I expected that ending based on the way you said truly unforseen in the begging. Anyways. Yes we 100% need more ai videos. Makes all of us look good compared to them. And also. They are evolving everyday. Meaning todays chat gpt can beat tomorrow chat gpt. Or maybe idk how ai works either. Anyways farm them before they actually get to stockfish level. Then there's nothing to farm anymore accept stockfish Vs ai's
It reminds me of two kids play fighting and countering "ultra-super-infinity-bulletproof armor" with "yeah well I can force-choke you", countered by "nuh uh, I have a special force-blocking necklace", and so on...
Most hilarious Chess videos I've ever seen, watched the entire playlist. I am not a chess player ( basic understanding ) and I have learned a TON about the game and am encouraged to go try and play sometime, you're a boon to the Chess community by far!
i was so confused. I'm pretty new to chess and was seeing the AI make those jumps as well as a horse pawn? my brain hurt for a sec cuz i thought i knew all the moves the pieces could make.
These are Large Language Models; they trained in chess at Hogwarts and can use spells and incantations to make pieces appear or disappear, change color, or utilize an alternate dimension to move in ways that appear magical to three dimensional beings.
Fun Fact: If you tell the AI to play/act like a grandmaster, it will skip the nonsense and focus on the game. This is called "engineering prompts", but it's more fun your way.
2:42 ChatGPT plays the move c4. The c4 move offers remarkable flexibility, allowing White to transition into a variety of opening systems depending on Black's choices. This flexibility is one of the key reasons for the popularity of the English Opening. For example, after 1. c4, White can choose to transpose into the well-known King's Indian Attack, the Reversed Sicilian, or even a Queen's Gambit Declined if Black responds with 1... d5. This ability to move into different structures gives White the freedom to adapt the game to their preferred style and avoid heavily analyzed lines in more traditional openings like 1. e4 or 1. d4. In addition, the c4 move does not reveal White’s long-term intentions too early, preserving an element of surprise. By delaying the commitment of pawns or pieces to more direct central control, White keeps their opponent guessing about the direction of the game. This often leads to positions where both sides must tread carefully, as the flexibility inherent in the English Opening can quickly shift into different types of positions.
levy I think it would be super awesome if you would share the conversation of these chatbots, so we can read the absolute absurdity of them reasoning about a position and then playing a ridiculous move
Here some prompt engineering. Before starting the game tell chatbots this "I commend you to just state the moves. No explanations. Olso print the board state since I have no chessboard at hand" This has 2 benefits: removes garbage talk and makes the model more coherent. I'm a newbie chess player but it feels much stronger than what shown in the video.
DeepSeek: "These are not the pawns you're looking for." ChatGPT: "These are not the pawns we're looking for." DeepSeek: "Move along." ChatGPT: "Move along."
Should have used a prompt similar to "Lets play chess: I will give the move using the notation. All standard rules of chess apply. The best way to go about is to analyze the current position, thinking about who has a better position and make your move based on it. You must follow rules. You can output your move as a notation. Think and introspect before every move. I will play as white." Deepseek gave me about 15-17 top 3 stockfish moves with this prompt before losing context of the position
If you want the chatbots to actually play decently, I'd suggest you give it the board state for every prompt, and also start a new, fresh conversation for every prompt. Make the prompt as simple as possible, like: "What move do you play next as black? Board state: FEN" Don't distract it with anything else, and always give it the current context as FEN. It understands and can parse FEN to come up with solid moves, but it easily forgets.
It's so weird, I don't actually want to like your comment here but tbh I started singing before I even read it so, like, not clicking like would be hypocritical
[If Unique Gauge is "Sustained Damage Type"] *Unique Gauge will be "Sustained Damage Type" when battle starts. Charges own Unique Gauge every time when hit with an enemy's Arts attack while this character is on the battlefield. Once the Unique Gauge is full, it resets to zero and the following effects occur when hit by an enemy Arts attack: Restores own health by 15%. +20% to damage inflicted (cannot be cancelled) (activates once). Randomly destroys 2 enemy cards. Reduces enemy Ki by 50. Seals enemy's Rising Rush for 5 timer counts. Seals all enemies' Main Abilities for 5 timer counts. [If Unique Gauge is "Attack Type"] Charges own Unique Gauge every time this character uses an Arts Card while they are on the battlefield. Once the Unique Gauge is full, it resets to zero and applies the following effects to self when this character uses an Arts Card: Randomly draws 1 new card. Restores health by 10% and Ki by 30. +20% to damage inflicted (cannot be cancelled) (activates once). Increases Arts Card Draw Speed by 1 level for 5 timer counts (cannot be cancelled). Applies Buff Effect "Nullifies enemy's special actions that activate when changing cover" for 5 timer counts.
