General LLMs

General LLMs, such as ChatGPT, Gemini, and Claude, are designed to handle a wide range of topics and tasks. They have been trained on diverse datasets and can generate human-like text based on the input they receive. They are the craze nowadays with people heaping praise and others not impressed by what they can do.

They are cool to a certain degree  but they are horrible in specific domains like chess. That's where customised LLMs come in

Customized LLMs

Customized LLMs are specialized versions of general LLMs. Yeah they are similiar to chatGPT but  are fine-tuned or trained specifically on data related to a particular subject matter (eg: chess). For instance, a Chess LLM would be trained extensively on chess-related content, including game strategies, historical games, player techniques, and more. This enables the model to act as an expert in chess, providing insights, strategies, and potentially acting as a coach.They data sets will include input from coaches and sucking in the 100,000 chess books that are out there (with proper license and permission of course).

Current State and Future Prospects

As far as I know there are no Chess LLMs yet. But creating a highly specialized LLM like a Chess LLM is but aa heart beat away seeing custom LLMs are already in development and use in various fields, such as legal advice, medical information, and customer support. 

Current Capabilities: 

Existing LLMs can already provide substantial information on chess. They can analyze games, suggest moves, and explain strategies, although they may not be on par with dedicated chess engines like Stockfish or AlphaZero in terms of raw computational analysis. ChessLLMs can work in conjuction with engines to increase their scope.

Development Timeline: With advancements in AI and the increasing availability of domain-specific datasets, creating a more refined and expert-level Chess LLM could be realized within the next few years. Significant progress could be seen in as little as 2-5 years, depending on the resources and focus dedicated to the project.

Overall, the concept of a Chess LLM is well within reach and aligns with current trends in AI specialization.

It's been talked about on this blog and my YouTube channel many times.

One example scenario is a chess player just after a game in a tournament calling his chess coach telling it about the game but the AI chess coach has already seen it via live stream and already know the nuances of that game and giving advice immediately as well as give assuarance to the student. It can act like a real human coach and may even be more efficient keeping track of all the games, tailoring specific exercises to iron on weaknesses spotted and monitoring the effectiveness from results.