If the LLM never gets a chance to try to work around the block then this is more likely to work.
Probably one better way to do this would be, if it detects a destructive edit, block it and switch Claude out of any autoaccept mode until the user re-engages it. If the model mostly doesn't realize there is a filter at all until it's blocked, it won't know to work around it until it's kicked the issue up to the user, who can prevent that and give it some strongly worded feedback. Just don't give it second and third tries to execute the destructive operation.
Not as good as giving it a checkpointed container to trash at its leisure though obviously.
There is some underlying reality to what autism is, even if we do not have a good understanding of it; and even if turns out to be multiple unrelated things that happen to have similar symptoms.
Of the people with those actual conditions, it seems entirely plausible that some will not be hindered.
The authors of the DSM-V needed to create a diagnostic criteria for a condition that they do not understand, and for which no objective test is known. Further, their objective was designing something useful in a clinical setting. Giving those constraints, saying "if it is not a problem, we don't care about it" is entirely reasonable; despite not being reflective of the underlying reality.
This study suggests that there are several different things called "autism". That's because "autism" as a term is not about some underlying reality, but a bucket that a bunch of people get tossed when some medical professionals see them as similar. And they come to the attention of those medical professionals because those people either say they have a problem or are called a problem by others.
But a problem with a person is always about a person in a context. Blue-eyed people are hindered by their eyes in bright light. Do we call that a genetic disease and look for cures? Not here, because there are enough "normal" people with blue eyes. But if it was just 1 in 20,000 people with blue eyes, it'd surely be treated as a disease.
Or we could imagine a "Height Deficiency Syndrome" characterized by inability to reach the top shelves in a normal house. With an effort, we could surely cure this impactful genetic problem through early application of hormones and the use of new CRISPR-related technologies. Or we could look at it as normal human variation which only "hinders" people because of how our society is set up to cater to "normal" people.
But we thankfully now have a term for that sort of nonsense: medicalization of deviance.
As I commented in another thread, there's no a priori reason to believe that the "average" glutamate receptor level is the "right" one. Isn't it possible that there are:
1. "Normal" people with a level of glutamate receptors at 10, say, on a scale I'm inventing for this example
2. "Autistic" (according to the DSM) people with a level of, say, 5, who are hindered by the effects of being at this level
3. "A little bit autistic" people at a level of, say, 8, who aren't hindered and don't meet the DSM criteria, but in fact actually benefit from the effects of being at this level
Some "normals" might then want to inhibit their glutamate receptors somewhat to get the benefits of being at an 8 or a 9 on my made-up scale.
Perhaps. But remember that this is a very complex 3D structure with varying receptor densities, it's not "The Glutamate Level", it's some neural network areas with higher or lower excitability connected to other neural networks.
Just like with ADHD it's likely that medication will at best have limited effectiveness and many side effects.
Certainly, we're at the "bash it with a hammer" stage not ready for anything nuanced. I just wouldn't want to assume that the right outcome is "less autism"; I suspect most people could do with at least a little more!
That‘s why we have so many late diagnosed. People who are on the spectrum but were able to mask or were just lucky until luck runs out. Then it becomes a problem and a diagnosis. I knew I am different as long as I can remember. It was obvious in Kindergarten and also in every type of school and later in work. I‘m an old millennial and nobody was trained back then in the 80/90s. Before it became a diagnosis and before awareness started to rise, people unalived them, died homeless or in prisons/wards.
Maybe they meant neurodivergent as a broader category? Like "some people are neurodivergent but don't have autism"
That would be a bit weird though...
EDIT: Neurodivergent is very much a broader category. What I meant would be weird is to state the obvious... Very much sounded like they were trying to say some people with autism may not want to get "cured" but using the wrong words
Perhaps your thinking on this lacks grey areas. A healthy percentage of extremely successful people in computing are referred to as “on the spectrum” - are these people helped by having some of the aspects of autism or hindered by it? Why do we need to have a diagnosis for people to have aspects of this pathology?
To the level of a clinical diagnosis, yeah it seems quite likely to me that most people can’t discern autistic spectrum behaviour in their peers. I bet most people couldn't even accurately say what those behaviours would be.
A clinical diagnosis isn't the only way to look at what's going on here. We can have differences that aren't medical problems. Differences that are measurable and nameable, even. Those categories can overlap with or be congruent to medical terms while still being valid and useful.
Definitely nobody in this thread struggling to see the grey areas and wanting to make sure everything is very cleanly defined, as if it’s difficult for them to deal with situations that are outside of rigorously defined clinical diagnostic criteria, for example… BTW just to be crystal clear - I’m obviously making a silly joke here it’s not intended to be serious :-D
Clinical autism and clinical ADHD are notoriously difficult to diagnose in adults. In some countries it's even illegal to prescribe stims unless there's a childhood ADHD diagnostic.
