automatic meeting summaries in particular aren't using the best models, I see speech to text mistakes constantly. it's like using the best model to process hundreds of hours of video conference per week would cost too much, but on the other hand doing it with a cheap model is worse than not doing it at all - especially if you relieved the scribe thinking ai had you covered
> People who rarely produced original writing are suddenly producing
these posts are beginning to make me wince. ai is giving voice to a lot of folks we probably wouldn't even be hearing from otherwise - because not everyone does their thinking in articulate prose, and extracting their realizations as shareable language takes effort, or long, embarassing iterations with ai "workshopping" to get to something they can read back and go yes this is what I am trying to say.
now the sentiment is that whatever these people had to say they could have just come out and said it - and that would be more passionate and less disjointed - no, not necessarily, and the more it gets repeated the more it's sounding elite and pompous to me. for example this:
> I had someone write me a personal message about a sensitive topic that was clearly workshopped with AI in an effort to nuance the conversation and not offend me. But the message became impersonal, dispassionate, and disjointed. It had all of the parts, but it didn’t make sense as a whole. I wasn’t interested in reading it, or responding.
this reads as someone choosing to ignore a personal message about a sensitive topic - because of an assumption they made about their writing: that they workshopped it when they should not have - because whatever it was they were struggling to say they should have isntead trusted the author to be able to understand the precise meaning of their raw, inelegant thoughts no matter how embarrassed they might be to share them in that form. they're complaining about a message they probably weren't otherwise going to see.
But how is that a positive that people that cannot give a fuck about writing coherent thoughts down to share them can now hog your attention with something that has no clever meaning whatsoever? Is talking just for the sake of talking a good thing in your book?
> people that cannot give a fuck about writing coherent thoughts down
That's clearly not the type of people they are talking about. For some people expressing their thoughts in words is really, really hard. It doesn't mean they don't give a damn. Just like being in a wheelchair doesn't necessarily make you lazy.
At the risk of sounding extremely insensitive, I am not interested in hearing the thoughts of someone if for them expressing them into words is really, really hard. I would be interested in reading the 30% or so that comes from them, but not at the cost of having to skim through the rest of the slop.
Do you realize there is no comment on this entire website of any real importance, including yours? Someone saying their reasons for agreeing with the article isn't something worthy of shut-downs and insults.
Everyone is an elitist, when you've enplaned do you assume the pilot is qualified to fly it or do you hope everyone gets a turn?
Admittance to the canon is no less a responsibility, would you trust the yoke of human culture to the artless, craftless, and naive?
It should be obvious to you, as a great epicurian of thinking, that the ability to spit out prose is orthogonal to the ability to think thoughts worth prosing about.
I strongly disagree with this statement. The careful selection of words that results in elegant prose is one and the same with the thinking. Sure, there are many ideas worth prosing about, but those are just a premise until someone goes through the effort to intentionally shape words around the premise.
>Orthogonal means intersecting or lying at right angles (perpendicular) in geometry, or mathematically independent and non-interacting in advanced fields
Because while I dislike AI writing, it matters in context. If someone is trying to formulate a coherent thought of work tasks required of me then I don't really care if it reads like AI as long as the point they're trying to get to me is clear.
If there's an attempt at something human, or emotionally important I will quickly balk at anything spat out through an LLM. If you're trying to communicate at the human level with me I want your input, flaws warts and all.
A card with just...nothing personal on it has as much value as machine text. I understand that you took the effort to think of me, but that's the bare minimum. If trying to express yourself isn't worth the time to make it personal then I'd almost appreciate if you didn't try at all because it hurts more to believe I'm not worth the effort.
How is ai giving them voice? All that generated slop does that you stop to want to read anything. And it is not their voice anyway. It is not their words, their ideas are mixed with ai generated points.
As for second paragraph, the writing "had no overall meaning". There is nothing to respond to if there is no overall meaning.
What we are seeing here is a Luddite reaction from artisans. A skill has been automated and those that have the skill are not happy about it.
