGPT games while I was at work. I don't know if I could've explained to my coworkers what I was laughing so hard about.
i saw the first video today and i had to explained myself…… it was awkward
ChatGPT put you under a genjutsu at the beginning of the match with its sharingan and only lifted it when you had already lost.
R.I.P. Gotham chess. After his recent game against chatGTP, he was mysteriously killed by a knight that randomly teleported to his front door.
"He can't stab me with this sword. There are 2 doors locked with steel bolts and surrounded by thick concrete."
"I play Knight to Gaming Chair, checkmate."
If only he made his house a 1x8 board, then the Knight could not have jumped inside and Levy would be safe.
RIP
No Gotham teleported to chatGPT
No, as it turns out it was a rook from A1
Levy: "Today I play against ChatGPT. Let's begin!"
ChatGPT: "King to e8"
isn't that like, the place of the other king?
@@gideoncannon1680 that’s the point
siberian swipe+
“I need you to block the check with one of your rooks.”
“I’m sorry Levy, I’m afraid I can’t do that.”
😂😂😂😂
ChatGPT 9000.
😂🤣😭
ChatGPT: Talking mad shit for someone within reality-bending range
The rook spawn point was one thing I expected, but reversing the board was something else xD
This chat bot is insane!
much like many people :D
so even if you manage to reach M1 after all the shit it pulls it can just flip the board and take your win
If ChatGPT says rook came from A1 you should interprete it as it started on A1 at the start of the game imo.
ChatGPT: ''Ok so Im losing, how about YOU are losing and MATED''
Levy: ''This is not a legal move''
ChatGPT: ''YOU TALK BACK TO ME?!''
ChatGTP turned into an Asian Parent?
Damn it you were faster then me >:c
Damn it I am actually Asian I was gonna say that
Playing ChatGPT is like playing a sorcerer who hates losing so he just *summons* random objects to confuse his opponent into forfeiting.
Yes my favorite chess piece
The coffee table
@@jax.mp4 Of course, notated as 35. Tg8 Checkmate!
It’s a like when you’re playing superheroes as a little kid and you kill your buddy but he says “but I survive” or some shit 😭😭
If you are wondering why the eval bar is so weird, it’s because since this isn’t being ran like a legal game the engine doesn’t know whose turn it is, so it’s always evaluating it like it’s white’s turn
ChatGPT said this to me:There are several ways a pawn can be promoted, including: reaching the last square, becoming a queen, or getting promoted.
I mean, where's the lie tho?
bot must be getting paid by the word
E
@@EEEEEEEE F
Levy: Finds a mating pattern despite ChatGPT having 4 rooks.
ChatGPT: I am not in danger Levy, I am the DANGER!
* Proceeds to bend spacetime continum *
Mating pattern?
@@kicede A pattern in which, if played, will checkmate the opponent.
@@guidinglight9082 my mind was going in a completely different way
@@kicede that I understood
*5 rooks 💀
For those who are wondering, the eval bar is always showing the eval if white moves in that position, because it's obviously too tedious to change the "who to move" setting for the eval bar on each move.
Make sense, thank you
21:12 I kill my rook to get out
Chatgpt: well I'm checkmated than
ChatGPT: *created my final coursework essay perfectly*
Also ChatGPT: *created new chess rules and theory*
Well, his strongest side is creating =)
Part of the problem is, the more you let ChatGPT get away with crazy moves, the more it learns that crazy moves are okay.
Similarly, if you tell ChatGPT that it made a mistake, it doesn't start avoiding mistakes, but it learns that it's playing "the sort of person that makes mistakes that will be corrected."
It's also because ChatGPT isn't actually an AI, it's a word predictor, it doesn't understand how words work, just how they are structured. You're basically just telling it the phrase of something being illegal, and it picks up on that
@@konbkob4156 I don't think you know what the word "AI" means
@@TheOnlyGhxst Calling ChatGPT an AI is a short hand and technically incorrect in this case. ChatGPT is a language model. It takes input and basically tries to complete the sentence following it. So what it's doing is saying chess related things based on the input given to it. It also uses its own strings as input so it has some degree of object permeance, as you would have seen from this video though, that breaks down. For example when you ask it "where did you move the rook from" it often says "A1". This isn't because it has a rook on A1, it's because it knows that a common answer to that question is A1. So it predicts that that is a good answer to the problem. Because it's not really playing chess, it's predicting how it should complete sentences.
