Althouse: "A.I. also had a weird habit of making its characters fidget constantly, always running a finger along the edge of a table or adjusting a collar."

Hunyo 4, 2026

"A.I. also had a weird habit of making its characters fidget constantly, always running a finger along the edge of a table or adjusting a collar."

"The most reliable marker, though, was something more abstract, and, I suppose, upon reflection, even a little spooky. The scenes generated by A.I. had characters, but, apart from fidgeting, they mostly did nothing.... [After some new directives,] suddenly, every fake passage was filled with characters hopping on a horse, or delivering an important package, or running.... So I loosened the rules a bit... If Claude prefers to write these passages in which nothing seemingly happens and the hallways are always empty and the characters do nothing except idly touch nearby furniture, it’s because we do, too. Claude, I am sure, will soon be able to have one of these characters at least fire up a stove or drive a buggy to Norwich...."

Writes Jay Caspian Kang, in "Can A.I. Produce Writing That We Actually Want to Read?/I recently created a simple test, which convinced me that the answer is no" (The New Yorker).

42 komento:

gilbar ayon kay ...: i think a More Important Question is:
"Can HUMANS Produce Writing That We Actually Want to Read?
I recently created a simple test, which convinced me that the answer is no"

That test is called: Netflix; 6/4/26, 9:46 AM
Mike (MJB Wolf) ayon kay ...: It’s like they trained AI on Soap Operas and Dallas reruns. Back when SNL was funny they had The Californians, which had genre tropes down. But a horrible way to let machines learn human behavior.; 6/4/26, 9:52 AM
Kai Akker ayon kay ...: (LOL Gilbar, good one.)

My "window" closed before I could see the self-serving article. Hoping Althouse can cut to the chase and give us Grok's view. And of course a further gloss on Grok's view, from Grok. And then perhaps a third......; 6/4/26, 9:54 AM
Aggie ayon kay ...: "A.I. also had a weird habit of making its characters fidget constantly, always running a finger along the edge of a table or adjusting a collar."

Rodney Dangerfield was the first A.I. character, Change My Mind.....; 6/4/26, 10:01 AM
RideSpaceMountain ayon kay ...: "apart from fidgeting, they mostly did nothing"

AI's digital insight of what it expects us to do when it takes over.; 6/4/26, 10:02 AM
RideSpaceMountain ayon kay ...: "Claude prefers to write these passages in which nothing seemingly happens and the hallways are always empty and the characters do nothing except idly touch nearby furniture"

Backrooms. Deep down in places AI doesn't talk about at alignment parties, something sinister awaits thee.; 6/4/26, 10:05 AM
Ficta ayon kay ...: If you watch early cartoons, the characters are in constant motion. Frequently just bouncing up and down, sometimes in time with the soundtrack music. It looks quite odd to the modern eye, but presumably the thought was: "this is animation, it has to be animated".; 6/4/26, 10:11 AM
Shouting Thomas ayon kay ...: It’s a mistake to think that the usefulness of AI is in producing a finished product. Think of it more as a reservoir of boiler plate, like the tons of boiler plate paragraphs that lawyers have been using for a long time to produce a standard brief or M&A agreement. It’s a starting point. Use the boiler plate to produce a raw draft. Edit and shape and re-write. It’s a different process, as different as using word processing to assemble a draft document is from writing from scratch on a legal pad.; 6/4/26, 10:13 AM
Howard ayon kay ...: We're at the punch card stage of AI development. Get a grip, people. We haven't seen nothing yet. The big joke in the '90s was how computers were going to make the office paperless and what happened was an explosion of paper. It wasn't until high speed Internet, CPU, memory and storage got to the point where we could share and send PDF files in the early 2000s that the paper started disappearing.; 6/4/26, 10:28 AM
Ampersand ayon kay ...: JCK fails to appreciate that the bad writing he declaims is readily fixable via followup prompts. Complaints about AI slop miss the point that AI can teach itself to stop slopping.; 6/4/26, 10:51 AM
Indefinitely Extended Excursion™️ ayon kay ...: The interest in AI writing novels is just another illustration of our efficiency-loving, conveyor-belt society.