For many writing is painful and the machines ease the pain. And like all machines it takes a while to use them skilfully. In particular remembering that like code it is write once and read many. What the reader needs should be at the forefront of a writers mind - both for code and prose.
Nobody gets on a motorbike and wins the TT in the first weekend.
Every person who has ever significantly won has done it by getting others to run for them. Tell me the billionaire that earned their fortune through the work of their own hands.
Writing can be painful because putting pen to paper forces you to think through every word and sentence. You immediately realize that the first thought that came to mind, wasn't a 100% correct idea, and it'll be more accurate and better formulated this way or that way. Writing is a sort of extended thinking, in which you're being held accountable for every word (by yourself). Some people argue that slowing down the writing process even more, for example by going from keyboard to pen and paper, gives you more time to process your thoughts, and leads to better and more articulate writing.
When I read AI-enhanced writing, I always feel like I have to peer through the blinds to see the writer's true intention. This feels very dishonest and partly offensive, because it can sometimes take several paragraphs before I realize that I'm reading AI output, rather than someone's processed thoughts.
I really, really despise this sentiment and the people that make it.
Writing has always been incredibly difficult for me, both due to incredible difficulties with handwriting and finding my own voice. That difficulty I find in writing is the friction that identifies what my voice is, the words I choose and the meaning I attempt to convey through my own understanding.
I cannot read AI-generated crap for the life of me because all of it simply has no meaning. I will go back and read through paragraphs, trying to grasp at the meaning of something that does not exist. If your argument is that AI-generated writing eases the pain then I will make the opposite argument: reading AI-generated slop increases the pain 10x. I would sooner read through an essay with endless punctuation issues and spelling mistakes than one passed through an LLM because one is an earnest attempt to communicate while the other is offloading all of the mental burden to me. Which gets to the root issue of a lot of the complaints around people slinging slop: you make it marginally easier for you, and then far worse for everyone around you.
"If your argument is that AI-generated writing eases the pain then I will make the opposite argument: reading AI-generated slop increases the pain 10x"
Then you didn't bother to read, or understand, my argument.
In particular this bit: "remembering that like code it is write once and read many. What the reader needs should be at the forefront of a writers mind - both for code and prose."
Are you sure you haven't got to the stage where you don't read anything properly at all?
Not the person you're addressing, but writing by hand remains a very valuable tool. It's eminently practical and far from obsolete. [1]
[1] "There's actually some very important things going on during the embodied experience of writing by hand," says Ramesh Balasubramaniam, a neuroscientist at the University of California, Merced. "It has important cognitive benefits."
You may think this is silly, but you actually think more when you write by hand than when you type on a computer, and especially more than when you just listen. When attending a lecture you should write notes on paper even if you just throw them away.
Nobody cares what your English grade is. Everyone would prefer to just hear you speak in your own voice. This isn't Luddism, this is the equivalent of people pushing their selfie through five Instagram filters until they look like an Oblivion character because they have no confidence in their own appearance.
This is not done in the service of readers, nobody wants the slop, it's writing dysphoria
the article we're commenting on quite literally cites a survey on reader behavior:
"Cynthia Dunlop recently shared survey results on reading AI-written articles entitled Report: How developers react to AI-scented blog posts. For the majority of readers, if they think an article is AI-assisted or AI-authored, they will stop reading (78%) and avoid the author in the future (71%)[...] the strongest result in the survey was people overwhelmingly preferred the author’s own writing (98%), with all its flaws and idiosyncrasies,[...]*
there's a say-do gap in play - similar to makeup/no-makeup studies: people claim to prefer natural/no-makeup while behavioral research reveals a preference for subtle/light/professionally-applied makeup
when it comes to ai people love to claim to hate it outright - this thread is full of descriptions of ai writing being impossible to read, totally incoherent, and superlatively bad/worse than average human writing. meanwhile evidence suggests that we might actually kind of like ai writing - with a growing portion of samples making hn's frontpage (https://www.salahadawi.com/hacker-news-ai-detector/state-of-...)