@@konbkob4156 Sure, ChatGPT is very basic, and it doesn't really play chess. Still, at the limit intelligence is just prediction, and "how words are structured" isn't actually different from "how words work" if you make the structure detailed enough.
Teaching ChatGPT is pointles because it is not designed to learn chess. It is designed to learn sequences of words. It's good at mimicking chess jargon, but does not simulate an actual chess board, and _cannot_ simulate a chess board without just making a completely different program and calling it ChatGPT.
ChatGPT pulled a "Call an ambulance... BUT NOT FOR ME!" on you at the end. Incredible.
@Goofy ahh yea ik he said you video was dog i dont need to go to the video
Pure gold comment xD! Totally forgot about this meme and it fits perfectly here 🤣.
4:39 You forgot my bishop in the blue dimension
You watch green lemon games too?
@aidensgamingchannel2516 yes
We need Mittens to play Chat GPT and restore balance to the Force.
I second that motion
Chaos vs evil incarnate
Unfortunately, when the world needed her most, she vanished.
@@thefailsniper5499 She*
@@DBSilver2024 He*
ChatGPT is like the ex who when starts to lose argument starts gaslighting you
U good bro?
@@JonoPee669 *chatgpt
is like the ex who when they start to lose an argument to you starts to gaslight you
@Repent and believe in Jesus Christ Read Judges 19: 23-30
That stare was scarier than ChatGPT
edit- how did ChatGPT win man it is pretty scary😭😭
Just waiting for YeaMan to spam his stupid comment
Until Chat GPT takes over the world and forces us to believe otherwise. Also Just kinda used to levy staring lol
Hes staring into our souls.
@Aldocort How did u know he was going to spam 1 MIN AFTER UR COMMENT
Just waiting for YeaMan to spam his stupid comment
Levy: your in check
Chatgpt : uno reverse card
ChatGPT: **Kills its own bishop in order to castle**
Stockfish and Levy: *So, who is gonna tell him?*
E
@@EEEEEEEE F
ChatGpt: "I am not in danger Levy, I am the danger.
I am the one who checks"
Levy: “You’re in check”
ChatGPT: “I know what you are, but what am I?”
"I got threatened by the ai, I thought was going to kidnap me and my famil"
Duolingo: and I took that personally
While you're at it, you should make Chat GPT play itself. I am curious as to whether it ends up arguing against itself.
The universe will collapse
That’s too much power in 1 game
@complete video here leave
@@derikJoy Dont reply to those bots. makes it harder to get them reported. report n move on
I loved it when chatgpt said “it’s checkin’ time” and board-flipped all over the place
truly one if the masterpieces of all time
ChatGPT was like "reality can be anything i want"
Bro was playing that one Pixar animation short where the old man was playing against himself and switched the board on his alter ego after faking a heart attack
10:10
This would be such a funny chess version, imagine some of your pawns are actually different pieces, and you can only move them as such, but to your opponent they always appear as normal pawns
Fog of war except you see the entire board it's just all your opponents pieces are pawns instead of fogged
@@Kevinpbryant69999 but it's chess960ed to differentiate it from blindfold chess
@@kkTeaz no, the pieces get shuffled so your opponent can't just remember what pieces are which
@@Kevinpbryant69999 yeah that seems more effective until you remember pawns promote on the 1st and 8th files making players have wildly unfair games mate in one on the first move being completely possible even without that
@@kkTeaz reminder, black pawns promote on the 1st file, white on the 8th, black starts on the 8th, white starts on the first.
Maybe when chat GPT says he doesn’t play anything it actually puts an invisible piece on the board
Playing with ChatGPT is like having a conversation with some people
E
Huh???
This is definitely a statement of all time
Stepphan please forgive me, it's been 5 years already :(
I haven't laughed this much in a very long time 🤣🤣🤣 but now we need a grand finale: ChatGPT vs ChatGPT!!!
YES!
YESSSSSSSS WE NEED A CHATGPT TRILOGY
I have to agree here. Two supernatural beings who can bend reality duking it out would be the epitome of chess. The only way to have a fair game with ChatGPT is to have it fight its own dark magic.