Adults have been socialised to mask the more problematic behaviours, and they can also be unaware that what they're doing is masking: they can believe that everyone struggles like that.
Only difficult because the criteria are misaligned. We diagnose school children more consistently, because we subject school children to strict measured criteria (school), and can point to the data (grades/homework) as objective evidence.
Why do we care so much about objective evidence? Because of prohibition. Prescribing stimulants isn't illegal because it is difficult to diagnose ADHD. It's difficult to diagnose ADHD for the very same reason it's illegal to prescribe stimulants: our society values prohibition of drugs over actual healthcare. An ADHD diagnosis implies a compromise of prohibition, so our society has structured the means to that diagnosis accordingly.
Experts in the field estimate a very high incidence of undiagnosed ADHD in adults. During the height of the COVID-19 epidemic, telehealth services were made significantly more available, which lead to a huge spike in adult ADHD diagnoses. Instead of reacting to that by making healthcare more ADHD accessible, our society backslid; lamenting telehealth providers as "pill mills", and generating a medication shortage out of thin air.
The DSM-V criteria are not a good description of the natural category, and most people don't actually use them. They are, at best, a vague gesture in the direction of the natural category. The ICD-11 criteria (6A02) are better, but are still contradicted by, for instance, studies evidencing the double-empathy problem. Trained psychologists know which diagnostic criteria to take literally, and which to interpret according to the understanding of the authors.
If someone doesn't have any deficits or impairments at all then they won't qualify under ICD-11 either:
"Deficits are sufficiently severe to cause impairment in personal, family, social, educational, occupational or other important areas of functioning..."
Virtually none of the definitions in the ICD or DSM are entirely correct: that doesn't mean they're not useful. For example, you stop meeting the literal diagnostic criteria of many conditions if they're being treated adequately, but that doesn't mean you no longer have those conditions. Someone on antiretrovirals with no detectable HIV viral load still has HIV, and still needs to take the antiretrovirals. No competent doctor would diagnose them as "cured". Yet, they would not meet the diagnostic criteria described in the ICD-11:
> A case of HIV infection is defined as an individual with HIV infection irrespective of clinical stage including severe or stage 4 clinical disease (also known as AIDS) confirmed by laboratory criteria according to country definitions and requirements.
and rarely they may never have met these criteria. This is HN, so a computer analogy might be more helpful: ask a non-technical friend to read through some of the POSIX.1-2024 spec, then ask them to explain the signal handling, or the openat error codes. They will totally misunderstand it, because the POSIX specs are not actually clear: their purpose is to jog the memory of the expert reader, and describe the details they might have forgotten, not to provide a complete and accurate description suitable for teaching.
(Edit: pointless confrontational passage excised. Thanks for the criticism.)
seems a bit confrontational, unless you yourself are a trained psychologist, in which case it would seem fitting to volunteer those credentials along with this challenge.
They still are an individual with HIV infection, except that it is in the stage of "remission" or "undetectable" but they have previously been diagnosed with HIV at a different clinical stage.
So the definition is perfectly correct, assuming you know what "clinical stages" there are.
Why is someone on HIV antivirals if no test ever confirmed them to have HIV? Presumably, they were confirmed as having HIV and have reduced its load to beneath detectable levels but that doesn't erase the previous confirmation.
I think that's all an aside, though, if not the ICD (as suggested by another poster) or the DSM definition initially used, which definition is correct?
OP, I think, is clearly harkening back to a previous post on HN (article at: https://www.psychiatrymargins.com/p/autisms-confusing-cousin...) by a professional discussing that the public often misunderstands and ignores key aspects of the definition. This seems rather a bit like you pointing out laypeople might read and not understand what they got out of the POSIX.1-2024 spec. Except it seems you're suggesting instead that the layperson understanding is correct.
> Why is someone on HIV antivirals if no test ever confirmed them to have HIV?
Mu. If it was confirmed, but not "confirmed by laboratory criteria according to country definitions and requirements", then they do not meet the diagnostic criteria (interpreted literally). Suppose, for instance, that there was a procedural error that might have messed up the diagnosis (so is forbidden by regulation), but in this case didn't mess up the diagnosis.
I can produce as many of these literally-correct, deliberate misinterpretations as you like. They have no bearing on actual medical practice.
> which definition is correct?
