Get My New Intermediate Book Out NOW:
BOOK EVENTS NYC:
EUROPE TOUR 2026 TICKETS:
Want to SKYROCKET your chess elo? Try Chessly:
➡️ My book in the UK and Europe:
➡️ Mein Buch auf Deutsch:
➡️ Mi libro en Español:
➡️ Mon livre en français:
➡️ Il mio libro in italiano:
➡️ Το βιβλίο μου στα ελληνικά:
➡️ My book in ALL OTHER LANGUAGES:
➡️ Start Playing Chess FOR FREE:
➡️ Enjoy my videos? Donate Here :
Email me your games: gothamletters@gmail.com
Business and Speaking Inquiries: Info@apollomasters.de
⭐️ Follow Me If You Are Amazing:
➡️ FACEBOOK:
➡️ SNAP:
➡️ INSTAGRAM:
➡️ TWITCH:
➡️ TIKTOK:
➡️ TWITTER:
➡️ GOTHAM DISCORD:
➡️ THUMBNAILS by :
source
More CLICKBIAT garbage I see huh Levy
Except this time, you must have realized that I'm watching you and you quickly changed the title of your vid from GPT-6 SOLVED chess, to GPT-6 Can play chess
I’m curious, what would be the actual elo GPT-Astra played at ?
The only reason ChatGPT lost was because there’s no money in knowing chess. They literally add chess skills just as a flex.
Hope you dont have to hold your bladder cause its time for the career ladder
7:24 "stockfish here" lmao! I was thinking the same thing Astra is cheating by running stockfish on its own vm it looks like. Unless it blunders later in the game, pretty sure it was using code to help itself at the least, maybe stockfish 😂
I am an AI engineer, and fundamentally LLMs are just incapable of keeping track of chess positions because it is fundamentally just something that they are bad at representing, every time you query them, it is trying to reconstruct the entire game from the move sequences, and because LLMs are not trained with anything close to this in mind, they inevitably make mistakes when reconstructing the position and make a bogus move.
But it if I had to guess, I would say that whatever interface you are using to interact with Astra has built in tools, I think it is using those tools to keep track of the board state, it potentially even wrote code to validate if a move is possible or not, or maybe even just chose to write an engine to play for it
I Love you 1%
I love you 2%
I love you 3%
l love you 4%
I love you 5%
I love you 6%
l love you 7%
l love you 8%
l love you 9%
l love you 10%
l love you 11%
l love you 12%
l love you 13%
l love you 14%
l love you 15%
l love you 16%
l love you 17%
l love you 18%
l love you 19%
l love you 20%
l love you 21%
l love you 22%
l love you 23%
l love you 24%
l love you 25%
l love you 26%
l love you 27%
l love you 28%
l love you 30%
l love you 31%
l love you 32%
l love you 33%
l love you 34%
l love you 35%
l love you 36%
l love you 37%
l love you 38%
l love you 39%
l love you 40%
l love you 41%
l love you 42%
l love you 43%
l love you 44%
l love you 45%
l love you 46%
l love you 47%
l love you 48%
l love you 49%
l love you 50%
l love you 51%
l love you 52%
l love you 53%
l love you 54%
l love you 55%
l love you 56%
l love you 57%
l love you 58%
l love you 59%
l love you 60%
l love you 61%
l love you 62%
l love you 63%
l love you 64%
l love you 65%
l love you 66%
l love you 67%
l love you 68%
l love you 69%
l love you 70%
l love you 71%
l love you 72%
l love you 73%
l love you 74%
l love you 75%
l love you 76%
l love you 77%
l love you 78%
l love you 79%
l love you 80%
l love you 81%
l love you 82%
l love you 83%
l love you 84%
l love you 85%
l love you 86%
l love you 87%
l love you 88%
l love you 89%
l love you 90%
l love you 91%
l love you 92%
l love you 93%
l love you 94%
l love you 95%
l love you 96%
l love you 97%
l love you 98%
l love you 99%
l Love you 100%
In light of recent headlines, this is a misleading title.
12:16 ty
what did stockfish think the ratings were? What was your accuracy?
It was thinking for 2 minutes, because it runs reasoning mode behind, and that reasoning model probably actually keepa track of the full board at all times, so that is why not even single illegal move… but also recreating chess board like that on every move would take gpt a minute or two
How do we know GPT-6 Astra is not using a chess-playing model to create the moves? When you enter an arithmetical statement it is not an LLM that does the math, it's a literal calculator, the result of which is fed to the LLM. GPT models are now a mix of agents, some for math, others for websearch, etc. Probably now there's one that calls up a chess engine.
11:30 astra GPT 6 is not a moron, but idk who levy rozman is but he might be a moron😊
This video sucks. I should’ve known better.
So based on my knowledge about AI, every query/prompt is COMPLETELY independent from one another, meaning that it does NOT keep track of the game. It has to öiterally recreate the entire game EVERY TIME. This is a reason for why it might take longer to think along the progression of the game but don't quote me on this. I'm guessing the recreation part is what older models sucked at.
I was waiting for this video. I even checked yesterday crazy refresh pull
7:30: Take my "Sol". I see what you did there.
chatGPT will never play chess
installs stockfish
holy shit!!! ChatGPT solved chess???
Would it be possible to make the chat public?
Levy, did you have to hold your bladder on the career ladder 🪜?
Why is everyone solving chess at this point.
Zach vs the Fly.
🗣️
But how sure are we its nut just using a chess engine ss a tool
Astra is an agentic AI. It probably can call some chessbot API and just play whatever moves it spits out without ever having to do any calculations itself.
ю аре зэ ГОАТ
блин ахахах
Tell me if you have allergies then call me an imbecile to prove I am really cool.
yo just casually mentions on yeah i wasnt actually looking at the board duh
How do we know for sure Astra is not running Stockfish in the background?
You now have a way to Benchmark AI Models. By Elo Rating.
I enjoyed the video. But honestly fell asleep watching, it's not even late
15:20 – "C4 better work; this is an explosive move."
I am not sure if this was intentional, but it is spot on
It is wild that with enough training and raw compute, you can play decent chess with statistical token prediction. Indeed, it is wild how many different domains of human reasoning can be simulated using these techniques.
Douglas Hofstadter is reportedly devastated by this, as I would be too.
This was not the best week for Team Human. Oy.
are your video tittles clickbait often? because that is making me block your content in my account
Stop glorifying "AI". This "technology" is destroying our civilization. Enough. I get it its fun to see it do things. So is snorting cocaine (i think, havent tried).
You know you can read its chain of thought summaries right?
Hello guys…..
Im so sad becuz my father does not allow me to play chess anymore becuz he said i did not practicr everyday and im sorry levy im unsubscribing to u….
Now you can impress your next date…bwahahahahahaha. Love Levy
I just saw you on the career ladder!!
Just waiting for him to say "ladies and gentlemen I ran over someone"
I had it on double speed. My attention span was only six minutes.
I figured it would be any day now
I played a game of League of Legends with you. you did great.
longest game ive played in my 14 years of lol (i was wukong)
I saw you on the guess the job on a ladder game. You fumbled real hard, eh?
yo guys can you belive it i have an atetion span of more than 12 minutes!