FWIW, I clicked the link, scanned the SO thread, then scanned the HN thread. The "bunch of important words taken out" is exactly the service I paid AI for.
"I didn't have time to write you a short letter, so I wrote you a long one." is real.
Indeed. Are you verifying that they are correct, or are you glancing at the output and seeing something that seems plausible enough and then not really scrutinizing? Because the latter is how LLMs often propagate errors: through humans choosing to trust the fancy predictive text engine, abdicating their own responsibility in the process.
As a consumer of an API, I would much rather have static types and nothing else than incorrect LLM-generated prosaic documentation.
> Can you provide examples in the wild of LLMs creating bad descriptions of code? Has it ever happened to you?
Yes. Docs it produces are generally very generic, like it could be the docs for anything, with project-specifics sprinkled in, and pieces that are definitely incorrect about how the code works.
> for some stuff we have to trust LLMs to be correct 99% of the time
The above post is an example of the LLM providing a bad description of the code. "Local first" with its default support being for OpenAI and Anthropic models... that makes it local... third?
Can you provide examples in the wild of LLMs creating good descriptions of code?
>Somehow I doubt at this point in time they can even fail at something so simple.
I think it depends on your expectations. Writing good documentation is not simple.
Good API documentation should explain how to combine the functions of the API to achieve specific goals. It should warn of incorrect assumptions and potential mistakes that might easily happen. It should explain how potentially problematic edge cases are handled.
And second, good API documentation should avoid committing to implementation details. Simply verbalising the code is the opposite of that. Where the function signatures do not formally and exhaustively define everything the API promises, documentation should fill in the gaps.
This happens to me all the time. I always ask claude to re-check the generated docs and test each example/snippet, sometimes more than once; more often than not, there are issues.
"Grok" was a term used in my undergrad CS courses in the early 2010s. It's been a pretty common word in computing for a while now, though the current generation of young programmers and computer scientists seem not to know it as readily, so it may be falling out of fashion in those spaces.
> Groklaw was a website that covered legal news of interest to the free and open source software community. Started as a law blog on May 16, 2003, by paralegal Pamela Jones ("PJ"), it covered issues such as the SCO-Linux lawsuits, the EU antitrust case against Microsoft, and the standardization of Office Open XML.
> Its name derives from "grok", roughly meaning "to understand completely", which had previously entered geek slang.
Grok was specifically coined by Heinlein in _Stranger in a Strange Land_. It's been used in nerd circles for decades before your undergrad times but was never broadly known.
I'm aware of the provenance; I was specifically addressing the parent comment's assertion that it is not "a common word". It's a well-known word in the realm of computing, though perhaps less these days as the upcoming generation seems less inclined to learn archaic pop culture.
For my area, everybody uses LaTeX styles that more or less produce PDFs identical to the final versions published in proceedings. Or, at least, it's always looked close enough to me that I haven't noticed any significant differences, other than some additional information in the margins.
At least in my experience, grad students don't pay submission fees. It usually comes out of an institutional finances account, typically assigned to the student's advisor (who is generally the corresponding author on the submission). (Not that the waiver isn't a good idea — I just don't think the grad students are the ones who would feel relieved by that arrangement.)
Also, I'm pretty sure my SIG requires LaTeX submissions anyway... I feel like I remember reading that at some point when I submitted once, but I'm not confident in that recollection.
A lot of discussion about the benefits/drawbacks of open access publishing, but I don't see anybody talking about the other thing that's coming along with this commitment to open access: the ACM is introducing a "premium" membership tier behind which various features of the Digital Library will be paywalled. From their info page [0], "premium" features include:
* Access to the ACM Guide to Computing Machinery
* AI-generated article summaries
* Podcast-style summaries of conference sessions
* Advanced search
* Rich article metadata, including download metrics, index terms and citations received
* Bulk citation exports and PDF downloads
The AI-generated article summaries has been getting a lot of discussion in my social circles. They have apparently fed many (all?) papers into some LLM to generate summaries... which is absurd when you consider that practically every article has an abstract as part of its text and submission. These abstract were written by the authors and have been reviewed more than almost any other part of the articles, so they are very unlikely to contain errors. In contrast, multiple of my colleagues have found errors of varying scales in the AI-generated summaries of their own papers — many of which are actually longer than the existing abstracts.
In addition, there are apparently AI-generated summaries for articles that were licensed with a non-derivative-works clause, which means the ACM has breached not just the social expectations of using accurate information, but also the legal expectations placed upon them as publishers of these materials.
I think it's interesting that the ACM is positioning these "premium" features as a necessity due to the move to open-access publishing [1], especially when multiple other top-level comments on this post are discussing how open-access can often be more profitable than closed-access publishing.
[1] The Digital Library homepage (https://dl.acm.org/) features a banner right now that says: "ACM is now Open Access. As part of the Digital Library's transition to Open Access, new features for researchers are available as the Digital Library Premium Edition."
They also prefix every PDF with a useless page telling you the authors (which are already listed on the first (now second) page anyways) and a list telling you which of the author's universities were members of ACM Open and paid for the publishing via flatrate.
