I think it is an information virus, but differently - it's homogenized everything, and made people dumber and lazier. It's poisoned public and professional discourse by reducing writing and thinking from the richness of humanity to one narrow style with a tiny latent space, and simultaneously convinced people that this is what good writing looks like. And it's erased thought from board classes of endeavor. This virus is much worse than the relatively benign symptoms described in the article.
There are public sources of information such as a curated WIkipedia, open content from Kiwix, Gutenberg Math books and OpenStreetMap for maps. Better, you can download offline and curated version of these so anyone can have a working snapshot anytime. That's good to avoid future AI tamperings.
As long as these as AI free, we are potentialy in the right direction.
We're going to have to go in the opposite direction and rely on directories or lists of verified human-made/accurate content. It will be like the old days of yahoo and web-indexes all over again.
It’s a method of storing information that makes it far more useful than previous methods.
Sure we can idealize feats of the human brain such as memorizing digits of pi. LLMs put more human behavior into the same category as memorizing digits of pi, and make the previously scarce “idea clay” available to the masses.
It’s not the same as a human brain or human knowledge but it is still a very useful tool just like the tools that let us do maths without memorizing hundreds of digits of pi.
> "Curious Yellow is a design study for a really scary worm: one that uses algorithms developed for peer-to-peer file sharing networks to intelligently distribute countermeasures and resist attempts to decontaminate the infected network".
Hat tip to HN user cstross (as I discovered the idea via Charlie’s blog):
These topics were first brought to my attention through his amazing novel Glasshouse. I’ve had the pleasure of having my first edition copy of the book signed by the author, and I then promptly loaned it indefinitely to a friend, who then misplaced it. The man himself is a friendly curmudgeon who I am happy to have met, and I have enjoyed reading about the future through his insights into the past and present.
Also I must acknowledge Brandon Wiley, who wrote the inspiration for Curious Yellow as far as I can tell.
i heard from an artist that Pinterest is full of AI-generated stuff now so artists looking for references have to go back to physical books for art references
It gives me recollection back to the Simpsons episode where Itchy and Scratchy writers go on strike. What follows was a beautiful scene of children rubbing their eyes in unfamiliar sunlight as they're forced to go outside, making up games, playing on playgrounds, all while Beethovens Pastorale hums in the background.
I'm all for it. Let big tech destroy their cash cow, then maybe we can rebuild it in OUR interest.
I think we should prefer content-farm content to be replaced w/generated content. ultimately it'll compress back to the prompt that generated it and that'll be easier to filter out.
> The only explanation is that something has coded nonsense in a way that poses as a useful message; only after wasting time and effort does the deception becomes apparent. The signal functions to consume the resources of a recipient for zero payoff and reduced fitness. The signal is a virus.
> Viruses do not arise from kin, symbionts, or other allies.
And what if intelligence as measured by computation complexity reaches a natural limit inevitably marked by a detached compassionate disposition? Does the cure then become the virus?
I want to suggest that the virus is even more insidious, and is an organism that feeds on VC money, and it is evolving via a substrate of human programmers to become more efficient at consuming it. And like an organism evolving towards survival, it gives no shits about the utility generated in return for the thing it eats.
And, as time goes on, it'll get more efficient at the consumption and waste less and less energy on the generation of utility. It is an organism that needs servers to feed and generates hype like a deep-sea monster glows its lure.
I'm unconvinced -- Certainly this will happen gradually, and there will be widespread public support for a solution. It's not to hard to imagine making online reviews or "high quality content" require verification tied to some per-citizen identification code (or asymmetric key). Maybe this makes it harder post anonymously on the internet, but at the very least we won't have the issue of proving identity.
Just wish we had a competent government to handle the upcoming transition. But even an incompetent one can have smart employees under it, and can give them the funding they need to accomplish this.
I don't think the results of the last two decades of mass information dissemination have been all that great. People didn't trust anything on the internet before GPT, and the internet was already a cacophony of screeching voices too. Smart people were already decoupling from social media, and moving meaningful interaction to enclaves well before ChatGPT became a factor. ChatGPT did not ruin the internet, it was unleashed on a broken internet to begin with.
If, as this article predicts, the result of GPT is that we don't trust information from the internet, and everybody moves away from it, that's great. Traditional journalism was better, as it turns out. Talking mainly to your friends rather than millions of people was better, as it turns out. I'm ready to go back to that, should it come to it.
But it won't. This essay is making a catastrophic prediction that won't come to pass. Whatever the future is, it's going to be something nobody is predicting yet. It'll be better than the doomsayers predict, and worse than what the cheerleaders say. It will be nothing like a simple magnification of the present concern over epistemology.
28 comments
[ 2.1 ms ] story [ 43.0 ms ] threadSure we can idealize feats of the human brain such as memorizing digits of pi. LLMs put more human behavior into the same category as memorizing digits of pi, and make the previously scarce “idea clay” available to the masses.
It’s not the same as a human brain or human knowledge but it is still a very useful tool just like the tools that let us do maths without memorizing hundreds of digits of pi.
https://en.wikipedia.org/wiki/Glasshouse_(novel)
> "Curious Yellow is a design study for a really scary worm: one that uses algorithms developed for peer-to-peer file sharing networks to intelligently distribute countermeasures and resist attempts to decontaminate the infected network".
Hat tip to HN user cstross (as I discovered the idea via Charlie’s blog):
http://www.antipope.org/charlie/blog-archive/October_2002.ht...
These topics were first brought to my attention through his amazing novel Glasshouse. I’ve had the pleasure of having my first edition copy of the book signed by the author, and I then promptly loaned it indefinitely to a friend, who then misplaced it. The man himself is a friendly curmudgeon who I am happy to have met, and I have enjoyed reading about the future through his insights into the past and present.
Also I must acknowledge Brandon Wiley, who wrote the inspiration for Curious Yellow as far as I can tell.
https://blanu.net/curious_yellow.html
That's how I view LLM's now. They are what follows computers in the evolution of information technology.
Humans still have an inherent need to be heard and hear others. Even in a pretty extreme scenario I think bubbles of organic discussion will continue
I'm all for it. Let big tech destroy their cash cow, then maybe we can rebuild it in OUR interest.
> The only explanation is that something has coded nonsense in a way that poses as a useful message; only after wasting time and effort does the deception becomes apparent. The signal functions to consume the resources of a recipient for zero payoff and reduced fitness. The signal is a virus.
> Viruses do not arise from kin, symbionts, or other allies.
> The signal is an attack.
―Blindsight, by Peter Watts
GPT Might Be an Information Virus - https://news.ycombinator.com/item?id=36675335 - July 2023 (31 comments)
GPT might be an information virus - https://news.ycombinator.com/item?id=35218078 - March 2023 (1 comment)
And, as time goes on, it'll get more efficient at the consumption and waste less and less energy on the generation of utility. It is an organism that needs servers to feed and generates hype like a deep-sea monster glows its lure.
Seems to be working out great so far. (=
Just wish we had a competent government to handle the upcoming transition. But even an incompetent one can have smart employees under it, and can give them the funding they need to accomplish this.
If, as this article predicts, the result of GPT is that we don't trust information from the internet, and everybody moves away from it, that's great. Traditional journalism was better, as it turns out. Talking mainly to your friends rather than millions of people was better, as it turns out. I'm ready to go back to that, should it come to it.
But it won't. This essay is making a catastrophic prediction that won't come to pass. Whatever the future is, it's going to be something nobody is predicting yet. It'll be better than the doomsayers predict, and worse than what the cheerleaders say. It will be nothing like a simple magnification of the present concern over epistemology.