AIs gunning for our precious freelancers

[rotting bones]

As for using LLMs in writing, IME they're mostly useful for converting shorter descriptions into longer ones.

If you want it to write whole novels, there are ways to get it to do that, but I wouldn't recommend it. Describe the novel in as much detail as you can, including stylistic details, and tell it to write a chapter outline. Then tell it to write one chapter at a time, mentioning the full title from the outline, like "Write Chapter 1: ..."

Expect utter shit.

[Starbeam]

Or just not deplete a reservoir and not use it.

[rotting bones]

Or just not deplete a reservoir and not use it.

1. LLMs don't necessarily dry reservoirs any more than 3D gaming. I run CodeQwen on a machine with a 16 GB graphics card. I ran LLaMA for the narrative project on a laptop with an 8 GB graphics card. I also don't do any 3D gaming, so I probably end up using less power than the leftist influencer bros you listen to.
2. Has anyone compared the environmental degradation caused by running an LLM for a few minutes vs an author obsessing over a piece of writing for months while consuming resources on a first world scale?
3. Blanket opposition to science doesn't come from a principled stance. These are people who are outraged that their livelihood is imperiled. Instead of being luddites, they should oppose the root cause of the problem, capitalism. As long as capitalism is in place, there is nothing we can do to prevent reservoirs from being depleted.

[rotting bones]

I wonder if the quality of ChatGPT's novels can be improved by enabling web search and using this kind of prompt:

Research the relevant topics online, then write a chapter outline for an x novel that does y in style z. The specifics of the research must be incorporated at the plot level, not just as aesthetics.

[linguistcat]

I wonder if the quality of ChatGPT's novels can be improved by enabling web search and using this kind of prompt:

Research the relevant topics online, then write a chapter outline for an x novel that does y in style z. The specifics of the research must be incorporated at the plot level, not just as aesthetics.

I haven't tried the search function on Deepseek, but you can include a web address for a page you want to include as part of the prompt and it'll go over any text on the website. And you can include several urls. So that's sort of halfway there.

[rotting bones]

I haven't tried the search function on Deepseek, but you can include a web address for a page you want to include as part of the prompt and it'll go over any text on the website. And you can include several urls. So that's sort of halfway there.

Good idea. I told ChatGPT to write a novel that would have been written by Vulcans using contemporary research: https://docs.google.com/document/d/16P0 ... sp=sharing

I think the story is about a bunch of scientists trying to bound Vulcan's growth within ecological parameters. It's hard to follow:

10.4 Collapse or Confluence?—The 30‑Minute Trial

The Annex implants a nano‑scale topoconductor ring into the temporal cortex of a volunteer adept, then initiates a meditation protocol called *Gol‑ein‑kita*, reputed to still neural chatter to near‑delta rhythms. For 1,832 seconds the hybrid network sustains qubit fidelity above 0.999, rivalling dedicated dilution‑refrigerator rigs ([news.ucsb.edu][8]). Then, as the adept’s heart rate rises, ionic turbulence spikes; fidelity dives; quantum wave‑function logs reveal an abrupt Penrose‑style objective collapse at 1,847 seconds, matching the gravito‑temporal threshold in Hameroff’s latest conference keynote ([consciousness.arizona.edu][10]). But instead of returning coherent state, the system devolves into thermal noise—proof that biological dampening still falls short.

...

12.8 Resolution — The Compression Protocol

The Council ratifies a single clause:

> *All future growth—biological, computational, or cultural—shall occur adiabatically, trading one conserved measure (energy, area, or rights) for another so that the total ledger entropy never rises.*

They christen the theorem Surak’s Law of Conservation of Diversity, a final synthesis of IDIC and thermodynamics.

Those are some of the most readable paragraphs in the output. I need to play with the style parameters. On the other hand, maybe Vulcan fiction wouldn't be human-readable. But then, the output doesn't even try to disguise contemporary research as occurring within the Star Trek universe. I'm not sure if that's my prompt's fault or the LLM's limitation. I did phrase it as Vulcans being a "thematic" component of the novel. That might have been a mistake.

Also, maybe Vulcans weren't meant to be Victorians. Maybe they were a projection of what nerdy kids like me aspired to become when we grew up.