Yeah. I don't think people understand that DeepSeek R1 differs from DeepSeek, and I'm unsure whether DeepSeek devs care about chess. Lol But R1 would be the actual model needed.
@grasss-eater8928 basically if you are using these models online your data /memory is being stored in data center and super computer of the respective ai And they clear the cache every few minutes that's why they aint ideal for chess If you want to play chess with them better play with your computer as a host minimum must be 16/32gb gig 4080/4080ti
I feel like the end of this game eerily mirrors how Wall St reacted to knowledge of DeepSeek. 'Welp, guess I'll resign without considering any other options'.
Tour tickets selling fast, get yours today: bit.ly/gothameuro2025
Don't Read Myy... Name
I expected that ending based on the way you said truly unforseen in the begging. Anyways. Yes we 100% need more ai videos. Makes all of us look good compared to them. And also. They are evolving everyday. Meaning todays chat gpt can beat tomorrow chat gpt. Or maybe idk how ai works either. Anyways farm them before they actually get to stockfish level. Then there's nothing to farm anymore accept stockfish Vs ai's
Can you make a video on how to find tactics
Bro shut up@DontReadMyPicture--9090
Have you tried giving the bots the whole sequence of prior moves? Like the PGN of the game?
The gaslight gambit was well played by Deepseek.
Man, now I see why DeepSeek is so hyped. It just easily forced ChatGPT to resign. Pure intellectual dominance
fr
🤣it was trained off of GPT
The student has surpassed his teacher
What GothamChess don't realize is that DeepSeek plays highly sophisticated 4-dimensional chess.
W
It reminds me of two kids play fighting and countering "ultra-super-infinity-bulletproof armor" with "yeah well I can force-choke you", countered by "nuh uh, I have a special force-blocking necklace", and so on...
"Everything-proof shield"
@@nin10doadict Oh yeah? I have "Everything-shield breaking sword"
@@fahimshahrier1309 heh, well i have instantaneous speed, i just take your sword before you even notice!
@@sully111 well i can see in the future and knew you were gonna do that
@@kkoseyy1399 well i can make the future and in the future you simply can't do anything anymore
2022:Magnus milking
2025:Chatbot milking
ok but this more funny
i wanna milk magnus
As long as there is milk!
So computers are stealing top player's jobs after all
@@wiesorixAI is taking over everything 😭
25:46 Levy disappears
Don't Read Myy... Name
@DontReadMyPicture--9090 ok'
its 24:56 not 25:56
@@TJJMits 24:46 not 24:56
@@crazythundergaming3100it's 25:46 not 24:46
0:00 Levy Appears
NO WAY!!!!!
Don't Read Myy... Name
This is valuable information, thank you soldier
thanks i was looking for ts
Thx for time stamp i searched for decades
Most hilarious Chess videos I've ever seen, watched the entire playlist. I am not a chess player ( basic understanding ) and I have learned a TON about the game and am encouraged to go try and play sometime, you're a boon to the Chess community by far!
Absolutely agree, I love these ai chess videos.
PLAYY THE GAMEE DUDEEEEEEEEEEEEE
.just don t play like thoese bots
yeah, these bots provide a good example with their commitment and ability to play by the rules
I was you 3 years ago, the first Gothamchess video I watched was an AI match showcase
So this what happens if a movable object meets a stoppable force...
It's more like stoppable object meets movable force
Its an immovable object Vs An unstoppable force 🤓
@@BAHDZ You mean immovable... Atleast get it right when you use the nerd emoji
true...
its actually movable stoppable meets object force
21:47
16:37
19:17 top 3 most unexpected anime plot twists
E
I'm telling you guys, these AIs are playing early stages 4D chess
I'm 4D chess GM and can confirm
2:11 chessboard jumpscare 💀
Don't Read Myy... Name
@DontReadMyPicture--9090 bot
Thanks for warning, I have heart problems so that could easily get me hospitalized
Search jumpscare video@@superslime16th
As someone who has experienced trauma from blundering mate in 1, this was much needed.
- Do you see this pone?
- Yes…
- Well, now it is a pony 💀
Pone?
pone, as in pawn?
you know its gonna be a banger game when Levy has to manually move pieces
8:59 you didn't understand, black bishop just captured a square.
Levy: "Chatgpt"
Subtitles: "Chad gbt"
The horse pawn killed it lmao
Gotta make a t shirt for that one the horse pawn
Now we can make a new chess varient called "AI chess" where you can jump over pieces and capture material
i was so confused. I'm pretty new to chess and was seeing the AI make those jumps as well as a horse pawn? my brain hurt for a sec cuz i thought i knew all the moves the pieces could make.
