alt Hacker NewsI really like Oxide's take on AI for prose: https://rfd.shared.oxide.computer/rfd/0576 and how it breaks the "social contract" where usually it takes more effort to write than to read, and so you have a sense that it's worth it to read.
So I get the frustration that "ai;dr" captures. On the other hand, I've also seen human writing incorrectly labeled AI. I wrote (using AI!) https://seeitwritten.com as a bit of an experiment on that front. It basically is a little keylogger that records your composition of the comment, so someone can replay it and see that it was written by a human (or a very sophisticated agent!). I've found it to be a little unsettling, though, having your rewrites and false starts available for all to see, so I'm not sure if I like it.
Replies
The problem with Ai writing is that its a waste of everyones time.
It’s literal content expansion, the opposite of gzip’ing a file.
It’s like a kid who has a 500 word essay due tomorrow who needs to pad their actual message up to spec.
I like the idea that various communications media have implicit social contracts that can be broken. In my opinion, power point presentations break an implicit social contract that is held in handwritten talks: if it's worth you displaying a piece of information, so that I the listener feel the need to take it in or even copy it down, it has to be worth your time to actually physically write it on the board. With power point talks this is not honored, and the average power point talk is much, much worse than the average chalk talk. I bet there are lots of other examples.
In 2020 at the start of covid, I did an experiment I called Project 35 where, for 35 days straight before my 35th birthday, I wrote 3 times per day, for 10 minutes each, I livestreamed it and whatever I wrote I would put directly into a book with no edits. While I didn't invite many people to join the calls (maybe fear, maybe just not wanting to coordinate it all), I found the process to be more raw, more human, and less perfect than 10x edited writing. It also helped me get better at typing in the moment and not rewriting everything, especially for social media, HN, and other places.
Anyway, it's at https://www.jimkleiber.com/p35/ if you wanna check it out, all sessions posted as blog posts, I think there's a link to the ebook (pay-what-you-want) and there may be audio (I recorded myself reading the writing right after each session).
If you check it out, please let me know :-)
I respect Oxide a lot. And here, too, with their non adoption of the marketing term (AI) for LLMs, ML.
Fun, I'd make playback speed something like 5x or whatever feels appropriate, I think nobody truly wants to watch those at 1x.
Years ago I wrote something similar to test a biometric security piece that used keystroke timings (dwell and stroke) to determine if the person typing the password is the same person who owns the account. Short version of a long story is that it would be trivial to get data for AI to reproduce human typing. Because I did it years ago using something only slightly more sophisticated than urandom.
You could totally make a believable timing generation model from a few (hundreds) recordings of human writing. Detecting AI is hard...
Based on the programs I was nudged to as a child, it was a surprise to no one but me that I scored higher verbal on the SATs than I did math, which I would have told you was my favorite subject. Despite the fact that French was my easiest subject. I can still picture the look on my french teacher’s face if I’d have mentioned this in front of him.
There are a lot of people like me in software. I’m tempted to say we are “shouted down”, but honestly it’s hard to be shouted down when you can talk circles around some people. But we are definitely in a minority. There are actually a lot of parallels between creative writing and software and a few things that are more than parallel. Like refactoring.
If you’re actually present when writing docs instead of monologuing in your head about how you hate doing “this shit”, then there’s a lot of rubber ducking that can be done while writing documentation. And while I can’t say that “let the AI do it” will wipe out 100% of this value, because the AI will document what you wrote instead of what you meant to write, I do think you will lose at least 80% of that value by skipping out on these steps.
I like the idea, but personally I would rather be thought a bot than show that I’m a human idiot who takes three tries to spell basic words.
This can only be fixed by authors paying humans to read instead of the other way around.
LLM-generated prose undermines a social contract of sorts: absent LLMs, it is presumed that of the reader and the writer, it is the writer that has undertaken the greater intellectual exertion. (Cantrill)
≈
The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it. (Brandolini)
To be fair, Oxide is a joke.
They want all this artisnal hand written prose under the candle light with the moon in the background. And you are a horrible person for using AI, blablabla.
But ask for feedback? And you get Inky, Blinky, Pinky, and Clyde. Aka ghosted. But boy, do they tell a good story. Just ain't fucking true.
Counter: companies deserve the same amount of time invested in their application as they spend on your response.
My biggest sorrow right now is the fact that my beloved emdash is a major signal for AI generated content. I've been using it for decades now but these days, I almost always pause for a second.