I've seen my share of ai slop and I get where the rage is coming from - I think folks could do a lot better than shouting down every use of ai as slop, especially while hypocritically using ai in their own writing. the pompous elitism on display in this thread alone is gross even for hn
the fact that the HN (or other) frontpages feature a non-trivial amount of AI content is to a large extent owed to the fact that many commenters don't even read the content they're commenting on, another problem in and of itself.
There is no elitism in demanding that people use their own words, or if you want to call it elitism then elitism is healthy and good. Imagine ten years ago you talk to someone online and they told you they outsourced their posting to a call center guy in the third world, that's what this is. Literacy and agency aren't gross, it's what we ought to expect of an educated adult. It's fundamentally a sign of honesty that if I talk to you or read what you write I want to hear words out of your mouth, not some meat puppet version of you hooked up to a chatbot.
demanding that people not use a technology that makes expression accessible to them, and insisting that they can only produce the sort of slop that you're sick of seeing, is not just elitist - it's indignantly pompous. it's neither healthy nor good - it's hateful and hurtful
> ..they outsourced their posting to a call center guy in the third world
nobody actually does this though - this exaggerated caricature dwells only in the hateful mind. to be fair there are content farms as there always have been and ai slop is bellowing from their stacks, but the article we're commenting on and your sentiment are not aimed at them - they're targeting the people we're not used to hearing from who might have interesting things to share now that technology is helping them organize their thoughts. you want to say "if you needed ai to organize your thoughts, your thoughts aren't worth organizing" there's nothing good or healthy about that sentiment - you think you're "demanding literacy of educated adults" but actually you're demanding a particular kind of linguistic fluency you happen to value - that's not a defense of literacy it's hallmark elitist gatekeeping
> What we are seeing here is a Luddite reaction from artisans.
They are hardly artisans. They are simply the vocal few who repeatedly conflate communication with art simply because the written word can be used for both.
"Easing the pain" is the foundation of many of today's problems. It's gotten too easy to avoid discomfort and effort. Also, if we are talking about disabilities, maybe we should start respecting different forms of communication instead of enabling people to "write" in a way that is not authentic to them at all, just so they can seem "normal".
Actually, yes. Even though you tried to make it sound ridiculous. I want to move towards a world where the worth of a person (and their thoughts, feelings, etc.) is not measured by an exclusive standard of "intelligence", especially if that standard is created/enforced by mega-corporations.
there's a growing contingent of anti-ai extremists shouting down anything that looks or feels like the thing they hate: actual slop. but there's a missing understanding of where slop really comes from - "you used ai instead of thinking" is just another guess, and mostly wrong since you literally cannot use ai without thinking - the prompts don't write themselves.
slop has been around a lot longer than ai - the contingent needs to realize what it is they hate. instead of "don't use ai to write" it's "don't use ai to write slop" or better: "don't write slop"
To be clear, I (the author of the article) do not hate AI. A lot of my current work is focussed on making AI work well.
However, I don't think it's a good idea to replace our own thinking with an AI, and that's what I'm trying to avoid with the way of working proposed in the article.
No. I hate AI and slop. I hated slop before AI, but with AI slop has proliferated to an unprecedented degree. If AI went away tomorrow, slop would go down with it to 1/1000ths of what it is today.
in my experience it’s pretty common to find big inverted indexes for text directly in the database - not necessarily large docs but certainly free text records in volume. using bm25 and unicode’s breakiterator is a very good way to build it. like putting lucene in the database basically - makes a lot of sense when the database is already large. places that bend over backwards to move search out of the db are usually trying to avoid having a very large db (and often end up with one anyway, getting the worst of both worlds)
They also end up with all the infrastructure and processes necessary to keep the external search system in sync, resync/reindex, pkey shipping back to their source of truth in queries, application-side joins and enrichment between both sources. It’s brutal.
Having everything in one place eliminates entire classes of development and especially operational problems.
when I was in school we were taught that the human brain was really two brains - the left and right hemispheres. it seemed to explain how you could learn a skateboard trick in your normal stance, then have to learn it again switch.
reply