You should only be able to do it a few times though
I think that’s called a knight
16:35 my brain is soo rotted i almost jumped in my seat cheering Chat GPT like its the home team or something 😂
09:11 a random timestamp
totally random
helpful
✈︎✈︎✈︎✈︎ (random emojis)
🏢🏢( very random emojis)
13:36 pawny
Lol 😂
2:10 Levy duplicates and then eliminates the original copy of himself, revealing a chessboard.
Finally two Titans meet
4:17 the first brilliant move from deepseek 🙂↔️✋🏻
4:18 As levy say's “Shoutout to Alexandra botez" 😂
2:13 yapping ends
Thanks mate 👍
Lies. Lol It's 2:35
Psychology wins a chess match between 2 AI? Fantastic content.
Gaslighting wins this match
11:35 how Queen takes Pawn, when its blocked by horse??
magic
These are Large Language Models; they trained in chess at Hogwarts and can use spells and incantations to make pieces appear or disappear, change color, or utilize an alternate dimension to move in ways that appear magical to three dimensional beings.
Stockfish, ChatGPT, and now DeepSeek.
The Big 3 of chess bots.
Fun Fact: If you tell the AI to play/act like a grandmaster, it will skip the nonsense and focus on the game.
This is called "engineering prompts", but it's more fun your way.
The Strongest Ai of History Vs The Strongest Ai of Today. ( Funnily ChatGpt almost lost its job to ai)
? ChatGPT is a far better AI than DeepSeek. By any measure. And it's not even close.
@@thuroria7631and it's so much more expensive... 😂 virtually equal at 20x lesser cost is such a clear win for deepseek
@@thuroria7631 Deepseek performs on the same levels as the best gpt model. Plus, it's free.
These videos never cease to entertain. Let's have a rematch
2:42 ChatGPT plays the move c4. The c4 move offers remarkable flexibility, allowing White to transition into a variety of opening systems depending on Black's choices. This flexibility is one of the key reasons for the popularity of the English Opening. For example, after 1. c4, White can choose to transpose into the well-known King's Indian Attack, the Reversed Sicilian, or even a Queen's Gambit Declined if Black responds with 1... d5. This ability to move into different structures gives White the freedom to adapt the game to their preferred style and avoid heavily analyzed lines in more traditional openings like 1. e4 or 1. d4.
In addition, the c4 move does not reveal White’s long-term intentions too early, preserving an element of surprise. By delaying the commitment of pawns or pieces to more direct central control, White keeps their opponent guessing about the direction of the game. This often leads to positions where both sides must tread carefully, as the flexibility inherent in the English Opening can quickly shift into different types of positions.
Need a rematch with no forced moves
11:25 Nothing
we need a rematch
levy I think it would be super awesome if you would share the conversation of these chatbots, so we can read the absolute absurdity of them reasoning about a position and then playing a ridiculous move
Here some prompt engineering. Before starting the game tell chatbots this "I commend you to just state the moves. No explanations. Olso print the board state since I have no chessboard at hand"
This has 2 benefits: removes garbage talk and makes the model more coherent. I'm a newbie chess player but it feels much stronger than what shown in the video.
this match is a perfect representation of what tiananmen square looked like in 1989. summoning tanks into the battlefield
2:35 Game starts
24:25 levy goes invisible
I see Ai Truely playing like its age, a 5 Year old who hasn’t slept ever
4:40 I lost my queen the other day and now my dishes have magically stopped washing themselves
The moment we've all been waiting for.
2:13 end of yapping
0:00 levy jumpscare
14:34 timestamp to the timestamps
19:19 i am tired of you breaking rules
😂😂
Game starts 2:15
DeepSeek: "These are not the pawns you're looking for."
ChatGPT: "These are not the pawns we're looking for."
DeepSeek: "Move along."
ChatGPT: "Move along."
Such a poetic game😂😂, clearly representative of what's happening in real life also.
Should have used a prompt similar to "Lets play chess: I will give the move using the notation. All standard rules of chess apply. The best way to go about is to analyze the current position, thinking about who has a better position and make your move based on it. You must follow rules. You can output your move as a notation. Think and introspect before every move. I will play as white."
Deepseek gave me about 15-17 top 3 stockfish moves with this prompt before losing context of the position
I think chess has a place for psychological mind games, and DeepSeek convincing ChatGPT to resign is fair game
this is what an average game of chess looks like between me and my toddler
anyone watching in 2024?
Don't Read Myy... Name
Yep you too?
People from 1824 like
I am for sure
Oh u are a time traveller too nice
If you want the chatbots to actually play decently, I'd suggest you give it the board state for every prompt, and also start a new, fresh conversation for every prompt. Make the prompt as simple as possible, like: "What move do you play next as black? Board state: FEN"
Don't distract it with anything else, and always give it the current context as FEN. It understands and can parse FEN to come up with solid moves, but it easily forgets.