[jcb]

Chat programs are a huge advance, sure. But the tempo is important to get right, since malloc is convinced human-level AI is coming next week.

Not really. It does seem quite plausible that they will develop in the next decade, though, given the incredible pace of development over the past few years. It might turn out that the last few cognitive tasks separating humans from AI are fantastically difficult of course, but the opposite seems equally likely.

(1) You are vastly underestimating the intelligence of a human.
(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?
(3) AIs have no human experiences. They have only consumed an enormous amount of documents written by humans describing human experiences.

[Raphael]

(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?

This.

[rotting bones]

(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?

This.

Honest question: I know the basics of aerodynamics had been worked out before, but how fully did we understand the details of bird flight when the Wright Flyer was built?

[Raphael]

(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?

This.

Honest question: I know the basics of aerodynamics had been worked out before, but how fully did we understand the details of bird flight when the Wright Flyer was built?

Not sure about it, but I think I've heard somewhere that earlier attempts at flying failed precisely because they tried to imitate bird flight too closely.

[rotting bones]

This.

Honest question: I know the basics of aerodynamics had been worked out before, but how fully did we understand the details of bird flight when the Wright Flyer was built?

Not sure about it, but I think I've heard somewhere that earlier attempts at flying failed precisely because they tried to imitate bird flight too closely.

Yes. I'm thinking that if they tried to mimic the motions of bird flight in a context where they wouldn't work, maybe that means they didn't understand why it works for birds either, at least at a technical level.

[Starbeam]

Or just not deplete a reservoir and not use it.

1. LLMs don't necessarily dry reservoirs any more than 3D gaming. I run CodeQwen on a machine with a 16 GB graphics card. I ran LLaMA for the narrative project on a laptop with an 8 GB graphics card. I also don't do any 3D gaming, so I probably end up using less power than the leftist influencer bros you listen to.
2. Has anyone compared the environmental degradation caused by running an LLM for a few minutes vs an author obsessing over a piece of writing for months while consuming resources on a first world scale?
3. Blanket opposition to science doesn't come from a principled stance. These are people who are outraged that their livelihood is imperiled. Instead of being luddites, they should oppose the root cause of the problem, capitalism. As long as capitalism is in place, there is nothing we can do to prevent reservoirs from being depleted.

1. What makes you say so? I'll listen
2. I have no idea what this refers to.
3. It does with me, even if i was being snappy there. This is engineering, not science; and it's not the building part that i have issue with. What you're doing is the perfection falacy/ nirvana fallacy. Just because i am not able to solve the root problem does not mean it's bad to address a branch of it. It's not as if i'm anything close to a proponent of capitalism. I'd post on the board's respective thread and rail against it, if it wasn't going to be necroposting. And i'm aware it's not the same thing as activism, the point is i can see the big picture too.

A lot of the "writing help" generative AI offers could be solved by learning how to talk directly, not "people pleasing >_<" or "~expressing oneself~" like an ingrate. Or maybe just looking something up for a few minutes. People never to learn to communicate concisely and i can't stand it. It's all a bunch of advertizement-like drivel.

[WeepingElf]

I wonder whether our current "Too many requests" problem and similar problems of other conlanging web sites are because someone is training an LLM to have it make conlangs, and thereby generating thousands of requests per second. Perhaps some Hollywood CEO has decided that David J. Peterson was too expensive, and they'd rather have an algorithm make the conlangs for their films.

[Starbeam]

One thing that kills me is it is completely possible to do automation without the resource-intensive and thieving slop methods generative AI uses. In fact, it's there in many cases and has been for quite some time. Of course, other automation can't do everything, but i fail to see why we need to drain so much water when nuclear energy could suffice.

[Raphael]

I wonder whether our current "Too many requests" problem and similar problems of other conlanging web sites are because someone is training an LLM to have it make conlangs, and thereby generating thousands of requests per second. Perhaps some Hollywood CEO has decided that David J. Peterson was too expensive, and they'd rather have an algorithm make the conlangs for their films.

Not sure if any Hollywood CEO is aware of the ZBB.

