The inherent flaw is that the dataset needs to be both extremely large and vetted for quality with an extremely high level of accuracy. That can’t realistically exist, and any technology that relies on something that can’t exist is by definition flawed.

That’s about consciousness, which is a much larger claim than the self being an illusion. You can have consciousness without a self, that’s what we call ego death. In theory, a conscious being could exist that’s always in a state of ego death, and have no understanding of the self and be utterly confused by why people find anything unintuitive about the teleporter paradox.

The solution that clears up all of these issues and results in a fully consistent view of the self is the one people like the least. There is no “you” or “me”, the self is an illusion the brain creates to make sense of things.

An LLM is a specific type of AI, and honestly one that’s a useless novelty at best and the death of information on the internet at worst. Machine learning has been used in chess for literal decades before LLMs, is Stockfish “a few if statements”?

Start with the win condition and work backwards from there, then as you learn each action you can take, put it into context for how it helps you. Then play a few turns against yourself to get an idea of the flow.

The Star Realms app is far more well-supported though, with all major expansions and all but the last 2 mini expansions available, while the Hero Realms app is just the base set with no apparent plans to support it further.

We need a use-it-or-lose-it clause for all copyrights. If the rights holder is not making a good faith effort to sell copies, they should forfeit their copyright entirely and the work in question goes straight to the public domain. 5 years is generous, I’d make the grace period 6 months.

As good as the Aeon’s End implementation is, it just makes me wish it had all the expansions available. I would sink so much time into a fully automated complete Aeon’s End collection.