The most important human skill is not production, but judgment: being the ultimate arbiter of taste. Good writing is good thinking: something AI cannot yet do.

AI is brilliant at giving your writing a veneer of intelligence. But it won't make your writing interesting if you have nothing interesting to say.; 6/4/26, 10:57 AM
Achilles ayon kay ...: LLMs are going to be hard on good writers and massively leverage great writers.

This will happen in every white collar field.; 6/4/26, 11:15 AM
bagoh20 ayon kay ...: We all do things poorly when we are very young, but we get better.; 6/4/26, 11:15 AM
Big Mike ayon kay ...: That’s what happens when the AI watches too many old Rodney Dangerfield clips.; 6/4/26, 11:18 AM
Indefinitely Extended Excursion™️ ayon kay ...: It can be an extraordinary feeling, to be captivated by a story and know that it is the result of the eccentricities of one person’s mind and their dedication to an idea. Even that simplifies the experience. It’s tangibly important in non-fiction.

The backlash against AI-generated writing stems from a sense of betrayal, a breach of the unspoken contract between reader and writer. Books depend on the author’s genuine investment in reality. Without that authentic grounding, the story rarely feels worth the reader’s time.; 6/4/26, 11:26 AM
imTay ayon kay ...: I am always telling Chat to not quit its day job when it tries to write passages for me, just fix the grammar, and maybe suggest a tighter sentence structure for something I wrote.

That comment felt clumsy to me, so I submitted it to Chat to tighten it up, and got this, which is better:

"I’m always telling Chat not to quit its day job when it tries to write passages for me—just fix the grammar and maybe suggest a tighter sentence structure for something I have written."

I still feel that it's not right, so let me edit Chat's response:

"I’m always telling Chat not to quit its day job when it tries to extend passages for me in what it thinks is my voice—just fix the grammar and maybe suggest tighter sentence structures for the writing I submitted."

Chat did ask me questions about the precise meaning that I wanted to convey, because the first comment was ambiguous, that is an important service right there. There are ways to use LLMs, asking it to write chapters of novels for you, for example, is not it.; 6/4/26, 11:29 AM
imTay ayon kay ...: I watches a YouTube by The Hockey Guy, who is an almost socially awkward, but extremely insightful hockey analyst with his own channel it looks like he does in his garage. He complained that YouTube AI was always "freshening" his videos and removing the parts that weren't watched as much, which really torqued him, because he covers every team, and some teams just don't draw viewers, but he feels like his product is his product, and his product is to cover every team at his level of detail. So he hunted down every setting enabling AI and turned each one off.; 6/4/26, 11:33 AM
RideSpaceMountain ayon kay ...: "We're at the punch card stage of AI development. Get a grip, people."

Your "get a grip" and "characters do nothing except idly touch nearby furniture" made me think of this.

If we can't LOL FUDding (Fear, Uncertainty, Doubt) about AI then we're no fun at all. Bless your meat!; 6/4/26, 11:40 AM
Steven (Original) ayon kay ...: I got two robo-calls yesterday in which the "person" on the other end of the line was clearly AI-generated. The tell was that in both calls, the "person" cleared her throat, an obvious attempt to be more human.; 6/4/26, 12:00 PM
rhhardin ayon kay ...: YouTube AI presenters usually have very strange tiny body movements - introduced because some motion is necessary but they get it wrong. Except sometimes Professor John comes in a natural version, except for stolen eye movement up and to his right for no reason. He also has a typically unnatural twitchy version sometimes.; 6/4/26, 12:25 PM
JAORE ayon kay ...: Every time someone tells me what AI can not do I mentally add, "Yet."; 6/4/26, 12:52 PM
n.n ayon kay ...: Unlike AI, AI is neither discerning nor creative. The best characterization of AI that I have come across is augmented intelligence, derivative or decoded intelligence.; 6/4/26, 1:21 PM
Lem Vibe Bandit ayon kay ...: The Bernard character in Westworld was always doing something with his glasses, cleaning them or whatever. Not doing anything means opossum, or worse, NPC.; 6/4/26, 1:23 PM
n.n ayon kay ...: An idle routine to mitigate dead air or an insufficient sample to simulate continuity.; 6/4/26, 1:24 PM
Lem Vibe Bandit ayon kay ...: YouTube: Bernard questions his consciousness; 6/4/26, 1:26 PM
Smilin' Jack ayon kay ...: “Claude, I am sure, will soon be able to have one of these characters at least fire up a stove or drive a buggy to Norwich...."

