I actually thought it might be AI because it was a highly specific task that would be a pain in the ass (unnatural, tedious, challenging) for a human to do but that you could summarize in a fairly short and sweet prompt.
I included an example of such a prompt on my take (which I know you have already read, but hey I never left a comment about it and the window was still open when I came back to safari)
I thought the attempt was pretty interesting on Claude’s behalf here. A little apologetic gesture at potential sentience? Hard to guess how that could have leaked in except maybe that people who write run on sentences are more likely to be woo?
I will say however that Claude's attempt is pretty obviously AI. The style is uncanny. Mine includes too many things AI would never write even with a great prompt. There's too much idiosyncrasy. The "no periods" challenge is, in this sense, a distraction: of course AI can do it!
I call it the AI Accent (some humans adopt it from too much communication with AI, very much contagious). I feel like I can spot ChatGPT, especially 4o, instantly. I have more trouble spotting Claude’s ticks. What did you notice? (If you don’t mind me asking.)
I found it to be a bit too much of a sucker punch and had no clear reason to doubt that AI wrote it. From my perspective, and before reading the comments especially Joel McKinnon's reproduction attempt, any reason I could come up with would probably be disproven by clever prompting.
well, I stopped at the fourth line or so. LLMs may have a teleological goal, but I am not really sure that they could not write such a text with a proper prompt, at least up to the latest lines.
Interesting way to make a point Alberto. A side point that may not have been intended, is that this "ultimate idiosyncrasy" demonstrates the downside of the being weird that you recently advocated. Punctuation and paragraph breaks make for easier cognitive flow, and doing without them is not at all conducive to readability, or even the attempt being made to read very far (I had to go back and plow through the whole thing to see if you had actually already made this point somewhere within the wall of text).
I still like the idea of being weird, but it has its limits if the intention is to engage with fellow humans. This is why most people don't bother to visit modern art museums and why very few poets followed ee cummings' idiosyncratic example of eschewing capital letters. I think there is a fruitful balance that can be struck, and do agree that LLMs are — by design — guilty of hewing too hard to the conventional side of that balance.
Hey Joel! Thanks for reading haha -- it was surely not a pleasurable read (that was the intention)! There's actually a second point made: I wrote it, not an AI, but who knows anymore, right? Having epistemic hygiene at a time when you can't believe your eyes means being critical at all times. I'm doing my part in keeping you guys sharp! (I don't think AI can write like this, but it probably can write without periods if instructed correctly)
I tried it! I took a recent post I wrote on my own substack (seldoncrisis.net) and asked ChatGPT to write it in exactly the same style. It started out fine, but eventually had to put in a period and added a paragraph break for good measure. A little later it repeated the period + paragraph break and I noticed that each paragraph got shorter until it was essentially writing pretty normally. You're right that ChatGPT does not like being weird.
I'm glad someone is doing the metafictional experiments that should be occasioned by the advent of the text-machine. Glad it was an objet d'art and not writing for its own sake, cuz that was dull as dirt
I don't know whether to curse you or commend you. I had one moment of question when there was no capital letter after a question mark but dismissed it as either a mistake or a stylistic choice. Should I consider it great progress that this is possible? Or a great loss for humanity that we will now be bombarded with writers who don't engage with the art of writing? Can it be both?
I found the structure to be engaging, and read to the end? if it can be called that, although I was pretrained by familiarity with Joyce's last chapter of Ulysses, as well as Gertrude Stein's texts, so that I could relax into the stream of language as not only medium but semantic point.
This made me think about how constraints are where meaning actually comes from, not something we optimize away. The ending works here because it doesn’t resolve.
I loved the experience of reading it, thank you. But as the comment show, it takes patience and the willingness to actively direct our attention - to engage with the creation of others, the "wall of text". Maybe soon, AI will help us figure out which human text might be worth the read, instead of creating walls of text itself.
And I wholeheartedly agree that AI will, over time, push us humans towards our own "weirdness and idiosyncrasy". I have been on a journey myself to "find my voice" within the flood of eloquence and have learned a lot in the process - about AI and about myself.
Well, that ended up being depressing.
Maybe the ending was too harsh! (I wrote it but I think leaving a disclaimer would kill the effect)
I read about half, skipped to the end and assumed it was AI, don’t see why AI couldn’t write a 900 word run on sentence 😵💫
Why did you assume it was AI? Any telltale sign or just because
Me having no patience to read the whole thing and then i believed you 😅 I also did, when I was reading that AI could not do this, think AI could.
I actually thought it might be AI because it was a highly specific task that would be a pain in the ass (unnatural, tedious, challenging) for a human to do but that you could summarize in a fairly short and sweet prompt.
I included an example of such a prompt on my take (which I know you have already read, but hey I never left a comment about it and the window was still open when I came back to safari)
https://blog.aptitude.guru/p/claude-craves-completion-prior-to
I thought the attempt was pretty interesting on Claude’s behalf here. A little apologetic gesture at potential sentience? Hard to guess how that could have leaked in except maybe that people who write run on sentences are more likely to be woo?
