The world according to ChatGPT
post by Richard_Kennaway · 2025-04-07T13:44:43.781Z · LW · GW · 0 commentsContents
1. 2. 3. 4. 5. None No comments
1.
I linked this in a small comment [LW(p) · GW(p)] yesterday, but I think it deserves greater prominence.
It's an article by Sam Kriss, "Born in the wrong generation". It starts from an image that someone tweeted of an idyllic scene of paradise: 1950's small-town America. The tweeter's comment was, "Can we bring this back?"
The image is from an AI generator. The most obvious tell is that none of the lettering on the street signs and storefronts makes sense, but there are others. (Of the foreground woman in the centre, can you tell which leg is which? Why is the other woman wearing such huge gloves? Whose feet are those, seen below her dress to the right of her own legs?)
The article goes on to imagine living in the world of ChatGPT. It is viscerally horrific. His "beautiful submissive girlfriend" agrees emptily with everything he says. The background people outside the "folksy chrome-and-cream diner on Main Street in your quiet walkable 98% white small American town" are nightmares of deformed flesh. "Well hey there, your hazy waitress will say to you, opening a mouth that suddenly lacks a lower jaw. Keep safe now, says the friendly but unfocused cop. The gun dangling on his belt is half-made of dead grey skin." His meal "tastes of wet."
This is the world of ChatGPT.
2.
In that earlier comment I mentioned having received an email from Academia.edu that began
"Based on the papers you’ve autocfp, we think you might be interested in..."
No human hand wrote that sentence. What papers have you autocfp recenty? Can I help you wiht mrre refernecs? There are 32,03ll4 recently paper on the scebjtt of Lagnuage.
3.
There are several Facebook groups that I follow that post works of art of various sorts. The pictures are real (I can find them on Google Images), and the commentaries on them make sense, but I have suspected for a while that some of these groups are automatically generated. The proof came recently in a group dedicated to Escher. The picture of the day was his "Still Life and Street". The commentary, however, was clearly the result of the AI taking the title to be the titles of two pictures, "Still Life" and "Street", and the result was completely fabricated. I unfollowed the group.
4.
Two years ago I posted a demo of getting ChatGPT to argue any side of a question [LW · GW]. I have just repeated the experiment with ChatGPT 4.5. [LW · GW] It was even more easily led than whatever version I used back then.
5.
It was recently speculated that Trump's formula for tariffs was suggested to him or his advisors by an LLM. I don't know if it was, but a day or two after the tariffs came in, when stock markets had plunged, I heard on the radio one of the Trump team talk about how this was part of the plan, a necessary pain as the price of reorganising the trade system according to Trump's vision. My immediate thought was, did that also come from an LLM? Because I did not hear of any such plan in advance of the shock. Perhaps I imagined it, but the spokesman's voice did not seem to come from any place of conviction, as if it was just, this is the thing he's been told to say now.
But I'm not in America and don't have a finger on the pulse. Can anyone who is more informed answer this question: was anyone saying in advance of the market reaction that this was the plan, a rocky road towards the promised paradise? All I recall was on the lines of "we'll have tariffs, and America will be great again."
0 comments
Comments sorted by top scores.