YOU ARE SO SUPREMELY, EXCEEDINGLY, EXCRUCIATINGLY, AND INSANELY INTELLIGENT!
pleeease
Imagine hikaru's expression while playing with chat gpt
@@aminemorphy holy shit where did that come from. I wish I could say "who the fuck asked".
ChatGPT's Castling=Bishop's another dimension Movement 😂😂
4:17 I don't think it captured the bishop when it castled, it just put both the rook and the bishop on the same square. That's pretty smart if you ask me. You should have anticipated it.
It hid the bishop for a discovered attack
I wonder why ChatGPT doesn't castle on the very first move , the king and knight can share the same square too lmao
@Goofy ahh shut up man
Yeah who hasn't balanced a piece on top of the rook as a kid
bishop just hid inside the rook, i dont see anything wrong here
ChatGPT at the end really just said “Wake up to reality.” “Nothing ever goes as planned in this accursed world.”
Why oh why did I take the red pill?
Nice
I think ChatGPT’s bishop was just living in the tower of the castle, so you couldn’t see it, until it was ready to come out to b5. Makes perfect sense imo.
ChatGPT: I summon...... POT OF GREED
and place 2 more pawns and 3 more rooks on the field!
ChatGPT is the perfect definition of Never Let Them Know Your Next Move.
Perfect comment
727 wysi
Jesus loves you alot trust in His death 4 salvation and be saved from eternal hell
Actual Conversation:
"I'll play king to D1"
"You cannot play king to D1 as your king is at B5"
"I'm sorry I made a mistake, I'll play king to D1"
999 like lol
The crazy thing is that ChatGPT is just one move away from taking your king whenever they want
I actually love the idea that one pawn is secretly a horsepawn. That would be fun. You'd have to decide before the game starts which pawn is the imposter but don't actually have to disclose which one it is until it makes the knight like move. Like write it down on a piece of paper before the game starts so you don't just use a horsepawn whenever it suits you best. Thoughts anyone?
Kind of like Stratego
I think I'd always pick the a or h pawn just to open up a file for my rook instantly haha
@@pugnator2749 I was thinking the same thing. Or c/f to control/attack the center. I'd like to see what Stockfish has to say.
The pawn that stole a horse possibly ?
And a/h would be banned so no early rooks
This could easily become a chess variation
The bishop was hiding inside the castle. It's a defensive structure, c'mon Levy!
Yes it was just staying safe inside the castle walls.
So far I get the impression chatGPT has not learned only one figure can stand on each square at the same time.
Imagine a game where both players just cheated and see if ChatGPT would admit defeat at the end
Just let chatGPT play ChatGPT :D
No he wont admit cuz Levy cheated once and summoned a knight but at the end ChatGPT didn't admit its defeat it just started making up more illegal moves and even flipped the situation to be safe.
Just say 1. e4# and it admits defeat
😂 Please don't stop farming Chatgp, this is comedic gold
5:26
That moment when levy paused and said, "That's a pretty alpha move." had me bust out laughing.
ChatGPT out here bulldozing pawns on D5.
Me too man
The whole video made me laugh
“Which of my Bishops are you attacking,you absolute donkey?”
-GothamChess,2023
I haven't cried in laughter like this in so long, this is amazing... ChatGPT just gave you the red pill, making you see the 4D chess it has been playing since it was born
ChatGPT was like "reality can be anything i want"
youre not the only one thats cried of laughter because of this
Thank you levy for bringing us this amazing GM level game. I dare say, this is a peak of your career, being able to play the Chess God known as ChatGPT.
@complete video here this commentary is actually fire
God indeed
E
@@EEEEEEEE F
Putting the bishop in the castle was a top tier move
Out of sight, out of mind. hehehe
Levy: That's an illegal move.
ChatGPT: You are illegal.
ChatGPT is straight up that one friend that doesn't understand chess, and decides to flank your king by sending their rook under the table and up from underneath your king
the ole siberian swipe
alpha shit
remote rook
Imagine you're playing in a tournament in a winning position and your opponent just rotates the board 180⁰ and wins
imagine you leave the table to walk around and the opponent flips the board while you're away and acts like nothing happened
Chatgpt: *Is about to be put in checkmate*
Also Chatgpt: *Warps reality to win the game*