JoshTriplett

159 ms·

2551.
by JoshTriplett 3y ago
I'm not talking about the ones that provide a user-contributable database of "what business owns this number, is it spam"; those are great. I'm talking about sites that just list all possible phone numbers in a range,
2552.
by JoshTriplett 3y ago
And the sites that list every possible phone number in a given area code, just so that they'll be a search hit for someone searching for a phone number. Zero value provided there.
2553.
by JoshTriplett 3y ago
Copy (before answering the "is this your final password" question) and paste seemed to work fine on everything except the chicken emoji, so I just had to paste and then manually enter and italicize a chicken and two bugs within th
2554.
by JoshTriplett 3y ago
"Congratulations! You have sucessfully chosen a password in 113 characters." This was a horrific abomination, and delightfully evil. My solution (which includes some spoilers, even though each game has somewhat different rules):
2555.
by JoshTriplett 3y ago
> For anyone else struggling with how to make the country name work with roman numerals or element names, you can lowercase the country name then it doesn't count as a roman numeral nor element. Unfortunately, "country name&quo
2556.
by JoshTriplett 3y ago
Yeah, Infinite Jukebox works surprisingly well, and it's fun to manually tweak some of the probabilistic looping points to create something fun.
2557.
by JoshTriplett 3y ago
And locking down the primary device a kid has access to creates the lure of the forbidden; that won't last any longer than their access to another device without those restrictions. That also creates an environment where if they do f
2558.
by JoshTriplett 3y ago
You're assuming that an AI model would be "inspired by" rather than "making derivative works of".
2559.
by JoshTriplett 3y ago
> But the team found that as they stretched the iron selenide, its electrons began to overwhelmingly prefer one orbital state over the other. This signaled a clear, coordinated shift, along with a new mechanism of nematicity, and superco
2560.
by JoshTriplett 3y ago
> If the answer is "humans get special treatment" then that's fine I guess but I think it's worth being explicit that that's the difference. That's absolutely the difference. Humans aren't copyrightable
2561.
by JoshTriplett 3y ago
> Would this also apply to home directory, eg ~/.meta/.netrc? That would be neat and make backing up a system and synching configs easier. ~/.config and ~/.local already exist for that purpose: applications should rea
2562.
by JoshTriplett 3y ago
And the blast radius of AIs continuing to violate licenses is not limited to large companies either, it's harming Open Source developers, independent artists and authors, and similar.
2563.
by JoshTriplett 3y ago
I'm familiar with the widespread practices of how AI models are trained, yes. There's an implicit "and we must be able to do this" in your argument, which is not at all evident. "But what about AI" is not an ar
2564.
by JoshTriplett 3y ago
> I completely agree that expanding copyright only hurts us all. As do I, but this is not expanding copyright. This is saying that as long as copyright exists , AI model training has to respect it too, and can't be a laundering ope
2565.
by JoshTriplett 3y ago
> Such a strict interpretation of copyright would kill Open Source, as any attempt at reverse engineering would be disallowed by that. Not at all, for multiple reasons. Reverse-engineering already has that problem: if any of the people
2566.
by JoshTriplett 3y ago
> If you find yourself saying "In other words," it means you didn't say it clearly enough the first time. Go back and rewrite the first attempt. Sometimes, particularly in didactic material, it helps to explain things in m
2567.
by JoshTriplett 3y ago
Clarifying: the part I'm arguing is "self-evident" is that an LLM is a derivative work of its training data, in the same sense that if you copy the text of a million books into a data file and compress that file reversibly in
2568.
by JoshTriplett 3y ago
> In Github's case for example, Github's TOS already includes a clause that if you upload code there, you grant Github a license to use the content to run Github's services... and copilot is one of them. You can't gra
2569.
by JoshTriplett 3y ago
Making LLMs more introspectable and explainable would be a huge win. But respecting licenses doesn't require going that far. 1) Restrict training to code under a specific documented set of acceptable licenses (permissive Open Source li
2570.
by JoshTriplett 3y ago
Developers are free to do so, if they want that. But that doesn't make it OK to train an AI on code not under such licenses.
2571.
by JoshTriplett 3y ago
A human is not copyrightable. And copyright is doing its job just fine in this context; AI training committing Open Source license violations en masse is the problem here. I'm all for copyright becoming substantially weaker or ceasing
2572.
by JoshTriplett 3y ago
> It's not clear that replacing the human with an AI somehow changes the equation so categorically that it becomes automatic to consider the output of the AI a derivative work. See below: there are good reasons for an AI LLM to be c
2573.
by JoshTriplett 3y ago
> What's the difference between using a pencil to write something and using an LLM to write something? The pencil is not a derivative work of a pile of copyrighted material. > Why does one produce something copyrighted why the ot
2574.
by JoshTriplett 3y ago
> Not having full control over the compiler is generally bad if your product needs to be supported for a long time or has to undergo some type of certification. rustup, and this mechanism in Go, don't take away your control over the
2575.
by JoshTriplett 3y ago
> 2. Why making a difference between AI and HI (Human Intelligence)? Because you can't copyright a human brain, and because humans (unlike machines) can themselves create works subject to copyright.
2576.
by JoshTriplett 3y ago
> Many licenses are reasonably clear that this kind of use is not acceptable, as is easily demonstrated by these "AI"s frequently producing exact matches without license statements. Which is unambiguously not allowed by many li
2577.
by JoshTriplett 3y ago
We don't need Open Source licenses to change to deal with AI. We need AI to respect Open Source licenses, or not use code under those licenses. I sincerely hope that one of the many court cases produces a verdict that says AI-generated
2578.
by JoshTriplett 3y ago
> - max of 9 relevant occupancy bits for a bishop - how? A bishop on e4 for example would have to check b1,c2,d3,f5,g6,h7 and a8,b7,c6,d5,f3,g2,h1 for a total of 13 bits. Similar mistake for a the rook. If I'm understanding correctl
2579.
by JoshTriplett 3y ago
No, I mean the latter: picking "not embedding a whole web browser" is worth it for the user-friendliness, even though it means having to deal with those browser differences, which have grown a lot less onerous than they used to be
2580.
by JoshTriplett 3y ago
Exactly; electron only gives you a benefit on that front if you don't otherwise have a need to be portable.

More ›