I think smart bulbs mi
drosophila
joined 8 months ago
Being able to summarize and answer questions about a specific corpus of text was a use case I was excited for even knowing that LLMs can't really answer general questions or logically reason.
But if Google search summaries are any indication they can't even do that. And I'm not just talking about the screenshots people post, this is my own experience with it.
Maybe if you could run the LLM in an entirely different way such that you could enter a question and then it tells you which part of the source text statistically correlates the most with the words you typed; instead of trying to generate new text. That way in a worse case scenario it just points you to a part of the source text that's irrelevant instead of giving you answers that are subtly wrong or misleading.
Even then I'm not sure the huge computational requirements make it worth it over ctrl-f or a slightly more sophisticated search algorithm.
Even if it takes 100+ years for quantum cryptanalysis to become viable I would rather we start switching over to better algorithms now.
Nerds don’t just want to teach people to drive. They want to teach them about the engine, the drive train, the underlying transportation infrastructure, and how to change their own oil and tires.
Maybe if more people knew how combustion worked and where the gasoline they burn comes from we wouldn't have as much global warming denialism.
Similarly, if people knew how their posts were served though Facebook, what server costs are, and what their revenue model was, it wouldn't come as such a surprise to them that their privacy was being violated.
But I think you're right though. I've given up on trying to convince the general public of literally anything, at least in the US where it's clear the cult of ignorance has soundly won. How can I tell someone that it's better to use an electric car if they're not willing to understand the carbon cycle? How can I tell someone it's better to be vaccinated if they're not willing to understand herd immunity? How can I tell someone that federated social media is better if they're unwilling to understand what federation even is?
Are you misreading “preparing” as literally any writing
"Prepare derivative works" means not just any writing, but literally anything creative. If you paint a picture of a character from a book, using specific details described in that book such as their appearance and name, you are creating a derivative work.
https://law.stackexchange.com/questions/78442/what-is-considered-a-derivative-work
Even that Wikipedia article goes into fair use.
Fair use carves out an exception for parody, criticism, discussion, and education. "Entertainment" or "because I like the series and these characters" are not one of those reasons. Fan fiction might qualify as parody though.
What effect on the market can there be for a fan remaster of a 20 year old game that isn’t for sale anymore? Hard to argue that doesn’t fall under fair use.
This is not how "the effect of the use upon the potential market for or the value of the copyrighted work" part of fair use works.
A company can create a work, sit on it for literally 100 years doing nothing with it and making not a single cent from it, then sue you for making a nonprofit fan work of it. Steamboat Willie is 95 years old and until just this year you could have been sued for drawing him. Note that, in the eyes of the law, Steamboat Willie is effectively a different character than Mickey Mouse.
Again, I cannot stress enough how it doesn't matter at all whether you are personally profiting from something or whether you are affecting a market. The word "potential" in that quote above is doing a lot of work:
A father in the UK wanted to put spiderman on the grave stone of his 4 year old son who loved the character. Disney said "no". Disney does not make tombstones. You are not eating into their profits by putting spiderman on a tombstone. And yet in the eyes of the law Disney has every right to stop you since they might decide to start up a tombstone business next week.
Nothing I have written here is legal advice.
EDIT: I am not a fan of any of this. I think you should be able to write nonprofit fanfiction without worrying that some corporation might sue you. I am on your side on this. But this is the reality we live in.
People are allowed to write fanfic and make fan movies and whatnot. The line isn’t crossed until money changes hands.
This is completely wrong. A company is fully within their rights to issue you a cease and desist for fan works. Some companies, like Disney and Nintendo, do this all the time (though sometimes people are able to fly under the radar).
If you see a free fan game or fan work of anything it's completely at the mercy of the company that owns the IP. If it's not taken down it's either because the company is cool with it, not aware of it, or can't be bothered to deal with it.
EDIT: https://en.wikipedia.org/wiki/Legal_issues_with_fan_fiction
People really have no idea how overbearing IP laws are. Technically even recordings of people playing video games (let's plays and the like) could be infringing. This hasn't been extensively argued in court because most game companies don't want to deal with the PR backlash that forbidding let's plays would cause (in addition to the free advertising they get). Though, once upon a time that didn't stop Nintendo from using YouTube's copyright system to claim videos of their games.
https://www.ign.com/articles/2013/05/16/nintendo-enforces-copyright-on-youtube-lets-plays
https://www.slaw.ca/2024/02/07/lets-plays-a-copyright-conundrum/
This is a symptom of the absolutely insane way digital payments work.
You give a company your card details and they're able to charge whatever they want, whenever they want, by default. That's like paying at a restaurant by handing the waiter your entire wallet and telling them to take out the cost of the meal.
Well, Fromsoft had a good run.
Maybe we'll get one more game out of them.
I asked them to support JPEGXL by default.
Redfall being a prime example. We kept hearing how Microsoft was happy to leave those studios to it, to give them the time and resources they needed and they still released dog shit.
Yeah, the studio that developed Prey (a dumbass name that zenimax forced them to use) went on to develop Redfall after Microsoft bought them.
Clearly they were a bunch of idiots before the acquisition who had no idea what they were doing, and the only problem afterward was that Microsoft didn't boss them around enough.
That's what Google was trying to do, yeah, but IMO they weren't doing a very good job of it (really old Google search was good if you knew how to structure your queries, but then they tried to make it so you could ask plain English questions instead of having to think about what keywords you were using and that ruined it IMO). And you also weren't able to run it against your own documents.
LLMs on the other hand are so good at statistical correlation that they're able to pass the Turing test. They know what words mean in context (in as much they "know" anything) instead of just matching keywords and a short list of synonyms. So there's reason to believe that if you were able to see which parts of the source text the LLM considered to be the most similar to a query that could be pretty good.
There is also the possibility of running one locally to search your own notes and documents. But like I said I'm not sure I want to max out my GPU to do a document search.