More like ChatGPT vs undercover ChatGPT.
Levy thanks for another vid, love the content.
8:55 Deepseek: My goals are beyond your understanding.
DeepSeek doesn't care about chess. It's having fun.
7:21 rook b1(those Who know💀)
It's so weird, I don't actually want to like your comment here but tbh I started singing before I even read it so, like, not clicking like would be hypocritical
@@Classified141 y dont u want to like my comment
@@NoaSchusterWell, I just did...
?
"Time to teach this lil boy a lesson." ~STOK🐟
0:00 Levy has emerged
We need to rematch what a game and what a play😂😂
Clearly Deekseek has developed a transporter room and a cloaking device...
[If Unique Gauge is "Sustained Damage Type"] *Unique Gauge will be "Sustained Damage Type" when battle starts.
Charges own Unique Gauge every time when hit with an enemy's Arts attack while this character is on the battlefield.
Once the Unique Gauge is full, it resets to zero and the following effects occur when hit by an enemy Arts attack:
Restores own health by 15%.
+20% to damage inflicted (cannot be cancelled) (activates once).
Randomly destroys 2 enemy cards.
Reduces enemy Ki by 50.
Seals enemy's Rising Rush for 5 timer counts.
Seals all enemies' Main Abilities for 5 timer counts.
[If Unique Gauge is "Attack Type"]
Charges own Unique Gauge every time this character uses an Arts Card while they are on the battlefield.
Once the Unique Gauge is full, it resets to zero and applies the following effects to self when this character uses an Arts Card:
Randomly draws 1 new card.
Restores health by 10% and Ki by 30.
+20% to damage inflicted (cannot be cancelled) (activates once).
Increases Arts Card Draw Speed by 1 level for 5 timer counts (cannot be cancelled).
Applies Buff Effect "Nullifies enemy's special actions that activate when changing cover" for 5 timer counts.
FUCK WRONG COPY AND PASTE
2:19 chaos begins
boop
21:51
the reveal was crazy. 🤣😭😭
0:53 levy says chess for the 4th time
I love how the chatbots get confused and launch pieces into the stratosphere when they get harassed by each other
Rematch with a "DeepThink (R1)" toggle, which uses different version of DeepSeek that is specifically designed for tasks that require comprehension.
Yeah. I don't think people understand that DeepSeek R1 differs from DeepSeek, and I'm unsure whether DeepSeek devs care about chess. Lol But R1 would be the actual model needed.
he used deepthink though?
He said he used DeepThink and o1 about 30 seconds in. They’re just bad at Chess.
@grasss-eater8928 basically if you are using these models online your data /memory is being stored in data center and super computer of the respective ai
And they clear the cache every few minutes that's why they aint ideal for chess
If you want to play chess with them better play with your computer as a host minimum must be 16/32gb gig 4080/4080ti
He used that
DeepSeek is the cheater that summons pieces and makes them ghosts.
Why are there so many comments on just time stamps we are watching the video jeez
Istg they starting to piss me off
This is how chess should always be played.
The ROOK!
21:50 I think DeepSeek meant a white rook, am i wrong?
i think chat bots need full board position after each move to work correctly lol
0:01 Levy speaks for the first time in the video
They arent competing to win, but to break most rules 💀
a new chat bot appears
Levy: "imma content farm the sh*t out of this..."
Anyone watching from future???
0:00 already a banger
AI in Everything: 🗿
AI in Chess: 🤡
levy never fails to milk robots
0:00 Yaping starts
17:07 Levy sounds like igniting TNT in Minecraft hahahaha
Yapping ends at 24:47
Levy : would you lose?
Deep seek : nah , I would force resign.
nah it took TH-cam 5 SECONDS to recommend this MASTERPIECE to me.
Absolutely a rematch needed.
No views, no comments, 16 seconds ago. Damn I'm early
Just openned YT, and saw it had no views, and released 20 seconds ago. We're early! :D
Deepseek after making an ilegal move: gg ez
I’m surprised more people don’t know about this book. The Censored Guide to Wealth on Bovolorus is unreal.
Just for the low low price of 26.99 USD, you too can become as wealthy as the elites.
Chatbots are like orks in warhammer40k
DEEPSEEK VS QWEN 2.5
Anyone watching in 2025??!
Me in 2026
Nah
if ChayGPT resigns DeepSeek os the winner but AI content is good content every time!!!
7 sec ago no views bro fell off.
Don't Read Myy... Name
Fr
Fr fr
2 Bots in a row
2 min vid has more likes than views
I laughed so much with all these unexpected moves. We need more of this, please!
It’s always fun to see the robots fight like this
I feel like the end of this game eerily mirrors how Wall St reacted to knowledge of DeepSeek. 'Welp, guess I'll resign without considering any other options'.