ChatGPT, A Bad Chess Coach

I watched the tournament GothamChess conducted between AI chatbots. Pretty entertaining.

I was stuck at 700 for a good while, once I found a good opening that I could understand like the Queen's Gambit, the Italian, or London, a lot of the pressure in the first few moves was gone, all I had to focus on was puzzles. Try making the game as simple as possible for yourself, don't go for any crazy plans or openings for steady growth.

ChatGPT is very funny to watch ngl

I have no clue how anyone would get the crazy idea that ChatGTP would be a good chess coach ? ChatGTP is below the level of a two year old at the game. Doesnt even know the rules, teleports pieces, reincarnates pieces that are already gone ...

YouTube is extremely rich in good channels on chess.

GM Naroditsky has the best with his speedruns. Just amazing amount of good advice, very condensed, and demonstrated in practical games.

But there are countless others.

I use lichess and chesspuzzles to get puzzles to train tactics (for example daily).

Replying to @SacrifycedStoat, I have stopped playing Blitz. Now I play 20+0 games.

The problem with all LLMs is that they just scrape their datasets for words that appear together irrespective of the underlying ideas, and then use the algorithm to figure out what words should follow the prompt. So because a bunch of books and online posts they scanned say "castle the king" as beginner advice, it'll spit out "castle the king". But to your point, it doesn't actually know or think about anything. It sees a bunch of games that have e5 as a move, so it'll spit out e5 irrespective of the board state, whether it's the opening or a mid-game pawn break, whether it's legal or blocks a check, whether there is even a pawn or piece that can even move there, etc.

What does chatgpt recommend? Outside of chess, it seems to be helpful...

hey, I learnt chess from ChatGPT!!!

And now I take on 2000ELO players

see here https://www.chess.com/forum/view/game-showcase/overconfident-700-elo-meets-2000-elo-and-finds-out

When I was looking online for more specific details of the rules of stalemate by repetition a chatbot response advised that castling three times during a game can lead to a stalemate by repetition. That tells me all I need to know about the usefulness of AI bots when trying to learn anything more than absolute basics.

It is hard to figure out how intelligent LLMs are, since their exact parameters arent clearly described. AI is usually implemented as neural nets, which are an attempt to recreate how biological brains work in the computer.

The human brain is estimated to have around 90 billion neurons. We currently estimate that in order to emulate the behavior of a biological neuron in a neural net in the computer, you need about 100 neurons in such a neural net. So you need 9 trillion neurons in a neural net to reach the same complexity as the human brain, at least in regards to neurons. There are many other cells in the human brain which also may contribute. Also brain cells are amazingly well connected, since these connections grow naturally. Thats a huge factor why computers are so much less efficient than biological computers, they have fixed connections.

ChatGTP for example is said to have 1.8 trillion "parameters", whatever that means. So its complexity would be around 1/5 of a human brain ? In theory, anyway.

Additionally we have never observed AI to be creative either, and people have checked for that. Its still just a machine. There is no actual understanding in all these functions of anything. Its a complex math formula that makes an output from an input. The math formula was trained with certain data, and it cannot produce anything that wasnt included in the training, and of course noise.

So yeah, its not surprising that LLMs fail to understand chess. A two year old can easily be teached chess on a higher level than a LLM can show.

Hello everyone!

My way is simple...

Soviet or Russian chess books editions!

They are quite affordable,as far,as money is concerned.

You can take second-hand books that are not in the best condition-it is cheaper.

Learn from the source themselves,and what Capablanca would say...experiment is the best teacher!

youtube

Nice!

Of course nowadays we can access larger amounts of contents on internet, and more rapidly. Videos, texts, virtual chess boards to move the pieces around and practice favorite openings lines etc., all available for our learning, be it for free or not. And we all know that good and even great physical chess books have been dormant on the shelves worldwide for quite a while now. But last year a friend of mine passed on to me a 1974 soviet book that was treated as a family treasure, first owned by his father. A very flattering gest

So the chess vintage gods made this well preserved book entitled "How to Open a Chess Game" come to my humble hands. Not surprisingly for me, this one happens to be a special one.

As expected, it is rooted on real games played at high stakes scenarios by the writers themselves and by other GMs, most of them world champions (including Fischer and more non soviet players games). On the heavy weight team of writers and editors are Larry Evans, Petrosian, Larsen, Glicoric, Keres and a young Karpov, with the latter taking part in the editorial board. No need to mention what chess represented to these guys, with hard work and expertise abounding in a chess literacy that stems from the proficuous knowledge producing of the late-19th-early-20th centuries era. They meant to put together a high quality body of work, with a lot of theory, but in an easy to read manner, in order to make it accessible for begginers to professional readers. And despite the density of info, they succeeded, for the 1974 RHM Press edition translation into english make palpable the fluidity and lightness of their writings.

Glicoric's "Elements of Opening Strategy" chapter is just brilliant. The analysis and presentation of the Ruy Lopez he provides based on his own game against P. Bidev (with black) in Belgrade 1964, also with references to Fischer-Spassky WC match in Reykjavik 1972, really puts on the table the uniqueness and the power of the knowledge the great players develop through their singular experiences, linking chess theory, the tricky reality of the live chess board and finally their particular interpretations and conclusions. Tips, notes, entire paragraphs, all little jewels you cannot find everywhere* are countless here, as they are in the hole book.

*specially with today's crazy rithm of knowledge production and distribution, when increase in quantity use to become decrease in quality, demanding us to be critical learners.

In his Ruy Lopez section, Glicoric consolidates a pretty much timeless source of knowledge for those who wish to be better players at: (1) placing the pieces behind the pawns, as the author builds upon the "philidorian" bases to present decades of evolvement on notions like the "radiation" of power the pieces can maintain from this position; (2) playing with bishops and diagonals, expanding horizons - Ruy Lopez’s “soul”?; (3) pieces coordination for increase in their effectiveness; (4) the utter importance of time; (5) the fight for the center.

To consider, a priori, that an old book can no longer provide any info we do not already possess or could easily access on the web, on topics like fighting for the center or piece activity, is such a sad mistake.

You sure you don't want to hear from the masters and chat with GPT instead? Can we rely peacefully on the courses we bought on chessable or udemy, or even on our excellent youtubers of choice, when it comes to deep understandings of the game basics or nuances? Maybe yes. And maybe we think we have a wider or profound view of something, but in thuth we're just scratching the surface.

What I know for sure is that consensual great/classical/seminal physical books must not be taken as obsolete in the face of internet general sources. It is just a different type of source of information. It may require more time, but the knowledge and wisdom they bring are made to change our way of thinking in strong ways. And mostly enduring ones.

How do I figure out which books truly fit me?

Who is using chatgpt for chess lol

I'm stealing that one.