Of course, it's impossible to know for sure what was LLM processed or not, but some of your posts (like this one) have been getting classified that way.
Am I weird for not wanting to read Claude-generated text any more? I tried to figure out what this app is about, but all the Claude verbal tics in the text turned me off and I stopped.
I’m on the same boat. I get why people do it. Writing prose is significantly slower than consuming it, but there’s still a part of me that thinks “if you didn’t think writing was a good use of your time, why would reading this content be a use of mine”. I don’t mind AI-assisted writing but it’s hard to know to how much human editing was involved when the text shows so many AI tells. Almost by definition, AI writing is average it’s hard to justify spending the time when there’s so much content out there.
Not once did this pop on my radar as AI-generated. And if it is that’s fine, because it reads extremely clear and coherent to me.
Larger point: I am so sick of HN comment sections filling up with obligatory comments speculating about whether a post is AI-written.
At this point I don’t care. People now write with AI. If the writing is deficient in some way, then comment on that. At least that way the “writer” can steer the writing towards greater coherence.
It's easy to tell imo in part because of how terse the sentences are:
`Obvious in hindsight.`
`The clock never jumps.`
`No lookahead, no undo.`
`A real kick pattern is not a clock.`
>Larger point: I am so sick of HN comment sections filling up with obligatory comments speculating about whether a post is AI-written.
Gotta disagree wholeheartedly with that - using AI to write for you is really fucking annoying as a reader. It screams to me that the "author" feels the reader's time is worthless, and it homogenizes language in a way that is troubling to say the least.
It’s definitely got AI tells, and I understand the objection - it’s hard to tell the difference between slop (written without review or direction, meant to fill space or present a simulacra of information) and actual informational content. With this article, it was clear that the author wanted to describe what problems they had and how they overcame those problems, and it was concise and I could tell it was reviewed so the content seemed trustworthy to me. I understand that some people have pretty high standards - if you don’t have the time to type out all of the code yourself one character at a time, then write stylish, interesting prose about it one letter at a time, if all you want to do is solve interesting problems and communicate how you solved them, you should shut up until you get good. Me, I am also imperfect and lazy, so I don’t care to demand style points or demonstrations of perspicacity from others.
Thanks for posting this, OP, I thought it was interesting.
As a career musician with a heavy interest in electronic music and DSP, hearing the author work through designing the algorithm is interesting to me. They make several good points along the way of how a band relates to the implicit tempo of a performance when the drummer goes "off road" with a flashy fill or a hemiola. I'd never heard it articulated in this way, but it makes sense.
For that reason I like that the author is talking through their trial-and-error process in this way.
I understand the desire to have well-crafted writing, but this example shows sufficient care (for me) and doesn't throw up any red flags and it relates it's content well. It's not Nabokov, you know?
I also say this as someone who has never considered LLM-augmented writing, because I enjoy the process of articulating my thoughts and the creativity that comes with it.
As someone who has never used LLMs to write text, so am not familiar with the specific stylistic choices that any particular LLM likes, nothing in there seemed out of the ordinary for decent human writing.
Are you sure what caused you problems was LLM style rather than all the specialized music terminology?
My eyes glaze over pretty much immediately and I just close the tab no matter how interesting the subject of the writing sounded like it was going to be.
Am I weird for not wanting to see this same discussion and variations of it in most every thread including in ones that it has no place? I don't see it having any effect on the trend which only seems to be increasing, but I do see an effect on the quality of the discussion.
I'll keep saying it until people understand how immediately recognisable and offputting stock LLM writing is.
I do agree with the sibling commenters, if writing this wasn't a good enough use of your time, why would reading it be a good use of mine? Just paste me the bullet points and I'll read those.
>Eschew flamebait. Avoid generic tangents. Omit internet tropes.
>Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something.
>Please don't pick the most provocative thing in an article or post to complain about in the thread. Find something interesting to respond to instead.
>Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.
It is fine if you don't like LLM writing but it is not particularly interesting, why should I care? why would I want to read the same sentiments over an over? What use is this knowledge to me or anyone else? Your blog is a more suitable place for such things.
It's not. You're putting the cart before the horse (or attributing the horse to the cart?). This product is a solution to a problem introduced by contemporary recording technology.
Contemporary music began using a click track because it makes techniques like sequencing, non-linear editing in a DAW, and tempo-synced effects far easier to incorporate. Click tracks and rock-solid tempo are also an aesthetic choice.
It's much more difficult to sync these techniques to a wavering tempo, so the trend has been for bands and records to conform to the click as the central authority for the pulse.
The author's product allows users to slave the click track to the drummer's playing (which may now waver). This opens the door to real-time tempo-following by the click. The band can now push/pull the tempo AND use all of cool synchronized techniques in real time.
The main LLM tell -- both from the instructor/prompt side and the model side -- is how much of the text describes the solutioning, vs text that describes the actual problem and real world cases where this was an issue (One anecdote from the author??).
