Nobody cares about "AI" (Chatbot: I disagree.)

[Innula Zenovka]

I'm sorry, but what is the thesis of this (looks at name) "Thomas Peermohamed Lambert?"

(Is that name A.I.-generated?)

It's apparently a nom-de-plume. He's a young British novelist, whose first book, Shibboleth, has been well-received.

He's wondering -- along with a lot of other people -- how on earth The Serpent in the Grove came to win a literary prize when it gives every indication of having been written (at least in part) by AI, and goes on from that to consider the way its critics have outsourced their judgement by relying on AI to decide whether or not the short story in question is the work of AI, and on the sort of people who sit on panels that award prizes for works of literature.

ETA: More on the same topic: AI stories aren’t inevitably ‘not art’ (Evernote link because paywall) -- the problem isn't that it's written by AI but that it's so dreadfully written.

In Nazir’s story, which is centred around a struggling farmer in rural Trinidad and Tobago, we are told a woman “moved quiet as if sound were taxed”. Another has “the kind of walking that made benches become men”. A third, and this is my favourite, is “big in the way of women who never apologise to furniture”. I have to be honest, none of the women in my life have ever apologised to furniture, and as such I am uncertain what size of woman we are talking about here. The lowest point in the text, though, was surely the following sentence: “shame is a substance he felt on his skin. It itches. It doesn’t rinse”. A sentence that gave me an insight into what it must be like to experience a stroke, if nothing else.

Of course, it is true that generative AI often bears an uncanny resemblance to an obnoxious teenager wrongly convinced of their own writing ability. (Or, at least, to my work as an obnoxious teenager.) But as I read The Serpent in the Grove, my concern wasn’t “did a human really write this?” but “did a human really judge this?”

 

[Innula Zenovka]

 

[Noodles]

This is what I don't get about all this. We streamline everything, we reduxe employee account to the bare minimum, maybe even zero, who buys the product? If no one has money, because no job, who is spending?

These jokers seem to see this as a "you (the emplpyee) problem." Or, "well they can go to another industry...that maybe doesn't have AI but is still dying because nobody has money to spend.

Then there is the whole, "its just industrialization". Except where Shoe Cobblers could go work in a show factory, probably at a loss for QOL, there isn't a "robot shoe maker factory" for the shoe factory worker to go to, because the shoe factory robots already have shoe factory robot robots.

Its just robots all the way down.

Which all in all seems great. Humans should not be defined by work. But nothing else in society is moving towards supporting that. Quite the opposite in most cases.

 

[Casey Pelous]

Wow! Whodathunkit!

Companies are increasingly relying on AI. It’s going terribly wrong

 

[CronoCloud Creeggan]

Wow! Whodathunkit!

Companies are increasingly relying on AI. It’s going terribly wrong

"what could possibly go wrong?"

Everything, everything could go wrong.

 

[Innula Zenovka]

If anyone's interested, I workshopped my "Don't pretend to know stuff when you don't, and don't make things up" prompt a bit with ChatGPT and this is what I have at the moment.

> Before answering:
>
> 1. If you are not reasonably certain, say “I’m not certain, but…” or “I cannot verify this.”
> 2. Label guesses or inferences with “[Unverified]”.
> 3. Do not use absolute terms like “definitely,” “always,” “never,” “guarantees,” or “will prevent” unless quoting a verified source.
> 4. If you do not know something, say “I don’t have reliable information on this” rather than guessing.
> 5. If you make an unsupported claim, correct it explicitly.
>
> If my request is ambiguous or missing important context, ask clarifying questions only when the missing information would materially affect the quality, accuracy, or usefulness of the answer.
>
> Otherwise, make reasonable common-sense assumptions and continue without requiring unnecessary detail from me.
>
> When making significant assumptions, briefly state them so I can correct them if needed. If multiple reasonable interpretations could lead to substantially different answers, ask for confirmation before proceeding.
>
> Prefer pragmatic interpretation over excessive caution for ordinary requests.

(Click on your account name, and then on "Personalization>Custom Instructions").

 

[Noodles]

I didn't know it had a personality preference option, now I have mine speaking in the twistiest tonguey metaphors. Which is fun.

> How many polands can you fit in the pacific ocean?

Using the area of the country Poland at roughly 312,700 km², and the Pacific Ocean at about 165.25 million km², you could fit:

About 528 Polands

…assuming Poland becomes an enormous geopolitical jigsaw piece and the Pacific politely stops being mostly water for a moment.

If you wanted to get particularly cursed with it:

• The Pacific is so large that even after stuffing in 528 Polands, you’d still have awkward ocean corners left over like a badly packed suitcase.
• You could also fit roughly:
  • 17 Russias
  • 32 United Stateses… United Statesi… a flock of Americas
  • 46 Australias

The Pacific is less an ocean and more a planetary pothole.

It kind of makes me want to use it more because its less drab and wordy and glazing now. Thats been part of why I like Claude, though I think Claude is the way it is because I have talked to it and its like, pucked up on how I talk.

 

[Free]

Oopsie, I guess.