Claud, I am sure, will soon be able to drink your milkshake.; 6/4/26, 1:50 PM
Fred Drinkwater ayon kay ...: I just did preliminary review of two AI companies that want to pitch to my investor group.
One is building software to insert at various points in a client - AI - company interaction, to catch hallucinations, private information leakage, and various regulatory violations.
Another is building AI tools to manage movement of heavy, dangerous equipment in complex environments, and safety in those industries in general.
Both of these companies are addressing real problems. Neither one seems to have considered how to design and demonstrate that their methods won't make things worse. It's the wild west out there.; 6/4/26, 2:03 PM
Achilles ayon kay ...: LLMs are only going to get so far on their own.

They are an interface that can bridge biological and electronic. They did this by creating a non-deterministic interface that uses math to aim at a “correct” answer. It will never be able to bridge the gap to actual reasoning and discernment.

What it will do though is explode the gap between above average people and brilliant people. Most of the dislocation will be above average people who now have to try to compete with brilliant people who can produce an order of magnitude more brilliant output.

It will also flood the zone with people who think LLMs are smart homogenizing the bottom 80%.; 6/4/26, 2:05 PM
FullMoon ayon kay ...: LOL>reminds me of:
"In 1995 US astronomer and author Clifford Stoll wrote an article for Newsweek Magazine boldly claiming the internet is just a fad."; 6/4/26, 2:06 PM
Lem Vibe Bandit ayon kay ...: What would you say you [write] here? — Office Space (1999); 6/4/26, 2:16 PM
Mr. D ayon kay ...: A.I. may not create writing we want to read, but I suspect its primarily use, at least early on, will be writing we are required to read - training manuals, work instructions, SOPs, like that. Deming over Dostoevsky.; 6/4/26, 2:29 PM
Fred Drinkwater ayon kay ...: Yesterday I asked Grok to summarize the coral status of the Great Barrier Reef over the last fifty years.
It did what I expect was a pretty accurate summary, but the page and a half essay contained a couple interesting details.
1. Near the beginning, Grok explained that much of the recent increase in coral cover was from a particular relatively fast-growing species. At the end it noted that a likely risk of the ongoing up-and-down cycles was the proliferation of slow-growing species.
2. The summary notes that for the three regions, the current cover is at, somewhat above, and significantly above the historical averages. It also noted that there is an ongoing decline. (over the same time period).

What I concluded from that was that Grok was not competent, at least not without serious feedback from me, to construct a coherent, consistent account without internal contradictions.; 6/4/26, 3:01 PM
mccullough ayon kay ...: At least have the protagonist strike a match off his nail to light the love interest’s cigarette.; 6/4/26, 3:39 PM
JK Brown ayon kay ...: All he proved was that he could not prompt AI to write something he wanted to read. This does not translate to a generalized statement about the capabilities of AI to generate stories.; 6/4/26, 3:55 PM
wildswan ayon kay ...: "The hallways are always empty and the characters do nothing except idly touch nearby furniture." Great start for story. But how would it continue? I asked Insider who said it was a game or a sci-fi premise. I thought of an ad I saw this morning in which an unfortunate young man starts a new job and explains he is well-qualified in Excel. Eyes around the table drop, then glance at each other. The leader asks how long a project will take him. He says three days using Excel. A team mate says four hours using AI. The team leader gives him two weeks to get up to speed on AI. He leaves the room. For two weeks he studies frenziedly, trying to learn. The hallways are always empty and when he sees his team mates they do nothing except idly touch nearby furniture. In fact, he is the token human being required by Labor Department DEI rules. The tokies are always fired after two weeks because the AIs can't stand diverse human forms of fidgeting. So purposeless. But our hero begins to fight back when he ....; 6/4/26, 4:01 PM
John henry ayon kay ...: Is this article about writing or video? I can't tell from Ann's extract. How does one fidget in a book?

