Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Are large language models a threat to digital public goods?

arxiv.org

Post reply on HN
Print view View original

11–20 of 163 posts

Re: Are large language models a threat to digital public goods?

#11

Post by TeMPOraL »

detourdog wrote: post #8
martingalex2 wrote: post #5

This is a very intriguing study as it would appear the fact that LLMs need public data to train on, its incentive on the marketplace of ideas is to reduce the very thing that gives it its power. It contains the seeds of its own destruction, destroying the web and open data ethos, and yet another data point pointing at another AI winter.

For reasons I can't articulate I see LLMs as a vehicle for removing the creators from their ideas. This is very different than search engines. If a search engine generates traffic for documented ideas it creates a community. An LLM based internet seems to remove the creator and shim itself in between for the sake of business.

It's tricky, because in many ways, this is achieving exactly what I, as user, want computers to do for me: give me information I requested, and only that. I very much do not care about who discovered/created/published it, except only if it helps me quickly ascertain the trustworthiness of said information. I do not want to be forced or prodded to establish relationships with creators or communities. I do not want their ads and upsells.

It's the same issue as with search engines providing "information boxes": huge win for me, but a mortal enemy for those who want to monetize anything resembling intellectual property.

> An LLM based internet seems to remove the creator and shim itself in between for the sake of business.

This sounds bad, and in some cases it is, but in others it is not. Content farms and recipe sites have creators behind them too.

Re: Are large language models a threat to digital public goods?

#12

Post by version_five »

I see this as a rough parallel to "is the printing press a threat to illuminated manuscripts". Maybe it is under a very narrow view, but overall, improved propagation of ideas has led to improved dissemination of ideas, and it will this time too. People who's narrow world has been disrupted will perform all sorts of mental gymnastics to tell us how we're going to be worse off for it, but we won't.

Ironically, internet discussion forums democratized specialist knowledge that you previously would have had to be in the right university or research circle to access. SO monetized that, and now we've moved on to complaints like this about further democratization of knowledge and (to quote another post) that knowledge being "stolen".

Re: Are large language models a threat to digital public goods?

#13

Post by endisneigh »

version_five wrote: post #12

I see this as a rough parallel to "is the printing press a threat to illuminated manuscripts". Maybe it is under a very narrow view, but overall, improved propagation of ideas has led to improved dissemination of ideas, and it will this time too. People who's narrow world has been disrupted will perform all sorts of mental gymnastics to tell us how we're going to be worse off for it, but we won't. Ironically, interne…

Doesn’t seem like a valid comparison

Re: Are large language models a threat to digital public goods?

#15

Post by RandomLensman »

TeMPOraL wrote: post #11
detourdog wrote: post #8

Earlier quoted context omitted.

For reasons I can't articulate I see LLMs as a vehicle for removing the creators from their ideas. This is very different than search engines. If a search engine generates traffic for documented ideas it creates a community. An LLM based internet seems to remove the creator and shim itself in between for the sake of business.

It's tricky, because in many ways, this is achieving exactly what I, as user, want computers to do for me: give me information I requested, and only that. I very much do not care about who discovered/created/published it, except only if it helps me quickly ascertain the trustworthiness of said information. I do not want to be forced or prodded to establish relationships with creators or communities. I do not want the…

And the cost of creating the information you want should be borne by whom? Even if it isn't monetizeable IP, how to share the costs?

Re: Are large language models a threat to digital public goods?

#16

Post by abc_lisper »

They used stack overflow to make their case, and report that user engagement has gone down after the release of ChatGPT. Could it not be the case that SO is less adept at finding related/duplicate questions than ChatGPT? Given the later's facility with the language, I would expect it to be. So I look at the paper to see if they accounted for that, and find this.

"Second, we investigate whether ChatGPT is simply displacing simpler or lower quality posts on Stack Overflow. To do so, we use data on up- and downvotes, simple forms of social feedback provided by other users to rate posts. We observe no change in the votes posts receive on Stack Overflow since the release of ChatGPT. This finding suggests that ChatGPT is displacing a wide variety of Stack Overflow posts, including high-quality content."

