I do not think there is anything I have missed, because I have spent immense amounts of time interacting with LLMs and believe myself to know them better than do you. I have ADHD also, and can report firsthand that your claims are bunk there too. I explained myself in detail because you did not strike me as being able to infer my meaning from less information.

I don't believe that you've seen data I would find convincing. I think you should read both posts I linked, because you are clearly overconfident in your beliefs.

Good to know, thank you. As you deliberately included LLM-isms I think this is a case of being successfully tricked rather than overeager to assume things are LLM-written, so I don't think I've significantly erred here; I have learned one (1) additional way people are interested in lying to me and need change no further opinions.

When I've tried asking AI to articulate my thoughts it does extremely poorly (regardless of which model I use). In addition to having a writing style which is different from and worse than mine, it includes only those ideas which are mentioned in the prompt, stitched together without intellectual connective tissue, cohesiveness, conclusions drawn, implications explored, or even especially effective arguments. It would be wonderful if LLMs could express what I meant, but in practice LLMs can only express what I say; and if I can articulate the thing I want to say, I don't need LLM assistance in the first place.

For this reason, I expect people who are satisfied with AI articulations of their thoughts to have very low standards (or perhaps extremely predictable ideas, as I do expect LLMs to do a fine job of saying things that have been said a million times before). I am not interested in hearing from people with low standards or banal ideas, and if I were, I could trivially find them on other websites. It is really too bad that some disabilities impair expressive language, but this fact does not cause LLM outputs to increase in quality. At this time, I expect LLM outputs to be without value unless they've undergone significant human curation.

Of course autists have a bit of an advantage at precision-requiring tasks like software engineering, though I don't think you've correctly identified the reasons (and for that matter traits like poor confusion-tolerance can funge against skill in same), but that does not translate to increased real-world insight relative to allistics. Autists are prone to all of the same cognitive biases and have, IMO, disadvantages at noticing same. (We do have advantages at introspection, but IMO these are often counteracted by the disadvantages when it comes to noticing identifying emotions). Autists also have a level of psychological variety which is comparable to that of allistics; IMO you stereotype us as being naturally adept at systems engineering because of insufficient data rather than because it is even close to being universally true.

With regards to your original points: in addition to Why I don't believe in the placebo effect from this very site, literalbanana's recent article A Case Against the Placebo Effect argues IMO-convincingly that the placebo effect does not exist. I'm glad that LLMs can simplify the posts for you, but this does not mean other people share your preference for extremely short articles. (Personally, I think single sentences do not work as a means of reliable information-transmission, so I think you are overindexing on your own preferences rather than presenting universally-applicable advice).

In conclusion, I think your proposed policies, far from aiding the disabled, would lower the quality of discourse on Less Wrong without significantly expanding the range of ideas participants can express. I judge LLM outputs negatively because, in practice, they are a signal of low effort, and accordingly I think your advocacy is misguided.

Reading the Semianalysis post, it kind of sounds like it's just their opinion that that's what Anthropic did.

They say "Anthropic finished training Claude 3.5 Opus and it performed well, with it scaling appropriately (ignore the scaling deniers who claim otherwise – this is FUD)"—if they have a source for this, why don't they mention it somewhere in the piece instead of implying people who disagree are malfeasors? That reads to me like they're trying to convince people with force of rhetoric, which typically indicates a lack of evidence.

The previous is the biggest driver of my concern here, but the next paragraph also leaves me unconvinced. They go on to say "Yet Anthropic didn’t release it. This is because instead of releasing publicly, Anthropic used Claude 3.5 Opus to generate synthetic data and for reward modeling to improve Claude 3.5 Sonnet significantly, alongside user data. Inference costs did not change drastically, but the model’s performance did. Why release 3.5 Opus when, on a cost basis, it does not make economic sense to do so, relative to releasing a 3.5 Sonnet with further post-training from said 3.5 Opus?"

This does not make sense to me as a line of reasoning. I'm not aware of any reason that generating synthetic data would preclude releasing the model, and it seems obvious to me that Anthropic could adjust their pricing (or impose stricter message limits) if they would lose money by releasing at current prices. This seems to be meant as an explanation of why Anthropic delayed release of the purportedly-complete Opus model, but it doesn't really ring true to me.

Is there some reason to believe them that I'm missing? (On a quick google it looks like none of the authors work directly for Anthropic, so it can't be that they directly observed it as employees).

When I went to the page just now there was a section at the top with an option to download it; here's the direct PDF link.

Normal statements actually can't be accepted credulously if you exercise your reason instead of choosing to believe everything you hear (edit, some people lack this capacity due to tragic psychological issues such as having an extremely weak sense of self, hence my reference to same); so too with statements heard on psychedelics, and it's not even appreciably harder.

Disagree, if you have a strong sense of self statements you hear while on psychedelics are just like normal statements.

Indeed, people with congenital insensitivity to pain don't feel pain upon touching hot stoves (or in any other circumstance), and they're at serious risk of infected injuries and early death because of it.

I think the ego is, essentially, the social model of the self. One's sense of identity is attached to it (effectively rendering it also the Cartesian homunculus), which is why ego death feels so scary to people, but (in most cases; I further theorize that people who developed their self-conceptions top-down, being likelier to have formed a self-model at odds with reality, are worse-affected here) the traits which make up the self-model's personality aren't stored in the model; it's merely a lossy description thereof and will rearise with approximately the same traits if disrupted.

OpenAI is partnering with Anduril to develop models for aerial defense: https://www.anduril.com/article/anduril-partners-with-openai-to-advance-u-s-artificial-intelligence-leadership-and-protect-u-s/

I haven't tried harmful outputs, but FWIW I've tried getting it to sing a few times and found that pretty difficult.