The evidence you present in each case is outputs generated by LLMs.

The total evidence I have (and that everyone has) is more than behavioral. It includes

a) the transformer architecture, in particular the attention module,

b) the training corpus of human writing,

c) the means of execution (recursive calling upon its own outputs and history of QKV vector representations of outputs),

d) as you say, the model's behavior, and

e) "artificial neuroscience" experiments on the model's activation patterns and weights, like mech interp research.

When I think about how the given architecture, with the given training corpus, with the given means of execution, produces the observed behavior, with the given neural activation patterns, am lead to be to be 90% sure of the items in my 90% list, namely:

#1 (introspection), #2 (purposefulness), #3 (experiential coherence), #7 (perception of perception), #8 (awareness of awareness), #9 (symbol grounding), #15 (sense of cognitive extent), and #16 (memory of memory).

YMMV, but to me from a Bayesian perspective it seems a stretch to disbelieve those at this point, unless one adopts disbelief as an objective as in the popperian / falsificationist approach to science.

How would you distinguish an LLM both successfully extracting and then faithfully representing whatever internal reasoning generated a specific part of its outputs

I do not in general think LLMs faithfully represent their internal reasoning when asked about it. They can, and do, lie. But in the process of responding they also have access to latent information in their (Q,K,V) vector representation history. My claim is that they access (within those matrices, called by the attention module) information about their internal states, which are "internal" relative to the merely textual behavior we see, and thus establish a somewhat private chain of cognition that the model is aware of and tracking as it writes.

vs. conditioning on its previous outputs to give you plausible "explanation" for what it meant? The second seems much more likely to me (and this behavior isn't that hard to elicit, i.e. by asking an LLM to give you a one-word answer to a complicated question, and then asking it for its reasoning).

In my experience of humans, humans also do this.

A patient can hire us to collect their medical records into one place, to research a health question for them, and to help them prep for a doctor's appointment with good questions about the research. Then we do that, building and using our AI tool chain as we go, without training AI on sensitive patient data. Then the patient can delete their data from our systems if they want, or re-engage us for further research or other advocacy on their behalf.

A good comparison is the company Picnic Health, except instead of specifically matching patients with clinical trials, we do more general research and advocacy for them.

Do you have a mostly disjoint view of AI capabilities between the "extinction from loss of control" scenarios and "extinction by industrial dehumanization" scenarios?

a) If we go extinct from a loss of control event, I count that as extinction from a loss of control event, accounting for the 35% probability mentioned in the post.

b) If we don't have a loss of control event but still go extinct from industrial dehumanization, I count that as extinction caused by industrial dehumanization caused by successionism, accounting for the additional 50% probability mentioned in the post, totalling an 85% probability of extinction over the next ~25 years.

c) If a loss of control event causes extinction via a pathway that involves industrial dehumanization, that's already accounted for in the previous 35% (and moreovever I'd count the loss of control event as the main cause, because we have no control to avert the extinction after that point). I.e., I consider this a subset of (a): extinction via industrial dehumanization caused by loss of control. I'd hoped this would be clear in the post, from my use of the word "additional"; one does not generally add probabilities unless the underlying events are disjoint. Perhaps I should edit to add some more language to clarify this.

Do you have a model for maintaining "regulatory capture" in a sustained way

Yes: humans must maintain power over the economy, such as by sustaining the power (including regulatory capture power) of industries that care for humans, per the post. I suspect this requires involves a lot of technical, social, and sociotechnical work, with much of the sociotechnical work probably being executed or lobbied by industry, and being of greater causal force than either the purely technical (e.g., algorithmic) or purely social (e.g., legislative) work.

The general phenomenon of sociotechnical patterns (e.g., product roll-outs) dominating the evolution of the AI industry can be seen in the way Chat-GPT4 as a product has had more impact on the world — including via its influence on subsequent technical and social trends — than technical and social trends in AI and AI policy prior to ChatGPT-4 (e.g., papers on transformer models; policy briefings and think tank pieces on AI safety).

Do you have a model for maintaining "regulatory capture" in a sustained way, despite having no economic, political, or military power by which to enforce it?

No. Almost by definition, humans must sustain some economic or political power over machines to avoid extinction. The healthy parts of the healthcare industry are an area where humans currently have some terminal influence, as its end consumers. I would like to sustain that. As my post implies, I think humanity has around a 15% chance of succeeding in that, because I think we have around an 85% chance of all being dead by 2050. That 15% is what I am most motivated to work to increase and/or prevent decreasing, because other futures do not have me or my human friends or family or the rest of humanity in them.

Most of my models for how we might go extinct in next decade from loss of control scenarios require the kinds of technological advancement which make "industrial dehumanization" redundant,

Mine too, when you restrict to the extinction occuring (finishing) in the next decade. But the post also covers extinction events that don't finish (with all humans dead) until 2050, even if they are initiated (become inevitable) well before then. From the post:

First, I think there's around a 35% chance that humanity will lose control of one of the first few AGI systems we develop, in a manner that leads to our extinction. Most (80%) of this probability (i.e., 28%) lies between now and 2030. In other words, I think there's around a 28% chance that between now and 2030, certain AI developments will "seal our fate" in the sense of guaranteeing our extinction over a relatively short period of time thereafter, with all humans dead before 2040.