Can anyone tell me how they are related and what it means?

PS: I also asked ChatGPT (4). Here's what it says https://chat.openai.com/share/7029a1e5-63d0-4cec-bf76-10b0b5...

Re: Are large language models a threat to digital public goods?

#17

Post by adventured »

RandomLensman wrote: post #9
flangola7 wrote: post #7

Earlier quoted context omitted.

AI winter? Labs haven't even began to scratch the surface of the data available to train models on. We may be running out of quality human-written text but we have yet to dive into: Video Audio Images Heat Motion/acceleration Lidar RF Sonar Radar Network traffic Atmospheric pressure Wind vectors Magnetic fields System/application logs Electrical current UV X-ray Microwave Ionizing particle emissions

Interesting, so we chuck in exabytes and more of data generated each day and then what?

The focus should eventually be on building AI such that they can be given access to data and then independently figure out new and creative ways to use it, rather than requiring humans to figure that out for them and then narrowly define: do this thing with the data I've provided. Given the scale of the data, they'll be better suited to that approach if it's breakthrough outcomes that we want.

Re: Are large language models a threat to digital public goods?

#18

Post by Swizec »

RandomLensman wrote: post #15
TeMPOraL wrote: post #11

Earlier quoted context omitted.

It's tricky, because in many ways, this is achieving exactly what I, as user, want computers to do for me: give me information I requested, and only that. I very much do not care about who discovered/created/published it, except only if it helps me quickly ascertain the trustworthiness of said information. I do not want to be forced or prodded to establish relationships with creators or communities. I do not want the…

And the cost of creating the information you want should be borne by whom? Even if it isn't monetizeable IP, how to share the costs?

> Even if it isn't monetizeable IP, how to share the costs?

The internet started thanks to ample government funding for research. So have many other technologies, including AI.

I wonder if there's a way we could all somehow pool our resources and use that to pay for common goods that we all use. What would we call such a scheme?

Re: Are large language models a threat to digital public goods?

#19

Post by TeMPOraL »

RandomLensman wrote: post #15
TeMPOraL wrote: post #11

Earlier quoted context omitted.

It's tricky, because in many ways, this is achieving exactly what I, as user, want computers to do for me: give me information I requested, and only that. I very much do not care about who discovered/created/published it, except only if it helps me quickly ascertain the trustworthiness of said information. I do not want to be forced or prodded to establish relationships with creators or communities. I do not want the…

And the cost of creating the information you want should be borne by whom? Even if it isn't monetizeable IP, how to share the costs?

I don't have good answers. I have some high-level intuitions.

One of them is that creation costs of information are fixed, while its usefulness is unbounded, so it doesn't make sense to try and reward creators for each access/view/use, in perpetuity.

Secondly, there's a lot of information laundering going on - any random book I read carries between a few to few hundred references to prior written work. What I pay for the book goes to the author and the publishers, but AFAIK it doesn't go to any of the authors and publishers of works referenced in the book. Wikipedia takes this one step further, effectively turning all that information free.

Thirdly, AFAIK copyright explicitly does not cover information/knowledge - it covers specific works. So Google showing me an info box with a recipe scrapped from some site could technically fall afoul of the law - but an LLM generating me a recipe based on associations created from being trained on millions of recipes, this feels like it should be in the clear, at least from user's POV.

Re: Are large language models a threat to digital public goods?

#20

Post by TeMPOraL »

abc_lisper wrote: post #16

They used stack overflow to make their case, and report that user engagement has gone down after the release of ChatGPT. Could it not be the case that SO is less adept at finding related/duplicate questions than ChatGPT? Given the later's facility with the language, I would expect it to be. So I look at the paper to see if they accounted for that, and find this. "Second, we investigate whether ChatGPT is simply displ…

> We observe no change in the votes posts receive on Stack Overflow since the release of ChatGPT. This finding suggests that ChatGPT is displacing a wide variety of Stack Overflow posts, including high-quality content.

My alternative hypothesis: ChatGPT is eliminating all the long-tail, garbage questions, so the vote patterns don't change because the missing questions weren't getting much of them anyway.

Post reply on HN