Improvements can be supported by poor researchers with tenure.

Just try getting that past an ERB. Holy crow.

Simply saying 'you could share it with this guy' is not like the other case, because the two people haven't been called out as special. You don't share windfall with everyone you meet, why should this guy be different? In the DG, the other player was brought into the game and spent time waiting around just like you. It calls them out.

I'm trying to find a way of constructing a closely analogous situation without self-consciousness, and am having a tough time.

They borrow our machinery for consciousness - it's not clear to me that they aren't.

Also, it's rare for a dream to be so coherent that a transcript would convince an unimpaired (conscious) human.

o.O

They're testing the new thing, whatever that is, with the idea that if it's better, we go out and do that instead. The test is in the comparison.

The very idea that they're testing this with the existing program as a control sounds like a criticism of the existing method, saying, "We can probably do better than this!"

Seriously - if you want to convince people to change the educational regime, you had better include the current educational regime as one your experimental groups (whether or not it's given the name 'control'). If you don't, the experiment will not directly address the question of whether whatever else you were trying was any better than what we've got now.

But... doing the experiment in the first place is to test whether you ought to change that background. It's already an assumption of our society, and the people doing the experiments are already trying to do better.

, or the simulating entity has mindbogglingly large amounts of computational power. But yes, it would rule out broad classes of simulating agents.

If it's an ancestor simulation for the purposes of being an ancestor simulation, then it could well evaluate everything on a lazy basis, with the starting points being mental states.

It would go as far as it needed in resolving the world to determine what the next mental state ought to be. A chair can just be 'chair' with a link to its history so it doesn't generate inconsistencies.

You have a deep hierarchy of abstractions, and only go as deep as needed.

Why do you say that? The low-hanging fruit of good ideas tends to get plucked, even the long shots - the primary exceptions are things that people refuse to do because they're wrong.

Really? I'd think that routing things through subclasses of AbstractAction and then hooking them up as Listener to the various access methods would make it comparatively straightforward.