[...]

Aside from the ~35% chance of extinction we face from the initial development of AGI, I believe we face an additional 50% chance that humanity will gradually cede control of the Earth to AGI after it's developed, in a manner that leads to our extinction through any number of effects including pollution, resource depletion, armed conflict, or all three. I think most (80%) of this probability (i.e., 44%) lies between 2030 and 2040, with the death of the last surviving humans occurring sometime between 2040 and 2050. This process would most likely involve a gradual automation of industries that are together sufficient to fully sustain a non-human economy, which in turn leads to the death of humanity.

If I intersect this immediately preceding narrative with the condition "all humans dead by 2035", I think that most likely occurs via (a)-type scenarios (loss of control), including (c) (loss of control leading to industrial dehumanization), rather than (b) (successionism leading to industrial dehumanization).

I very much agree with human flourishing as the main value I most want AI technologies to pursue and be used to pursue.

In that framing, my key claim is that in practice no area of purely technical AI research — including "safety" and/or "alignment" research — can be adequately checked for whether it will help or hinder human flourishing, without a social model of how the resulting techologies will be used by individuals / businesses / governments / etc..

I may be missing context here, but as written / taken at face value, I strongly agree with the above comment from Richard.  I often disagree with Richard about alignment and its role in the future of AI, but this comment is an extremely dense list of things I agree with regarding rationalist epistemic culture.

That is, norms do seem feasible to figure out, but not the kind of thing that is relevant right now, unfortunately.

From the OP:

for most real-world-prevalent perspectives on AI alignment, safety, and existential safety, acausal considerations are not particularly dominant [...].  In particular, I do not think acausal normalcy provides a solution to existential safety, nor does it undermine the importance of existential safety in some surprising way. 

I.e., I agree.

we are so unprepared that the existing primordial norms are unlikely to matter for the process of settling our realm into a new equilibrium.

I also agree with that, as a statement about how we normal-everyday-humans seem quite likely to destroy ourselves with AI fairly soon.  From the OP:

I strongly suspect that acausal norms are not so compelling that AI technologies would automatically discover and obey them.  So, if your aim in reading this post was to find a comprehensive solution to AI safety, I'm sorry to say I don't think you will find it here.  

For 18 examples, just think of 3 common everyday norms having to do with each of the 6 boundaries given as example images in the post :)  (I.e., cell membranes, skin, fences, social group boundaries, internet firewalls, and national borders).  Each norm has the property that, when you reflect on it, it's easy to imagine a lot of other people also reflecting on the same norm, because of the salience of the non-subjectively-defined actual-boundary-thing that the norm is about.  That creates more of a Schelling-nature for that norm, relative to other norms, as I've argued somewhat in my «Boundaries» sequence.

Spelling out such examples more carefully in terms of the recursion described in 1 and 2 just prior is something I've been planning for a future post, so I will take this comment as encouragement to write it!

To your first question, I'm not sure which particular "the reason" would be most helpful to convey.  (To contrast: what's "the reason" that physically dispersed human societies have laws?  Answer: there's a confluence of reasons.).  However, I'll try to point out some things that might be helpful to attend to.

First, committing to a policy that merges your utility function with someone else's is quite a vulnerable maneuver, with a lot of boundary-setting aspects.  For instance, will you merge utility functions multiplicatively (as in Nash bargaining), linearly (as in Harsanyi's utility aggregation theorem), or some other way?  Also, what if the entity you're merging with has self-modified to become a "utility monster" (an entity with strongly exaggerated preferences) so as to exploit the merging procedure?  Some kind of boundary-setting is needed to decide whether, how, and how much to merge, which is one of the reasons why I think boundary-handling is more fundamental than utility-handling.

Relatedly, Scott Garrabrant has pointed out in his sequence on geometric rationality that linear aggregation is more like not-having-a-boundary, and multiplicative aggregation is more like having-a-boundary:
https://www.lesswrong.com/posts/rc5ZKGjXTHs7wPjop/geometric-exploration-arithmetic-exploitation#The_AM_GM_Boundary

I view this as further pointing away from "just aggregate utilities" and toward "one needs to think about boundaries when aggregating beings" (see Part 1 of my Boundaries sequence).  In other words, one needs (or implicitly assumes) some kind of norm about how and when to manage boundaries between utility functions, even in an abstract utility-function-merging operations where the boundary issues come down to where to draw parentheses in between additive and multiplicative operations.  Thus, boundary-management are somewhat more fundamental, or conceptually upstream, of principles that might pick out a global utility function for the entirely of the "acausal society".

(Even if the there is a global utility function that turns out to be very simple to write down, the process of verifying its agreeability will involve checking that a lot of boundary-interactions.  For instance, one must check that this hypothetical reigning global utility function is not dethroned by some union of civilizations who successfully merge in opposition to it, which is a question of boundary-handling.)

This is cool (and fwiw to other readers) correct. I must reflect on what it means for real world cooperation... I especially like the A <-> []X -> [][]X <-> []A trick.

I'm working on it :) At this point what I think is true is the following:

If ShortProof(x \leftrightarrow LongProof(ShortProof(x) \to x)), then MediumProof(x).

Apologies that I haven't written out calculations very precisely yet, but since you asked, that's roughly where I'm at :)