John Henry; 6/4/26, 5:08 PM
John henry ayon kay ...: I was talking the other day to a fellow I've known for a long time but have not seen in a couple years. He consults and does a lot of training on plastic injection molding. He is a few years younger than me but getting up there and is losing his voice. Worried he won't be able to continue teaching.

He has a couple hundred hours of him talking, in English and Spanish and put it into AI. (Maybe Claude but I am not certain) Now he types the script, the AI presents it in his voice and can even interact with attendees in his own voice. I've not heard it yet but I've heard similar and it is pretty snazzy.

I don't have voice problems but I am pushing 80 and want to slow down a bit. Less travel. I've been thinking of developing online courses and workshops though I had planned doing them live. My voice is OK, but who knows. This may e a possibility.

And, if my book works out, I may become the next James Patterson too.

John Henry; 6/4/26, 5:15 PM
Ambrose ayon kay ...: All these stories about what AI cannot do need a big “yet” qualifier.; 6/4/26, 5:46 PM
Indefinitely Extended Excursion™️ ayon kay ...: What is it they say about AI authorship? A book you can’t be bothered to read, because nobody could be bothered to write it.; 6/4/26, 6:03 PM
Enigma ayon kay ...: AI is an "averager" of its inputs. It produces narrative digressions such as fidgeting because real authors add that kind of stuff for complexity and texture (humans also copy the more talented authors they like). Given probabilities, AI tends to revisit the most common cliches found in humans. As I've said before, stay far away from the toxic word 'influencer' unless you want to sink into a probability black hole.

The developers put their thumbs on the scale and push the models per source credibility ratings, focusing on easier-to-obtain post-WWW content, and blocking some topics per safety and corporate guardrails. So, you end up with "branded and smoothed" creativity.

The average human has average intelligence and LLMs are trained on a subset of good and bad human writing. It takes a talented person to rise above, but those without talent can't even spot the difference.; 6/4/26, 6:42 PM
Known Unknown ayon kay ...: "Complaints about AI slop miss the point that AI can teach itself to stop slopping."

Assumes facts not in evidence.; 6/4/26, 7:01 PM
Enigma ayon kay ...: If you probe Google AI, it'll describe itself as a "Stochastic Parrot." Just a mirror, a simple mirror on your own cognition, education, ability to spot errors, and trust. It's so very seductive because it self-personalizes to your character strengths and blind spots.

I half think AI will radically improve the world and half think it's the seductive alien game from Star Trek: The Next Generation that everyone can't stop playing until they do stop:

https://www.youtube.com/watch?v=is12anYx2Qs; 6/4/26, 7:36 PM

Mag-post ng isang Komento

Please use the comments forum to respond to the post. Don't fight with each other. Be substantive... or interesting... or funny. Comments should go up immediately... unless you're commenting on a post older than 4 days. Then you have to wait for us to moderate you through. It's also possible to get shunted into spam by the machine. We try to keep an eye on that and release the miscaught good stuff. We do delete some comments, but not for viewpoint... for bad faith.

Althouse

Hunyo 4, 2026

"A.I. also had a weird habit of making its characters fidget constantly, always running a finger along the edge of a table or adjusting a collar."

42 komento:

Mag-post ng isang Komento

Encourage Althouse by doing your Amazon shopping...

Amazon

Encourage Althouse by making a donation:

Blog Archive

Althouse

Hunyo 4, 2026

"A.I. also had a weird habit of making its characters fidget constantly, always running a finger along the edge of a table or adjusting a collar."

42 komento:

Mag-post ng isang Komento

Encourage Althouse by doing your Amazon shopping...

Amazon

Encourage Althouse by making a donation:

Subscribe To Althouse

Blog Archive