[linguistcat]

... but i fail to see why we need to drain so much water when nuclear energy could suffice.

Do you think the water is to power the data centers? It's to cool them. And it's not like it's gone forever, it goes back to the water cycle. The main issue is that more pure water (low in salts and other minerals) is also the type of water that life prefers to drink, while less pure water is bad for the pipes and can cause them to clog. So when we use water that is preferable for cooling data centers, we aren't using it for drinking or watering crops or whatever. Still an issue, but not the same you seem to think there is from your phrasing. If you meant something else I'd love to hear. I'm not strongly pro or against in this case, I'd just like for people to be pro or against for actual reasons and not imagined bs. But that was all I could image from you comparing AI water usage to nuclear power.

[malloc]

(1) You are vastly underestimating the intelligence of a human.

Election results over the past decade would seem to suggest otherwise. Any species which reelects Trump, Erdogan, and Modi must have its intelligence seriously questioned. That said, my main contention has been that people underestimate the intelligence of computers these days, dismissing all of their incredible capabilities as parlor tricks or otherwise unrelated to intelligence.

(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?

Perhaps so, but the mechanics of human cognition have attracted enormous attention and critical breakthroughs could come at any time. Even without total understanding of the human brain, though, AI has made incredible progress and seems poised to overtake humans in numerous fields.

[keenir]

(1) You are vastly underestimating the intelligence of a human.

Election results over the past decade would seem to suggest otherwise. Any species which reelects Trump, Erdogan, and Modi must have its intelligence seriously questioned.

primates can be bribed. thats not a revelatory idea.

That said, my main contention has been that people underestimate the intelligence of computers these days, dismissing all of their incredible capabilities as parlor tricks or otherwise unrelated to intelligence.

thats because you think there is only one kind of intelligence.

(2) We don't even fully know how a human brain works. How can we excede it if we don't even fully know how it works?

Perhaps so, but the mechanics of human cognition have attracted enormous attention and critical breakthroughs could come at any time. Even without total understanding of the human brain, though, AI has made incredible progress and seems poised to overtake humans in numerous fields.

yes...like doing the same repetitive tasks endlessly, and performing math. oh gods, how will we ever survive alongside counting aids? O.O

[Raphael]

That said, my main contention has been that people underestimate the intelligence of computers these days, dismissing all of their incredible capabilities as parlor tricks or otherwise unrelated to intelligence.

thats because you think there is only one kind of intelligence.

Exactly. That's one of the main problems with malloc (another is that he often seems to think in cliches).

To repeat myself:

Why can every human find their way around the kitchen while many cannot play chess?

Because there's no single thing called "intelligence", there are different mental skills. Having a lot of one mental skill doesn't in any way guarantee that one will be good at another mental skill.

To use a personal example, back when I was in school, I knew people who were a lot better than me at math, but who, at the same time, seemed to have serious trouble understanding what seemed like very simple concepts to me in physics. And math and physics are usually seen as two closely related fields.

Among people more famous than the ones with whom I went to school, there are both people who seem to be very smart about specific STEM subjects, while believing idiotic things about humanities-related subjects (Anatoly Fomenko), and people who seem to be very smart about humanities-related subjects while believing idiotic things about specific STEM subjects (Thabo Mbeki).

Perhaps so, but the mechanics of human cognition have attracted enormous attention and critical breakthroughs could come at any time. Even without total understanding of the human brain, though, AI has made incredible progress and seems poised to overtake humans in numerous fields.

yes...like doing the same repetitive tasks endlessly, and performing math. oh gods, how will we ever survive alongside counting aids? O.O

Yes. To repeat myself again:

malloc, do you still believe that Einstein got so famous for being good at calculating things in his head like a calculator app?

[WeepingElf]

I wonder whether our current "Too many requests" problem and similar problems of other conlanging web sites are because someone is training an LLM to have it make conlangs, and thereby generating thousands of requests per second. Perhaps some Hollywood CEO has decided that David J. Peterson was too expensive, and they'd rather have an algorithm make the conlangs for their films.

Not sure if any Hollywood CEO is aware of the ZBB.

The CEOs themselves perhaps not, but the people whom they pay to develop the conlanging AI would.