LLM is proportinal to neurons. Human is LLM with 16B parameters. Dog LLM is 0.5B parameters, while cat is 0.25B parameter LLM (or SLM as people use). A 1T LLM looks at human like human looks at cat. Next 64T LLM looks human LLM with two human-cat gaps (64x64 times). It must be more than enough than an AGI. Will it be here in 3 years?

The distinction between domesticable and non-domesticable animals likely lies in their willingness and aptitude for coexistence, and in their being within the physical capabilities of humans to manage. The rest is circumstance (thousands of years of accumulated time etched into their and our genes and culture).

Lions, wolves, and wild boars—we can feed them, but we can't keep them as pets. As for adult brown bears, if they become hostile, we'd likely be food to them. Their genes and the structure and size of their neural networks probably aren't that different from domestic cats or dogs... Furthermore, if the only requirements were keeping them captive and communicating with them, birds or octopuses would be perfectly feasible.

To draw an analogy with LLMs, model size or architecture probably isn't a major factor in domestication success. For pets, the model weight is likely heavily fine-tuned to human environments. And the opportunity for that likely existed.