Fuck AI

Ah, now I get it. We aren't actually working from the same framework. By reasoning from first principles I take it you mean rationality/logic?

Yes, precisely.

The problem with that is that mathematics, logic, reasoning, and so on can't actually prove anything. If we used logic we can determine that no evidence can be definite as things like dreams, hallucinations, illusions and so on exist. The only conclusion you can really reach is that perhaps everything is made up, and you can't be certain anything is real in other words solipsism. That's where "first principles" come in I guess.

Ah yes, the old Cartesian demon who holds your consciousness imprisoned in a dream world making you question whether or not you exist at all. That's actually a very good, if not the perfect example of what I mean. I'm not sure how familiar you are with Descartes' Meditations, but outside the well-known realization of "I think, therefore I am", the method by which he defeats said demon is actually precisely the sort of thing I'm proposing.

To be more practical, what I was trying get at is basically the difference between having and being. Anything you have is likely to be temporary. Anything you are is likely to be constant. So you might ask yourself, are you autistic or do you have a condition called autism? If you can see the difference in perspective each statement offers, then you'll understand what I was on about.

You see, language is in fact even more basic a tool than reason and logic, because language is how we organize our perception of the world. Reason and logic simply arise out of language because language must have a certain structure in order to be meaningful at all and not just a random collection of words. LLMs clearly have the ability to learn that structure in a way that allows them to produce perfectly understandable sentences in any human language we choose to train them on, but they cannot really produce any good answers for questions that they haven't been specifically trained on.

Yes, they might still effectively hallucinate an answer anyways, and it might even sound correct, but unless you call them out on it when they start making stuff up, they won't even notice it happening. Clearly, they cannot actually reason through their own arguments, they simply produce something that imitates the human reasoning process well enough to pass muster approx. 90% of the time or so.

By first principles I assume you mean assumptions, as you won't get anywhere with logic without some kind of assumption. Since I don't know what your first principles are I am not going to be able to follow your reasoning, as if I would probably be starting with a different set of assumptions about the world.

As I tried pointing out above, a language model doesn't actually reason very well, it just imitates what humans do because it operates on prior knowledge acquired by its training. Meanwhile, humans have the ability, as Descartes' Meditations show, to throw away ALL of their prior assumptions about the world and start over from scratch, so to say, using only as much of their prior knowledge (i.e. the tools of language, logic, and reason) as strictly necessary, and in doing so, might reach new conclusions about the world that were previously inaccessible. Meanwhile an LLM will just make a wild-ass guess that seems to make sense, but often doesn't.

Generally though I don't believe logic/reasoning is a good tool for understanding people and things related to people like politics. It's good for bounded contexts with a well known state or rules like computers, or physical phenomenon. Depending on your worldview humans are either too badly known and too complex for another human to perform logic on them, or are simply not logical to begin with. Since it's not an effective strategy it's not something I am interested in using on people. I suspect a lot of disagreements where people are screaming at each other that the other isn't being logical come from having different assumptions rather than one being illogical.

Humans CAN be wildly illogical, that's true, but you can choose not to interact with such people (at least on the Internet, IRL it can of course sometimes be more difficult to do). Just like Descartes tests his demon, you can administer tests to them to see if they're willing to agree on some sort of shared ground rules for having a discussion that may be of mutual benefit, like we did in the previous comments and are continuing to do right now.

Again, a language model doesn't do that, it operates based on the rules it learned from its training corpus, and those are fairly fixed until you do another round of training that incorporates new information. Autism appears to be somewhat similar in that regard, in the sense that prior knowledge about how the world works (i.e. past exerience) is overweighted in comparison to what's actually happening (i.e. current experience).

Okay now you are saying thing with at least some degree of scientific evidence. The evidence for everything else you have said up until now has been pretty much "I made it the fuck up". I mean to be fair psychology isn't a real science and diagnostic categories are largely based on intuition rather than neurobiological evidence, so you aren't that far off.

I'm not sure if it's worth getting lost in the weeds of debating whether psychology is a real science or not, so I'm going to suggest we don't pursue that train of thought at the moment.

The LLMs I have worked with have been much more demure, they fairly easily admit they made a mistake (and can probably be coerced into doing so even if they actually haven't), and are willing to reason about political positions very different from their own liberal bias. Pretty much the opposite of stubbornness and debate bros. By being stubborn I am if anything behaving less like an LLM, as LLMs haven't been stubborn in my experience. Maybe you have had a different experience, if so I would like to here it.

To be fair, anything either of us has to say on this matter would likely fall under the category of circumstantial evidence. I for one certainly haven't done anything that could be considered scientific in this regard, and I am merely operating based on my memory of conversation I have either personally had, or have seen posted somewhere on the Internet.

Also the restricted and repetitive behavior thing is about special interests/hyperfixation. It's not actually applicable here as far as I know.

See my reasoning above for why I believe it DOES actually apply. I could be wrong, of course, but that's why I tried to explain how I arrived at this conclusion.

I don't think you have a modern understanding of neurodiversity or of neurotypes. A lot things that were once thought to be limitations of autistic people weren't limitations of autistic people at all.

I will freely admit that I haven't spent a huge amount of time familiarizing myself with the latest research on this, and I'm likely approaching it from a very different angle than you are, which might explain some of our difficulties communicating about this subject.

For example it was thought we lacked empathy by some psychologists (and still is) even though now we know of the double empathy problem. It's a incompatibility/communication issue, not an ability one. I would suggest you do some reading, then you might understand what I am getting at. It's also understood there are some limitations neurotypical people have that autistic people do not.

That's very interesting, and seems to validate my intuitive belief that autism is a condition that makes certain types of cognition more difficult, but not entirely impossible. Which means that with the right meds and/or mental effort, it may be possible to overcome it or at least greatly reduce the severity of its sypmptoms.

There was actually an interesting study done which showed that NT people don't behave morally when they aren't being watched, unlike autistic people who behave the same regardless of if they are being watched or not. The thing you said about most people behaving like NPCs is potentially one of those limitations of neurotypicals I am talking about here.

I have some interesting thoughts about that one, but it would require a rather lengthy explanation on where I'm coming from, so perhaps I'm going to save them for another time.

It's a shame you haven't been evaluated if that's something you wished for. Do you mind telling me what symptoms you think you might have? I understand if that's not something you want to discuss publicly or with me in particularly.

I'm pretty sure I have had all the symptoms I mentioned from the Wikipedia page at one time or another, and I continue to struggle with them from time to time. I also find it hard to make friends because most people seem to find my way of communicating exceedingly difficult, while I have had great difficulties with their tendency to make smalltalk.

That said, not sure what a diagnosis would do for me now, unless I was trying to get on disability benefits, perhaps. While it might have helped make my life a bit easier in the past, I'm somewhat concerned getting diagnosed now would just turn into an easy excuse for not making an effort.

There is also a tactic where people ask if someone needs help in a disingenuous way as a form of ad-hominem attack. Essentially calling some crazy while trying to make it sound like they are legitimately concerned. I don't think you are doing this, at least not intentionally, but I hope you understand that this could be read this way.

I'm certainly familiar with this tactic, but I don't think it HAS to necessarily be used nefariously, as it could just serve as a conversation starter. Perhaps it's a bit like asking someone to coffee after smashing a brick through their window, but I hope I have demonstrated enough sincerity so far as to not be credibl