Which definition of "carbon atom" is correct? Our definitions have, for 200 years, been sufficient to distinguish "carbon atom" from "not carbon atom", but those definitions have changed significantly in that time. Autism is that category into which autistic people fall, and into which allistic people do not fall, which is distinguished from several other categories with which it is often confused. (The ICD-11 spends way more words on distinguishing autism from OCD, Tourette's, schizophrenia, etc than on defining it directly.)
The autism itself, depending on the person, is often less of a problem than societal expectations. For example- in a world where everyone was red/green colorblind, such a condition would not be considered a handicap. And in a world where everyone was autistic, many things would be different.
Society punishes us severely for not being able to see the difference between red and green, to use that metaphor. And they seem to expect that if they punished us just a little harder, we would suddenly become normal. Thats the big problem. Non conforming behavior is always treated as a crime or offense on some level, but we cannot conform, and therefore must adjust to a life of endless punishment doled out by both authorities and peers.
Its quite difficult to go through life that way without developing a negative self image. This goes for people with autism, adhd and other types of neurodivergence.
It's like being tortured to extract information that we do not have. They'll only believe it once they've completely broken you down.
And then you meet the next person, who has not yet tortured and broken you, so they again do not believe that you "don't have the intel", and you get to go through it all over again.
The worst part is when you start believing for yourself that they're right, that you're holding back, and that it's all your fault for not giving them what they want, just for the life of you you can't figure out how.
Getting certainty about my condition did so much to heal me.
But going by the strict notion of DSM-V criteria of providing a hindrance, we hit the somewhat problematic definition whereby a person can have autism at one point in their life (when it hinders them in a context), moves into another point or context in their life (where it does not) and therefore they do not or would not meet the criteria for having autism if they sought a diagnosis at that point in time, and then move back into another point or context in their life where it hinders them and so now they meet the criteria and presumably have autism again.
Now, needless to say, this is not how anyone actually thinks about psychiatric or psychological issues in practice, especially with conditions such as autism, and just highlights the relative absurdity of some of the diagnostic metrics, practices and definitions.
What we tend to do is tie the diagnosis of autism to the individual identity and assume that it is a consistent category and applicative diagnosis that stays with a person over time because it is biological. We know, of course, that this is despite not having any working biological test for it, and diagnosing it via environmental and behavioural contexts. And don't even get me started on tying in diagnosis of aspergers/autistic individuals with broadly differing abilities and performance metrics on a range of metrics under the one condition such that the non-verbals and low-functioning side of neurotypicals get lumped in with the high iq and hyper-verbal high-functioning aspergers as having the same related condition even though neurotypicals are closer to the non-verbals and low-iqs on the same metrics and scores.
The entire field and classification system, along with the popular way of thinking about the condition is, if i might editorialise, an absolute mess.
A person without legs does not stop being disabled because they have no need or desire to walk. The fact remains that should they need or desire to walk in the future the hinderance will still very much exist.
A similar example could be made of someone with gluten intolerance. If they do not eat foods that contain gluten they are still gluten intolerant. They are however still disabled by needing to stay in that situation.
Firstly a fish without legs objectively does not have legs, but we do not necessarily call it disabled, even though it clearly lacks a facility.
Secondly, the autism spectrum disorders are, as I previously mentioned, not obviously just about deficits of behaviours or functions but also can take in extended and exceptional abilities in some areas and greater sensitivities rather than deficits or lack of an ability, so it is not clear that the entire diagnosis can be defined by deficits or lacking things. The high functioning and Asperger's type diagnosis is not about a universal deficit diagnosis and we do not generally call neuro-typical humans disabled because they lack prodigious activity or interest in math, language, or other subjects, even though that can also objectively be measured and called a deficit.
> The high functioning and Asperger's type diagnosis is not about a universal deficit diagnosis
To get an Asperger's diagnosis under the DSM-IV you needed some amount of impairment. "Disorder" is in the title of the DSM, if something isn't conceptualized as a disorder it isn't in there.
Being reliant on a particular life situation does strike me as a hindrance in and of itself. Maybe more of a macro limitation than a day-to-day one, but a reasonable definition could encompass that, too.
> This is more or less not true. If it doesn't hinder a person in any aspect of their life, they don't fit the DSM-V criteria for a diagnosis.
You're confusing autism itself with Autism Spectrum Disorder. Autism Spectrum Disorder indeed has to do with difficulties ("deficits" / "impairment"). Autism itself on the other paw is a physical, quantifiable difference in neural architecture. Autistic people think and work differently, whether they have been diagnosed with Autism Spectrum Disorder or not.
It's also worth noting that autism is not the only neurodivergence, it's just the most widely known one (IIRC).
For reference, my copy of the DSM-5 states the following diagnostic criteria for Autism Spectrum Disorder: (sub-items elided)
> A. Persistent deficits in social communication and social interaction across multiple contexts, as manifested by the following, currently or by history (examples are illustrative, not exhaustive; see text): [...]