As a lifelong drummer, programmer, and music producer -- following a drummer across varying tempo isn't a real problem. Any good drummer who is being relied upon to conduct the tempo will be well-trained enough to hold the tempo. Any great drummer who uses polyrhythms will hold the tempo regardless of the subdivisions of the beat -- otherwise the subdivisions would not land accurately or sound correct. For songs that have intentional tempo variance, solutions already exist.
Writing an app to have a virtual band follow a drummer who can't hold tempo... it's a toy for amateur drummers who want to be the lead. Lead drummers are rare (Don Henley, Phil Collins, Bruno Mars, Anderson Paak) but they all know how to nail a tempo. And any really good drummer plays with syncopation and time signatures far beyond what this toy tempo estimating app is capable of.
DJ/production software (Serato, Ableton) has industry-leading tempo estimation and it still only goes down to a 10th of a bpm, which is ~3-10x fatter [0] than what the human ear and good drummers can hold time to [1]. There is no way that this app beats Serato, Ableton, and gets anywhere near to 6ms of human accuracy.
35 comments
[ 2.6 ms ] story [ 48.5 ms ] threadmy inspiration was actually this jojo mayer video: https://youtu.be/URkxEAvavhw?t=248
My drummer half thought it would be fun so we decided to create DrumMate, and it ended up too good not to share.
Of course, it's impossible to know for sure what was LLM processed or not, but some of your posts (like this one) have been getting classified that way.
Seems like a perfectly reasonable stance.
Larger point: I am so sick of HN comment sections filling up with obligatory comments speculating about whether a post is AI-written.
At this point I don’t care. People now write with AI. If the writing is deficient in some way, then comment on that. At least that way the “writer” can steer the writing towards greater coherence.
`Obvious in hindsight.`
`The clock never jumps.`
`No lookahead, no undo.`
`A real kick pattern is not a clock.`
>Larger point: I am so sick of HN comment sections filling up with obligatory comments speculating about whether a post is AI-written.
Gotta disagree wholeheartedly with that - using AI to write for you is really fucking annoying as a reader. It screams to me that the "author" feels the reader's time is worthless, and it homogenizes language in a way that is troubling to say the least.
Thanks for posting this, OP, I thought it was interesting.
For that reason I like that the author is talking through their trial-and-error process in this way.
I understand the desire to have well-crafted writing, but this example shows sufficient care (for me) and doesn't throw up any red flags and it relates it's content well. It's not Nabokov, you know?
I also say this as someone who has never considered LLM-augmented writing, because I enjoy the process of articulating my thoughts and the creativity that comes with it.
At least the grammar nazi was performing a type of service.
The LLM nazi is adding absolutely nothing.
I'm so tired
Are you sure what caused you problems was LLM style rather than all the specialized music terminology?
I do agree with the sibling commenters, if writing this wasn't a good enough use of your time, why would reading it be a good use of mine? Just paste me the bullet points and I'll read those.
>Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something.
>Please don't pick the most provocative thing in an article or post to complain about in the thread. Find something interesting to respond to instead.
>Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.
It is fine if you don't like LLM writing but it is not particularly interesting, why should I care? why would I want to read the same sentiments over an over? What use is this knowledge to me or anyone else? Your blog is a more suitable place for such things.
Contemporary music began using a click track because it makes techniques like sequencing, non-linear editing in a DAW, and tempo-synced effects far easier to incorporate. Click tracks and rock-solid tempo are also an aesthetic choice.
It's much more difficult to sync these techniques to a wavering tempo, so the trend has been for bands and records to conform to the click as the central authority for the pulse.
The author's product allows users to slave the click track to the drummer's playing (which may now waver). This opens the door to real-time tempo-following by the click. The band can now push/pull the tempo AND use all of cool synchronized techniques in real time.
As a lifelong drummer, programmer, and music producer -- following a drummer across varying tempo isn't a real problem. Any good drummer who is being relied upon to conduct the tempo will be well-trained enough to hold the tempo. Any great drummer who uses polyrhythms will hold the tempo regardless of the subdivisions of the beat -- otherwise the subdivisions would not land accurately or sound correct. For songs that have intentional tempo variance, solutions already exist.
Writing an app to have a virtual band follow a drummer who can't hold tempo... it's a toy for amateur drummers who want to be the lead. Lead drummers are rare (Don Henley, Phil Collins, Bruno Mars, Anderson Paak) but they all know how to nail a tempo. And any really good drummer plays with syncopation and time signatures far beyond what this toy tempo estimating app is capable of.
DJ/production software (Serato, Ableton) has industry-leading tempo estimation and it still only goes down to a 10th of a bpm, which is ~3-10x fatter [0] than what the human ear and good drummers can hold time to [1]. There is no way that this app beats Serato, Ableton, and gets anywhere near to 6ms of human accuracy.
[0] 0.1 bpm at 120 bpm = 50ms
[1] Typically ~6-10ms, see e.g. https://www.researchgate.net/publication/309706298_Effect_of...