We demonstrate that small pretrained foundational generative language models with millions of parameters can learn the latent rules of a process from data associated with the process. Inspired by Stefan Zweig's novella "Schachnovelle," also known as "The Royal Game" in English, we show that 28M and 125M parameter pretrained foundational small language models (SLMs) can be instruction fine-tuned with 1,000-to-1,000,000 examples to learn the rules of chess, propose legal moves, and accurately solve chess problems. We also explore the impact of successive language model fine-tuning epochs on improved outcomes and demonstrate reductions in model hallucinations by increasing the number of instruction fine-tuning examples.

Human preference or taste within any domain is usually a difficult thing to identify or predict with high probability. In the domain of chess problem composition, the same is true. Traditional machine learning approaches tend to focus on the ability of computers to process massive amounts of data and continuously adjust 'weights' within an artificial neural network to better distinguish between say, two groups of objects. Contrasted with chess compositions, there is no clear distinction between what constitutes one and what does not; even less so between a good one and a poor one. We propose a computational method that is able to learn from existing databases of 'liked' and 'disliked' compositions such that a new and unseen collection can be sorted with increased probability of matching a solver's preferences. The method uses a simple 'change factor' relating to the Forsyth-Edwards Notation (FEN) of each composition's starting position, coupled with repeated statistical analysis of sample pairs from both databases. Tested using the author's own collections of computer-generated chess problems, the experimental results showed that the method was able to sort a new and unseen collection of compositions such that, on average, over 70% of the preferred compositions were in the top half of the collection. This saves significant time and energy on the part of solvers as they are likely to find more of what they like sooner. The method may even be applicable to other domains such as image processing because it does not rely on any chess-specific rules but rather just a sufficient and quantifiable 'change' in representation from one object to the next.

Artificial intelligence hasn't taken over the world ... yet. But while humans can still outperform computers on most high-level intelligence tasks, at this point most people would concede the game of chess to the machines. The best chess-playing computer programs can already school just about any average human player, and they've proven capable of beating our grandmasters too. But maybe there's still some hope for us, even when it comes to chess. Scientists with the Penrose Institute have devised a unique chess problem that's fairly simple for humans to solve, but which seems to irreparably stump even the most sophisticated of chess programs.

A chess problem could help scientists finally unravel whether quantum theory can explain human consciousness. Sir Roger Penrose created the puzzle to prove the human mind can never be matched by a computer because it exhibits quantum effects. This means the brain doesn't follow the rules for the classical properties of matter, like a computer. Instead, it follows for a new concept of matter altogether that leaves cracks for consciousness and intuition to appear. Now, the Oxford university professor, who has set up a new institute, has invited everyday puzzle enthusiasts to pit their wits against the problem to test his theory.

We introduce a new artificial intelligence (AI) approach called, the 'Digital Synaptic Neural Substrate' (DSNS). It uses selected attributes from objects in various domains (e.g. chess problems, classical music, renowned artworks) and recombines them in such a way as to generate new attributes that can then, in principle, be used to create novel objects of creative value to humans relating to any one of the source domains. This allows some of the burden of creative content generation to be passed from humans to machines. The approach was tested in the domain of chess problem composition. We used it to automatically compose numerous sets of chess problems based on attributes extracted and recombined from chess problems and tournament games by humans, renowned paintings, computer-evolved abstract art, photographs of people, and classical music tracks. The quality of these generated chess problems was then assessed automatically using an existing and experimentally-validated computational chess aesthetics model. They were also assessed by human experts in the domain. The results suggest that attributes collected and recombined from chess and other domains using the DSNS approach can indeed be used to automatically generate chess problems of reasonably high aesthetic quality. In particular, a low quality chess source (i.e. tournament game sequences between weak players) used in combination with actual photographs of people was able to produce three-move chess problems of comparable quality or better to those generated using a high quality chess source (i.e. published compositions by human experts), and more efficiently as well. Why information from a foreign domain can be integrated and functional in this way remains an open question for now. The DSNS approach is, in principle, scalable and applicable to any domain in which objects have attributes that can be represented using real numbers.

Composition conventions are guidelines used by human composers in composing chess problems. They are particularly significant in composition tournaments. Examples include, not having any check in the first move of the solution and not dressing up the board with unnecessary pieces. Conventions are often associated or even directly conflated with the overall aesthetics or beauty of a composition. Using an existing experimentally-validated computational aesthetics model for three-move mate problems, we analyzed sets of computer-generated compositions adhering to at least 2, 3 and 4 comparable conventions to test if simply conforming to more conventions had a positive effect on their aesthetics, as is generally believed by human composers. We found slight but statistically significant evidence that it does, but only to a point. We also analyzed human judge scores of 145 three-move mate problems composed by humans to see if they had any positive correlation with the computational aesthetic scores of those problems. We found that they did not. These seemingly conflicting findings suggest two main things. First, the right amount of adherence to composition conventions in a composition has a positive effect on its perceived aesthetics. Second, human judges either do not look at the same conventions related to aesthetics in the model used or emphasize others that have less to do with beauty as perceived by the majority of players, even though they may mistakenly consider their judgements beautiful in the traditional, non-esoteric sense. Human judges may also be relying significantly on personal tastes as we found no correlation between their individual scores either.