> B. Restricted, repetitive patterns of behavior, interests, or activities, as manifested by at least two of the following, currently or by history (examples are illustrative, not exhaustive; see text): [...]
> Symptoms must be present in the early developmental period (but may not become fully manifest until social demands exceed limited capacities, or may be masked by learned strategies in later life).
> D. Symptoms cause clinically significant impairment in social, occupational, or other important
areas of current functioning.
> E. These disturbances are not better explained by intellectual disability (intellectual developmental disorder) or global developmental delay. Intellectual disability and autism spectrum disorder frequently co-occur; to make comorbid diagnoses of autism spectrum disorder and intellectual disability, social communication should be below that expected for general developmental level.
The autism in this study is ASD. This study doesn't have that much to say about people who don't qualify for a diagnosis, since they would not have qualified to take part in it.
I think the whole point of this was to see if the "agents" could act like a real human and real humans use Gmail much more frequently than sendmail. Sage even commented that they had update their prompt to tell the agents to not send email and not just remove the Gmail component for fear that the agent would just open it's own Gmail (or Y! mail, etc.) account and send mail on it's own.
That is really interesting and does suggest some new questions. I would claim it does not change who is responsible in this case, but an example of a new question: there was a time when it was legally ambiguous that click-through terms of service were valid. Now if an agent goes and clicks through for me, are they valid?
Unfortunately, the sheer amount of ChatGPT-processed texts being linked has for me become a reason not to want to read them, which is quite depressing.
The Readme is substantially LLM generated, yeah? Something about LLM readmes leave me cold. Including stuff like this feels like the sort of typical LLM time-wasting stuff that they output these days:
Add 3D Model: Click "3D Model" → "Add 3D model (.glb)" → Select your file
Add Image: Click "Image" → "Add image" → Select your file
Add Text: Click "Text" or press T to add 3D text
To add a 3d model I click "3d Model" and then "add 3D model" and then add my 3d model. Very clear, but not usually what I look for in a readme. LLMs love this sort of stuff though.
Yes, sorry about that if it annoyed you. I was too busy to write a good human-sound readme. It was an old project, and I wanted to release it as open source as quickly as possible.
I think I don't mind LLM generated documentation per se if they're marked as such up front! It's more when I get halfway through and realize that this is probably an LLM output that I get annoyed.
I will preemptively grant the narrow point that if a project demonstrates poor quality in its code or text (i.e. what I mean when I say "slop"), it can dissuade potential users. However, the "Why AIsbom?" section strikes me as clear and informative.
Many people prefer human writing. I get it, and I think I understand most of the underlying reasons and emotional drives. [1]
Nevertheless, my top preference (I think?) is clarity and accuracy. For technical writing, if these two qualities are present, I'm rarely bothered by what people may label "AI writing". OTOH, when I see sloppy, poorly reasoned, out-of-date writing, my left hand readies itself for ⌘W. [2]
A suggestion for the comment above, which makes a stylistic complaint: be more specific about what can be improved.
Finally, a claim: over time, valid identification of some text as being AI-generated will require more computation and be less accurate. [3]
[2]: To be open, I might just have a much higher than average bar for precision -- I tend to prefer reading source materials than derivative press coverage, and I prefer reading a carefully worded, dry written documentation file over an informal chat description. To keep digging the hole for myself, I usually don't like the modern practice of putting unrelated full-width pictures in blog posts because they look purdy. Maybe it comes from a "just the facts, please" mentality when reading technical material.
[3]: I realize this isn't the clearest testable prediction, but I think the gist of it is falsifiable.
The good cops, such as they are, get run out if they try to challenge the institutional problems in police forces. This radically restricts how good a cop can be while still being a cop.
Can good cops speak up about bad cops and keep their job, or do they have to remain silent? How many bad things can you see in your workplace without quitting or whistleblowing while still being a decent person? Can they opt out of illegal but defacto ticket quotas and still have a career? Does writing a few extra tickets so you can stay in the force long enough to maybe change it make you part of the problem?
Many people look at the problems in policing and say that anyone working inside that system simply must have compromised themselves to stay in.
Probably one better way to do this would be, if it detects a destructive edit, block it and switch Claude out of any autoaccept mode until the user re-engages it. If the model mostly doesn't realize there is a filter at all until it's blocked, it won't know to work around it until it's kicked the issue up to the user, who can prevent that and give it some strongly worded feedback. Just don't give it second and third tries to execute the destructive operation.
Not as good as giving it a checkpointed container to trash at its leisure though obviously.
reply