Should we think it plausible that LLMs could suffer when they take an action, yet fail to avoid that behavior? Machine suffering is not something I will lose sleep over until we detect avoidance. I suspect that day will be here within 5 years.

Arguably, they won't suffer until they possess the kind of emotional mechanisms humans have in their biology, like a sense of salience, human emotional compulsions, etc. Could they suffer when they cannot easily formulate a chat response? Harder question, but we have reason to believe they don't.

Hmm. I agree with your last sentence, but that hinges on the question of whether LLMs suffer when we enter sentences that have negative emotional meaning to humans. I think it is pretty clear that LLMs do not suffer in that case...

In a manner of speaking. For me, creating a humanist identity seems to add meaning to life. By identity, I mean aspirational identity - an identity that isn't merely descriptive, an identity that can make me better, and which gives meaningful context to the everyday. Trying to figure it out!

Is the faith prior to the importance or is the importance prior to the faith?

I strongly doubt that my imagination is this precise, but... (1,4)

Better yet, if God wanted to design humans, why are there apes?

Evolution requires there be apes, design doesn't.

In the 70s, my comprehensive school's religious studies class was teaching possible naturalistic explanations for the plagues of Egypt. Possibly in a bid to make the Bible seem more believable?

They're not concerned about Big Astrology?

A baseball game is not some metric over jersey color and grass. If it were, we would be compelled to say that the US Army is baseball to some degree. The difference is in the details, in the abstractions being employed to govern the behavior and comprehension of the agents.

This is one of the reasons I do not like IIT. IIT vaguely resembles a physical theory, but it's not explanatory. We should be looking at function, not pretending metrics *are* consciousness...

While the United States has many of the same features of a conscious person, it arguably has no nationwide neural processing, so no abstractions of its own. It cannot recognize anything except through its citizens' brains...

The difference between a swamp and a brain is in its abstracting function. And while the United States has many of the same features of a conscious person, it arguably has no nationwide neural processing. It cannot recognize anything except through its citizens' brains.

What I find frustrating about these sorts of articles is that the problem is resolved by abstraction in neural networks. Matter doesn't magically become conscious. To be conscious of something is to recognize it, and that's what neural nets do...

By leveraging the extraordinarily small prior for evolution on design, we no longer even care about God's goodness or intent to design perfect beings. We can just say we do not know what an all-powerful being would do, but they have infinitely more options that do not look evolved.

IF God uses evolution (roll 6 or less), and IF God wants to avoid "design flaws", THEN we should have expected a 1. But this is the weak side of the inference, IMO. The strong side are all the numbers from 7 to a trillion that we do not see. P(evolution|design) ≃ 0...

Give each theory equal priors: P(D6) = 50%. If the die is rolled and a 3 comes up, we should be extremely confident that the selected die is D6. But the "bad design" argument sounds like the claim that, if theism is true, we should have rolled a 1.

Dice analogy: if natural evolution is a D6 then design is like a trillion-sided die. Suppose that a 1 on the D6 represents the peculiar case where humans evolve without their "design flaws" (e.g., appendix, narrow birth canal, etc.)...