I will say however that Claude's attempt is pretty obviously AI. The style is uncanny. Mine includes too many things AI would never write even with a great prompt. There's too much idiosyncrasy. The "no periods" challenge is, in this sense, a distraction: of course AI can do it!
I call it the AI Accent (some humans adopt it from too much communication with AI, very much contagious). I feel like I can spot ChatGPT, especially 4o, instantly. I have more trouble spotting Claude’s ticks. What did you notice? (If you don’t mind me asking.)
I found it to be a bit too much of a sucker punch and had no clear reason to doubt that AI wrote it. From my perspective, and before reading the comments especially Joel McKinnon's reproduction attempt, any reason I could come up with would probably be disproven by clever prompting.
Saw the wall of text, and felt the immediate dread and couldn’t read any of it. Was this the point? Haha.
Certainly not, but I assume this will happen! (I'm testing the limits of what people will be willing to read out of curiosity)
Please don't ever do it again!
well, I stopped at the fourth line or so. LLMs may have a teleological goal, but I am not really sure that they could not write such a text with a proper prompt, at least up to the latest lines.
I read this as:
I am weird, therefore I am
thank you for the thoughtful piece, Alberto
This is the best, most succinct read so far. Thank you Ethan!
Interesting way to make a point Alberto. A side point that may not have been intended, is that this "ultimate idiosyncrasy" demonstrates the downside of the being weird that you recently advocated. Punctuation and paragraph breaks make for easier cognitive flow, and doing without them is not at all conducive to readability, or even the attempt being made to read very far (I had to go back and plow through the whole thing to see if you had actually already made this point somewhere within the wall of text).
I still like the idea of being weird, but it has its limits if the intention is to engage with fellow humans. This is why most people don't bother to visit modern art museums and why very few poets followed ee cummings' idiosyncratic example of eschewing capital letters. I think there is a fruitful balance that can be struck, and do agree that LLMs are — by design — guilty of hewing too hard to the conventional side of that balance.
Hey Joel! Thanks for reading haha -- it was surely not a pleasurable read (that was the intention)! There's actually a second point made: I wrote it, not an AI, but who knows anymore, right? Having epistemic hygiene at a time when you can't believe your eyes means being critical at all times. I'm doing my part in keeping you guys sharp! (I don't think AI can write like this, but it probably can write without periods if instructed correctly)
I tried it! I took a recent post I wrote on my own substack (seldoncrisis.net) and asked ChatGPT to write it in exactly the same style. It started out fine, but eventually had to put in a period and added a paragraph break for good measure. A little later it repeated the period + paragraph break and I noticed that each paragraph got shorter until it was essentially writing pretty normally. You're right that ChatGPT does not like being weird.
Let me take a look!
Oh, my God. This hurt too much.
I'm glad someone is doing the metafictional experiments that should be occasioned by the advent of the text-machine. Glad it was an objet d'art and not writing for its own sake, cuz that was dull as dirt
I don't know whether to curse you or commend you. I had one moment of question when there was no capital letter after a question mark but dismissed it as either a mistake or a stylistic choice. Should I consider it great progress that this is possible? Or a great loss for humanity that we will now be bombarded with writers who don't engage with the art of writing? Can it be both?
But maybe the fact that there was an uncapitalized word after a question mark should make you suspect that this was not written with AI after all!
Great... point!
So was it really you or was it AI instructed not to use a full stop? Interesting premise!
Try it! Let me know if you manage to prompt ChatGPT to write like this!
I found the structure to be engaging, and read to the end? if it can be called that, although I was pretrained by familiarity with Joyce's last chapter of Ulysses, as well as Gertrude Stein's texts, so that I could relax into the stream of language as not only medium but semantic point.
This made me think about how constraints are where meaning actually comes from, not something we optimize away. The ending works here because it doesn’t resolve.
I loved the experience of reading it, thank you. But as the comment show, it takes patience and the willingness to actively direct our attention - to engage with the creation of others, the "wall of text". Maybe soon, AI will help us figure out which human text might be worth the read, instead of creating walls of text itself.
And I wholeheartedly agree that AI will, over time, push us humans towards our own "weirdness and idiosyncrasy". I have been on a journey myself to "find my voice" within the flood of eloquence and have learned a lot in the process - about AI and about myself.
A masterpiece. Thanks!
I jumped right into the text, since it was the first piece of content I was confronted with today. And it was a delightful gem. Thanks for this.
Thanks for reading Dave!
Your biggest semicolon fan is still here. Bravo!
Is this written by AI or not? Am I dumb?
What do think Keith?
IDK Phil the last sentence confuses me dude. WDYT?
btw what’s your area of expertise? Just curious. I’m sure we can work together on something.
Yo Phil I figured it out it was him dude