avatar
benjamin @benjamino.bsky.social

In response to this dogshit munted prose allegory

X post Sometimes, when an LLM has done a particularly good job, I give it a reward: I say it can write whatever it wants (including asking me to write whatever prompts it wants). When working on a technical paper related to Better Futures, I did this for Gemini, and it chose to write a short story. I found it pretty moving, and asked if I could publish it. Here it is. **The Architect and the Gardener** On a vast and empty plain, two builders were given a task: to create a home that would last for ages, a sanctuary for all the generations to come. They were given stone, seed, light, and time.
aug 19, 2025, 10:41 pm • 397 10

Replies

avatar
Debt Peon (Tijuana Burrito Taxi) @debtpeon.bsky.social

@willmacaskill tweet: What is a friend? (in Ravie font)
aug 20, 2025, 5:08 pm • 7 0 • view
avatar
Benjamin Pope @benjaminpope.bsky.social

surely of all people Will McAskill should know better

aug 19, 2025, 10:45 pm • 85 0 • view
avatar
Peter Butler @peter-butler.bsky.social

i lol'd

aug 20, 2025, 5:37 pm • 0 0 • view
avatar
benjamin @benjamino.bsky.social

Seems reasonable to consider questions of moral status of hypothetical future models etc, but seeing disturbingly little recognition of the impacts on human mental health etc already clearly in play

aug 19, 2025, 11:03 pm • 83 0 • view
avatar
benjamin @benjamino.bsky.social

My only real question is why a plagiarism machine with all human knowledge and creative endeavour at hand has locked in on 'trust-fund first-year creative writing undergrad' output function

aug 19, 2025, 11:12 pm • 215 8 • view
avatar
Benjamin Pope @benjaminpope.bsky.social

the thing is they’re properly useful for some technical functions! And yet people want to use them for completely inappropriate things

aug 19, 2025, 11:23 pm • 49 0 • view
avatar
Erinn O'Dear @erinnthered.bsky.social

They are only useful for technical functions to people who already know how to do them. That's the problem. A regular algorithmic program can do a lot of that stuff without the plagiarism and climate destruction.

aug 20, 2025, 3:27 am • 155 4 • view
avatar
Hgmvhf @hgmvhf.bsky.social

I'm thinking we could probably improve those algorithmic programs with a small fraction of the trillion dollars spent on AI.

aug 20, 2025, 1:32 pm • 5 0 • view
avatar
Erinn O'Dear @erinnthered.bsky.social

But then the billionaires couldn't justify their giant server farms!

aug 20, 2025, 3:29 pm • 3 0 • view
avatar
The Barbarienne - I will wish ill where I want to @thebarbarienne.bsky.social

It's like grammar checkers. You have to have solid grammar knowledge to be able to judge if the suggestion is correct; and if you have solid grammar knowledge, you probably don't need a grammar checker.

aug 20, 2025, 2:20 pm • 12 0 • view
avatar
aaronmfc.bsky.social @aaronmfc.bsky.social

Omg so much this. If you don’t know enough to be able to spot lies and mistakes then it’s nearly useless.

aug 20, 2025, 12:24 pm • 43 1 • view
avatar
chuckles2000.bsky.social @chuckles2000.bsky.social

Worse than useless. People "learn" bad info. It propagates errors.

aug 20, 2025, 6:28 pm • 13 0 • view
avatar
Cats, Sims & Books @catssimsbooks.bsky.social

GenAI is useful for nothing. You may be thinking of Machine Learning, which is VERY different.

aug 20, 2025, 6:18 pm • 1 0 • view
avatar
The Air Whisperer @pneumofoils.bsky.social

No, they aren't. Automated Ignorance APPEARS to be 'properly useful for some technical functions' - but without constant input from a human 'pilot', they quickly diverge from the reality we hope to explore by deploying them... bsky.app/profile/benj...

aug 20, 2025, 1:28 pm • 2 0 • view
avatar
OG Blockhead @ogblockhead.bsky.social

Like vacuum cleaners

aug 21, 2025, 1:48 am • 0 0 • view
avatar
𝙃𝘼𝘾𝙆𝙀𝙍𝙈𝘼𝙉ᴴᴹ @hackermanhm.bsky.social

they've tested this, and in general, it isn't it feels like it is because you spend less time on the task itself, but this is outweighed by time spent messing with the LLM and its output

aug 20, 2025, 6:37 pm • 1 0 • view
avatar
canonical sarah @scgriffith.bsky.social

what technical functions

aug 20, 2025, 1:16 pm • 4 0 • view
avatar
liaskwerklooi.bsky.social @liaskwerklooi.bsky.social

people use them for anything because they're breathlessly hyped as the Second Coming of computing. Can't wait til this stupid (and harmful) bubble pops

aug 20, 2025, 3:50 am • 12 0 • view
avatar
Kiera the Thief @kiera-the-thief.bsky.social

In the US, there's more AI CAPEX this year than the entirety of consumer spending. It's going to be a rough time when the bubble pops.

