Who says AI doesn't create new kinds of businesses: I would hire someone to re-transcribe all the LLM output I have to trudge through back into human language. I don't have trouble with the volume, I can just feel my brain trying to eject my temporal lobes out my ears as each passing word makes becoming alingual more appealing.
10
Replies
2
Boosts
@jonny we could have had a world where first LLMs are like linux or wikipedia, open weights and trained with dataset where people know what's going into it.
Instead the whole damn thing's running on disgruntled office speaks and bottom of the barrel SNS posts.
Before LLMs I don't think I read a sentence to paragraph-level writing style that was viscerally annoying to me -
there were annoying people that said annoying things, there were people who didn't spell that well or had challenging grammar which I don't find annoying at all. the fallback writing style for people who couldn't write well was more or less just a transcription of their conversational voice which is fine and actually nice sometimes. ESL writing is good not bad to me because I love to read the ways different languages put thoughts together, a little semitranslation puzzle. The only thing I can think of as annoying is on larger scales like how a book or article is constructed, e.g. Douglas Hofstadter, Nathanial Hawthorne, Stephan Wolfram all drive me up a wall with how they get high on their own supply and repeat everything 20 times.
atypical writing voices tended into strange, interesting, and human, and it was easy to empathize and read past or become fascinated by how someone could come to write that way. I miss that so dearly, where now everyone who is self conscious about their writing voice sounds the same, and I hate the way it sounds.
Oh wait I thought of an exception - LinkedIn writing style. Maybe that's just it, I hate LLM writing style because its very similar to LinkedIn writing style
@jonny just go fruit ninja on that… input. 🥷 hiii-ya!
Anthropic published that p-space, "global workspace", steerable latent space thing - you can tell it doesn't really work because that would be something you would immediately want to productize, being able to do surgery on the internal state of a model would have immediate and obvious marketability in many of the cases where LLMs are most poorly suited. But if it did work, I would also give in and pay money to retranscribe all the LLM output with the LinkedIn voice vector nulled out.
@jonny I was once subjected to a presentation about how to use LLMs to rewrite your LinkedIn posts.
And it struck me during the presentation that to me, the previous text sounded *much more authentic*, sincere, and interesting than the LLM's output.
"Why would people want to sound exactly like every other post on LinkedIn?"
I guess we're in the minority. 😢
@NfNitLoop @jonny I think, unfortunately, many corporate cultures operate on the old saying...
"the nail that sticks out will be struck by the hammer"
I don't think the structure of organisations which want workers to be completely interchangeable parts, as a form of wage suppression, see individuality as anything but a liability.
@jonny it's not just a new kind of business, it's an adventure. The interest is real. Half the reading time. Only twice the price.
