Good analysis, just a few nitpicks: AR is the future… when it matches human visual abilities, which may take several decades more, be it through glasses, or through a neural link. There is a deep uncanny walley in either case to overcome.

I mean we'll see. It's why I put it in the opinion section but I don't see AR taking off. Pokemon Go was fun for some but overall I don't see another AR game doing that level of success for a long time. Especially since Console, PC, and traditional mobile games do far better sales/profits than it.

LLM AI has already revolutionized AI, it’s the holy grail “glue” to keep different AIs working together, including itself.

Sorry, I meant game AI rather than general-use AI. I agree, for general use, it's done some interesting things but people try to use it everywhere and for everything. Like games and text generation in games and they create things like "convince this AI to give you a secret code" LLM AI and toss it in a game and it sucks. The biggest issue is players getting stuck in unintended ways which seems to happen constantly with it. Additionally, for AI response and choices, there are already better systems out there like Utility AI or GOAP.

Overall, I don't see LLM revolutionizing game AI in any area for it. Lastly, it doesn't seem to be that useful for math or pattern extrapolation/interpolation. An example I used with ChatGPT 3.5 was simply: O's value is 0, N is -1, and P is 1. What value is A and What value is Z? After some explaining that numbers can only be assigned once, it got a value for A of -14 (correct) and a value for Z as 25 because "Z is 25 letters after O" It took several prompts to get it to admit Z was 11. Lastly, making J = 0 and telling to pivot all values around J being 0, it messed up again.