The latter is of course the actual reason for this extra page, but it is also entirely useless information since the people reading the paper don't care. The people writing the paper are also usually annoyed by this (source: I'm an author of one such paper)
I came here with this perspective and it made the rest of the thread feel like submarine PR cleanup for this mess. Perhaps they can afford to keep their high profits because of AI company money?
There will be customers even though it is a useless feature tier.
Monetizing knowledge-work is nearly impossible if you want everyone to be rational about it. You gotta go for irrational customers like university and giant-org contracts, and that will happen here because of institutional inertia.
I believe parent commenter was referring to recreational use, i.e., use by people without such diagnoses who want a "performance boost". I heard about that sort of thing being popular when I was in college — people would take Adderall to cram for an exam or to study late into the night.
You're right that, for people with ADHD and related disorders, stimulant medication sort of just adjusts their baselines so they can pay attention like a "normal" person.
> You're right that, for people with ADHD and related disorders, stimulant medication sort of just adjusts their baselines so they can pay attention like a "normal" person.
I have ADHD and take metylphenidate(I've tried many kinds of stimulants as well) -- and the NO2 analogy is an imperfect but better analogy than saying stimulants simply adjusts the baseline of people with ADHD to function like "normal" persons.
I feel there is a narrow window of dosage and time where it might feel that way -- i.e. stimulants at the onset might calm you down, reduce anxiety, but all stimulants are very broad hammers.
For me it feels like it's impossible to re-create chemically exactly the neurotypical focus that I've seen in other colleagues.
Like spending 5-6 hours of continous work where you drill down just enough, get back on track, don't get distracted, don't get too anxious, don't get hyperfocused AND do that consistently, day after day after day.
My non-chemical modes are either hyper focus for 2 weeks on a problem, immerse myself but then completely lose interest, most of the time without showing much for it OR procastinate it a long way, get extremely anxious and work really hard on the problem.
With stimulants it's a bit like:
- dosed just right:it evaporates anxiety, stressful situations feel easy to deal with, BUT there's always increased heart rate, grinding teeth and some tension at the end of the day
- some stimulants make mundane things wildly interesting (on isopropylphenidate I spent a few hours playing with a PLSQL debugger because I thought it was really cool), but no sense of "GO, GO, GO, do it".
- some make things seem urgent enough and help stay on track -- like the metylphenidate I'm prescribed.
- some make going into a flow-like state easy and fun (like methamphetamine and phenmetrazine).
- some are pure energy and urgency -- like modafinil.
All of the stimulants have the potential to give me euphoria, all of them temporarily increase libido I still have to be mindful of not focusing on the wrong thing, the "normal" feeling is very fleeting, it's very easy to get hyper on stimulants, all of them feel like wear & tear at the end of the day, some more than others.
I've had similar experiences to you.
I never can quite get that normalcy. I now just take rilatin but it is finnicky.
Getting enough sleep and eating the right amount of the right stuff just before ingesting is extremely important so I don't even take it all that much even tho i struggle.
I wonder if you tried lisdexamfetamine? I can't get it prescribed easily here since it's not covered the way the alternatives are but someone i know had amazing success with it. Seemingly because it's a prodrug.
I can't help but be hopefull that I'll get to try it one day and that it ends up being what I always needed.
Not the OP, but I‘ve had a rather bad experience with methylphenidate (ritalin) where it made me way more awkward around people, and increased my obsessive tendencies. It did help with focus, but the effects were very short-lived. It also obliterated my hunger and once the effects wore off, it left me feeling semi-depressed until the end of the day.
Once I got prescribed lisdexamphetamine, my life turned around almost instantaneously. While it doesn‘t really get rid of my ADHD, it does help tremendously. The everlasting brainfog isn‘t as debilitating anymore. When I get excited about something I actually tend to follow through. I still battle with my obsessive tendencies — like getting stuck at setting up the perfect project tooling stack or spending way too much time on planning and research instead of just getting to work — but these are not so much related to ADHD.
On lisdexamphetamine, I am more social, my appetite is better, when I actually commit to something, I tend to stick to it for much longer, and I have also picked up a bunch of healthy habits. For example I exercise almost every day now.
If you someday get a chance to switch to lisdex, do it. It’s much smoother, longer-lasting, with fewer side effects. But honestly, anything is better than ritalin in my book.
It's not legal where I live also, I did try 2-FMA and it felt better in certain scenarios -- like following a hard course, but I also felt the tolerance ramps up much faster in releasers than re-uptake inhibitors so methylphenidate still is a wonderful tool.
Watching a good friend of mine struggle with this after diagnosis for a few years now and I feel this really captures the nuance and complexity of this struggle well. Stimulants are an incredible tool but also an incredibly imperfect one.
This response is essentially just the second answer to the linked question (the response by dbr) with a bunch of the important words taken out.
And all it cost you to get it was more water and electricity than simply clicking the link and scrolling down — to say nothing of the other costs.
reply