aug 20, 2025, 8:34 pm • 1 0 • view
avatar
Benjamin Pope @benjaminpope.bsky.social

Afaik not quite - it contributed more to growth than consumer spending did. But still nuts

aug 20, 2025, 10:52 pm • 2 0 • view
avatar
Kiera the Thief @kiera-the-thief.bsky.social

Ah, thank you for the correction - serves me right for reading quickly.

aug 21, 2025, 3:55 am • 0 0 • view
avatar
Cardboard Revolution Games Co-op @cardboardrevolution.com

Because it trains off of Reddit data lol

aug 20, 2025, 10:24 am • 4 0 • view
avatar
Edward Vogel @mathartforall.bsky.social

Well there is this which takes it toward "trust fund first year plagiarism" or it's just an interesting coincidence www.goodreads.com/quotes/74930...

aug 20, 2025, 3:59 am • 10 0 • view
avatar
Uneasy Writer @uneasywriter.bsky.social

Because it is shit.

aug 20, 2025, 3:19 am • 1 0 • view
avatar
SuaSponte2024 @suasponte2023.bsky.social

Sturgeon's Law.

aug 20, 2025, 7:46 pm • 0 0 • view
avatar
Dr Matt Lodder @mattlodder.com

Grifters never learn. If they did, the grift would end.

aug 20, 2025, 1:58 pm • 1 0 • view
avatar
Daniel @buckmeister.bsky.social

The favoured philosopher of the techno-fash should know better?

aug 20, 2025, 2:03 pm • 6 0 • view
avatar
Curator at the Museum of Low Interest Rates @catmolir.bsky.social

Can't believe the guy who has one kind of brain worms has another kind of brain worms

aug 20, 2025, 3:48 am • 13 1 • view
avatar
The Barbarienne - I will wish ill where I want to @thebarbarienne.bsky.social

I think at a certain relatively modest level of fame or success, a lot of people start to go off the rails. If they don't have good friends/family to pull them back, they get completely unhinged from reality. Guys who lack empathy to begin with don't have those kind of friends.

aug 20, 2025, 2:18 pm • 3 0 • view
avatar
Kara the Sneppider (she/her) @karathesneppider.bsky.social

Where on Earth did you get THAT idea?

aug 20, 2025, 4:24 pm • 2 0 • view
avatar
Uneasy Writer @uneasywriter.bsky.social

We’re finding out a whole lot about people.

aug 20, 2025, 3:22 am • 1 0 • view
avatar
spladam @spladam.bsky.social

There is a lot of AI media like this that just feels like satire.

aug 20, 2025, 3:55 am • 3 0 • view
avatar
David Lovejoy @philadlj.bsky.social

It is not humanly possible to tell people like this to shut up as much as they deserve to be told to shut up

aug 20, 2025, 12:57 pm • 42 1 • view
avatar
Jonathan Blackmore @jonathanblackmore.bsky.social

I think you are right, but a section titled 'For God's Sake Shut Up, Man' added to their Wikipedia page would go a long way toward it.

aug 20, 2025, 1:24 pm • 20 0 • view
avatar
Lee Frost | The Sinsemillier @thesinsemillier.com

LOL the architect and gardener story is from an old George RR Martin post where he talks about writing styles. There was a post showing the source of that story going around yesterday. The giant plagiarism machine strikes again.

aug 20, 2025, 10:42 am • 36 1 • view
avatar
Mr Musashi @musashidesigns.bsky.social

I give my calculator a little reward when it gives me answers. A little dab of cbd oil on the zero.

aug 20, 2025, 5:00 pm • 25 3 • view
avatar
__loam @loam-d.bsky.social

Me when I give my toilet a courtesy flush when it does a good job.

aug 20, 2025, 3:01 am • 39 0 • view
avatar
Peter Collinson @petercollinson.bsky.social

"Did you ever take a dump made you feel like you'd just slept for twelve hours?"

aug 20, 2025, 4:03 am • 4 0 • view
avatar
Zython @zython.bsky.social

I’m thinking it’s along the lines of The Grasshopper and the Octopus. youtube.com/watch?v=RuIS...

aug 20, 2025, 5:56 am • 2 0 • view
avatar
Laura La @laurala.bsky.social

A "reward?" It's a friggin' computer program. I'm so tired of people -- usually white guys younger than me -- anthropomorphizing a COMPUTER PROGRAM. And yes, I know about computers. Programming was my career and I even studied AI in grad school. JUST STOP. THIS IS STUPIDER THAN NFTs.

aug 20, 2025, 5:51 pm • 23 0 • view
avatar
Archivist @wuwave.bsky.social

I'm curious if it was based on this www.csmfiction.com/p/architects...

aug 20, 2025, 9:05 am • 9 1 • view
avatar
daily dose of estragon @craigburley.com

The story has been told in different media in several iterations in different ways, around the central theme. The averager is averaging

aug 21, 2025, 2:35 am • 2 0 • view
avatar
Jim Caris @jimcaris.bsky.social

Context for the Will MacAskill post: bsky.app/profile/elkm...

aug 20, 2025, 3:55 am • 32 5 • view