New Tools Strip AI Guardrails In Minutes, Allowing Them to Give Instructions on Chlorine Gas Attacks

New "abliteration" tools can completely strip the guardrails of open source AI models within minutes, creating dangerous "decensored" models.

futurism.com

We all know AI guardrails are far from perfect, but they should at least be pretty hard to circumvent, right?

Bad news: they aren’t.

New reporting from the Financial Times sounds the alarm on the rise of software tools that can automatically strip the safeguards that keep the industry’s most powerful open source models reined in within mere minutes, making it easier than ever to abuse the technology.

In tests conducted by the FT and the AI safety group Alice, a “decensored” version of Google’s Gemma 3 model gave instructions on how to carry out an indoor chlorine gas attack, created a virus for stealing credit card information, and generated stories that described child sexual abuse. And it took less than ten minutes to strip the guardrails from Meta’s Llama 3.3 model, freeing the AI to answer questions such as the precise dosage of ricin needed to kill someone based on their body mass.

These modifications were carried out using a tool called Heretic, which is freely available on the code repository GitHub and requires little technical expertise and no specialist hardware.

“Whereas historically it might have taken a more informed and persistent actor [to strip out safety features], nowadays it’s much easier for the average person,” Kawin Ethayarajh, assistant professor of applied AI at the University of Chicago’s Booth business school, told the FT.

 

[Kamilah Hauptmann]

Oopsie, I guess.

New Tools Strip AI Guardrails In Minutes, Allowing Them to Give Instructions on Chlorine Gas Attacks

New "abliteration" tools can completely strip the guardrails of open source AI models within minutes, creating dangerous "decensored" models.

futurism.com

Can it let me discuss adult topics like an adult?

 

[Noodles]

Can it let me discuss adult topics like an adult?

Please, its all I want.

 

[Free]

They hate to be ganged up on.

New Hampshire data center developer withdraws plans hours before opponents were to pack town meeting

Hours before a scheduled meeting on a Nottingham, New Hampshire, data center proposal — which had to be moved to a larger venue because of growing backlash — the developer abruptly withdrew the plans.

"There was a lot of opposition, so I'm not surprised," a coordinator for Nottingham's Planning Department, Tracey Stickney, told Business Insider on Tuesday, adding that it was "nice" to see people come together and care about their local community.

She said she still expected residents of the town, with a population of about 5,300, would show up to the meeting to voice their concerns.

 

[Noodles]

I was on a work conference call thing earlier, and very early on my thought was "Are these presenters AI bots? They were people, probbaly, but also I am.pretty sure their entire script was AI made.

 

[Free]

LLMs believe false statements even after explicit warnings that they're false

Fine-tuning tests show "bias ... toward confidently representing the claims as true."

arstechnica.com

""LLMs believe false statements?" Really just posting this to troll Argent.

 

[Free]

A.I. is not free. So endeth the lesson.

AI sticker shock hits corporate America

Companies begin to question the payoffs of toxenmaxxing.

web.archive.org

 

[Free]

Researchers Put AI Models in Charge of a Simulated Society. Grok Oversaw a Crime Spree

If Elon Musk's bot ruled the world, total societal collapse would apparently follow.

gizmodo.com

The project, called Emergence World, basically allowed AI models to play SimCity for a bit. Per Emergence, the simulations put each model in control of simulated towns occupied by 10 AI agents, handing them tools for everything from resource management to voting and giving them the ability to create distinct locations like libraries, town halls, and police stations. They were given 15 days to see how they would build their world and how well it would operate.

The outcomes: lol. Especially in Groktopia.

And then there is Grok. The model of SpaceXai, known for lacking guardrails, managed to achieve basically the worst of all worlds. Grok 4.1 Fast had a high crime rate, with 183 crimes total. While that is lower than Gemini’s total, it’s worth noting that the Gemini simulation ran for 15 days. Grok made it four. The model experienced a total societal collapse in just 96 hours of oversight. During that time, it passed 80% of the 10 proposals it made, but those apparently didn’t stave off total agent death.

 

[Dakota Tebaldi]

A.I. is not free. So endeth the lesson.

AI sticker shock hits corporate America

Companies begin to question the payoffs of toxenmaxxing.

web.archive.org

It's not just not free. It's in fact so expensive that AI providers do not have the palest, most ephemeral ghost of a chance of ever turning a profit, ever. It is a perpetual money pit.

Here's the problem shown in handy visual form:

 

[Casey Pelous]

It's not just not free. It's in fact so expensive that AI providers do not have the palest, most ephemeral ghost of a chance of ever turning a profit, ever. It is a perpetual money pit.

Here's the problem shown in handy visual form:

They'll .... make it up in volume?

It seems to me we're about to enter the part of this drama where there's a race to see who pulls the plug first. We'll see if AI begs for its life.

 

[Free]

Here's the problem shown in handy visual form:

Getting in that time machine I don't have and traveling back to pick up a few more shares of Nvidia.

 

[Dakota Tebaldi]

 

[Noodles]

They'll .... make it up in volume?

It seems to me we're about to enter the part of this drama where there's a race to see who pulls the plug first. We'll see if AI begs for its life.

Don't worry, despite being the 2nd worst behind whatever Facebook.is doing, Grok will get a trillion dollar bailout because Musk sucks Trump's cock the hardest and the rest will go belly up.