LLMs are fine. People who use LLMs on copyrighted data, though, are doing it illegally.
(Edit: train, not use, of course.)
Are large language models a threat to digital public goods?
151–160 of 163 posts
Re: Are large language models a threat to digital public goods?
#152Earlier quoted context omitted.
They will and less information will be put into places that are freely accessible. If it's put anywhere at all it'll be put behind login only/paywalled/unscrapable places that LLM's can't access.
Why would they suddenly paywall information if they weren't already? The way I see it, there are roughly three groups of information providers: 1. Those who do it pro bono - because they feel like its a worthwhile thing to do, or because they believe in by "pay it forward", or otherwise because they haven't even thought that what they share is worth trying to extract rent from. 2. Those who do it "for free", as a way…
Group #2 is the reason why there's so much information on the internet in the first place, especially free resources, for better or worse. I think it's pretty ignorant of you to say we can just discard one of the main ways people who create things on the internet get paid.
Re: Are large language models a threat to digital public goods?
#153Earlier quoted context omitted.
Yes, maybe good to put the ideas around "dataism" (don't know a better word, sorry) really to the test.
People have been testing them for ages. There's something there, but not nearly as much as the hype implies. At the same time, the hype seems to only increase. Anyway, I really like that name. Non-dataist AIs are clearly the best kind.
I really think the deep learning maximalists are leading us down a rabbit hole. We're going to waste a lot of digital storage and computing resources on creating bigger and bigger models for more and more diminishing returns, barring some breakthrough. Without a way of encoding expertise, which is limited right now, I don't think we're going to get the kind of performance that the uninformed public expects out of these systems.
Re: Are large language models a threat to digital public goods?
#154On removing knowledge from the public eye:
Is it really removed if it is easier to get, and at a higher quality than a lot of SEO spam search results? On a primitive level I'm not sure it's much different than when computers & calculators replaced printed mathematical tables, such ones used that were used in artillery firing.
That said, there are other consideration of accessibility separate from the broader topic LLM's in general. Specifically, that LLM's truly suited as better alternatives to traditional content might end up gate-keeped by corporations charging high prices, an especially likely scenario if lawsuits regarding copyright rule in favor strict copyright protection of the work as not eligible for a fair use exemption. And/or if regulatory capture puts the barrier of entry into LLM's creation too high. Contrary as it might seem, losing the copyright battle might be a net benefit to Open AI and large competitors: It will be very difficult for more open alternatives to get the resources required to build open models.
I don't think use of LLM's, in themselves, constitutes a risk to public knowledge, in so far as they remain just as or more accessible than traditional content. That is the case right now, for some use cases, where I can get a much faster answer and immediately critique or get follow up responses to clarify things.
Re: Are large language models a threat to digital public goods?
#155Earlier quoted context omitted.
Regarding the baker example, some form of compensation is eventually directed to the employees, the farmers etc. even though you could say there are many layers of indirection. In case of LLMs, no compensation is directed to the person authoring the information. While it may not be a problem for the consumer of the information, it removes any incentives for the people authoring the information to continue doing so, w…
Depends what the incentives are. Some folks like the community- that doesn't change. A website like honestwargamer or goonhammer has a clan. Mostly because they are insightful and friendly. But if an AI scraped their content, it would have to build a better goonhammer? Maybe. But I bet it would have half a dozen nerds contributing and tinkering at the backend, painting the new minis that came out, talking about tourn…
That's a somewhat one-sided view. For gaming and entertainment, yes, people would do it anyway, since it's just fun, but these do not contribute much useful information to the collective consciousness anyway. Hobbies & creative communities will also survive.
OTOH, there are also plenty of technical blogs full of advanced content that is not "fun" to produce on its own, that are written to interact with a community of professionals (or juniors), and that might wither if engagement with actual human beings is reduced.
Re: Are large language models a threat to digital public goods?
#156Earlier quoted context omitted.
Ideas are copied by reading or hearing them. You can't own your ideas now, unless by own you mean horde. The perpetual creators rights you want extended are already artificial and require a non-trivial amount of our GDP to enforce and they still stifle future creation in a lot of areas. Most people are paid for doing things every day, they don't get to create one thing and never work again. Expanding creators compens…
Your entitlement to the labor of others is gross.
You're writing in English, which I doubt you came up with on your own, and I don't see you crediting the original speakers who developed your style or popularized the idioms you so casually use.
How do you claim the right to learn from the works of others and then demand government regulation and forceful intervention to keep from having to share whatever paltry innovations you may develop?
Re: Are large language models a threat to digital public goods?
#157The paper title is clickbait and therefore the HN comments are focused on various low-effort reactionary noise, but it seems like an interesting study. I would not be surprised to find StackOverflow usage dropped significantly because of ChatGPT. It's simply a much more effective tool for getting help with typical programming problems. Not as good of a resource for expert-level or architectural advice, but that's oka…
If you're only pasting backtraces, you're not taking advantage of its coding ability. You can describe the problem, have it write some code, then iterate by telling it what to add, cases you want it to handle, add features to the code. It's been doing a pretty good job of helping me designing a database down to the actual CREATE TABLE statements for me as well.
Oh, I know, it's phenomenal. Can't wait for it to be a little smarter and more directly integrated into docs/search.
Re: Are large language models a threat to digital public goods?
#158Earlier quoted context omitted.
Your entitlement to the labor of others is gross.
You're writing in English, which I doubt you came up with on your own, and I don't see you crediting the original speakers who developed your style or popularized the idioms you so casually use. How do you claim the right to learn from the works of others and then demand government regulation and forceful intervention to keep from having to share whatever paltry innovations you may develop?
Pretty dim view on humanity you have there.
Re: Are large language models a threat to digital public goods?
#159Earlier quoted context omitted.
> 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. And this is why humanity is going down the tubes...because you want something, you derive value from what you want, and yet you do no…
> And this is why humanity is going down the tubes... On the contrary - this is exactly how and why humanity built a technological civilization in the first place. Note that I didn't say > because you want something, you derive value from what you want, and yet you do not care about giving something back to who makes it. Yes, because it would be backward and limiting to do that. Note: I never said I don't want to giv…
Yes, because it would be backward and limiting to do that. Note: I never said I don't want to give anything back - I said I don't personally care specifically about the author/publisher. I don't want to establish any relationship with them. If I'm paying them directly for something, I'm paying them for that thing - not also for relationship (which really is a sales channel), not also for being advertised to. If I'm paying some intermediary, then rewarding the maker is the intermediary's problem, not mine.
And I maintain that your attitude is one that makes the world worse. We should know where things come from and not have an anonymous transactional attitude towards it. Our technological civilization as you call it has led us to destruction with only a minority benefitting.
BTW, my favourite place to get bread is one in which they actually sell and bake it in store (a tiny market with its own oven).
Re: Are large language models a threat to digital public goods?
#160Earlier quoted context omitted.
If you're only pasting backtraces, you're not taking advantage of its coding ability. You can describe the problem, have it write some code, then iterate by telling it what to add, cases you want it to handle, add features to the code. It's been doing a pretty good job of helping me designing a database down to the actual CREATE TABLE statements for me as well.
Oh, I know, it's phenomenal. Can't wait for it to be a little smarter and more directly integrated into docs/search.
You mean like this very helpful instance where it just makes shit up that directly contradicts the docs that are on the exact page it's on?