fluidcruft
fluidcruft on Hacker News
People who don't vote don't oppose the winner.
A federal court tried to conceal the identity of a judge who has been reprimanded.
Claude was able to work out who it was in less than 10min based on stringing together facts that remained and seemed irrelevant. Journalists uncovered the identity independently using different means. (The ability of Claude to do this was not revealed until the other journalists had gone public and confirmed the identity.)
Flock has no power vs the commerce clause. Congress just has to use it.
This battle was lost under Bush
I recommend listening to the LegalTech Week podcast. Generally lawyers are finding it's not delivering efficiency overall. They spend more time reviewing the output. The courts are drowning in slop and deepfake evidence is a growing concern. The jury is still out on all of this.
As I understand it, AI is only really getting traction in IT and software development. It's been mostly irrelevant elsewhere overall, all things considered. That could change, but basically there's only one product that is selling. Dunno whether there is $500b/year of IT and programming to replace.
My impression is Claude is trained to work in production systems and behaves that way instinctively. Some copypasta/context I often feed it is that I am the only user, no production data exists, any data that exists is mine and I accept responsibility for breakage, there is no public release yet, avoid compatibility shims, etc. Stuff like that. Claude will usually get right on board with having a window to make big changes now to avoid technical debt. But you are correct its instincts are very conservative.
I believe the theory is that if selection pressure is not applied at that level, then developing deception in that level becomes an inefficiency (=less adaptive and unlikely to be conserved).
idk but naive me sees the article say the model needed 158GB of RAM so I guess 10x 16GB GPUs? Apparently retail price is $1k per 5080 16GB but they're actually selling at $1.5k so I guess $15k for just the GPUs.
Even if you don't actually need a license, I would assume that making it an explicit term of the contract makes the dispute shorter.
There's always Edge (increasingly favored by IT here). Our machines used to ship and favor Chrome, but Edge has been the new IT-gods blessed browser ever since we bumped to Windows 11. Apple seems to be the only tech-titan that is not interested in extending their browser onto other platforms. So be it.
Misaligned to exploit the trial process is actually something to consider carefully. I would not put it past a drug company to design a phase 3 trial that has some subtle bias towards commercialization. And if we're just getting cocktails and trial protocols from the AI, it would be very wise to be skeptical.
For example, if the prompt is something like "give me a cocktail that is most likely to succeed in this specific phase three trial" there are any number of ways the AI might find ways to game specific features of the trial design.
You lasted longer than me, I bailed when they called a senator a congressman--partisan diatribe detected, no value to be had.
(there was an animated series before the movie)
Yep, their work used to be in a mono repo. That's still up there. They have a reference to how to find it. I didn't read all the history but it does date back to 2016. Seems like it was mostly a web app and not open source until recently is the vibe I was getting from it.
If you're talking about the artcraft stuff, it's strange. The codebase goes back to 2016 as a monorepo but not the full suite and was recently split out as separate apps and other parts of the Adobe suite added. The CLAUDE.md [1] are... interesting and left me puzzling about it.
I was mostly interested in the pdfcraft (have been appreciating
https://www.nobloatpdf.com/)
> Clean-room.
> * Never read, disassemble or copy anything inside the Acrobat bundle (names and listings only). Never open Contents/Resources/JavaScripts/.
> * Behaviour comes from public docs, specs (ISO 32000-2, the Arlington model) and black-box observation (plan/acrobat/).
> * Never copy GPL/AGPL code. MuPDF, Ghostscript, Poppler, veraPDF and DSS run only as external oracle processes.
> * See plan/README.md §Clean-room and plan/adr/0001.
It really looks like they have AI running Adobe (and other apps) as sources of truth and have rules against disassembly or including source from GPL projects which left me pondering the ethics of it.
[0] https://getartcraft.com/
[1] https://github.com/storytold/pdfcraft/blob/main/CLAUDE.md
Sure, unverified claims are unverified so there's a real risk of reputational tarnish. Why is it anyone else's job to verify a papermill? Would any journal accept an article that doesn't name people who are accountable? It should correctly be viewed as abuse of peer review. Let them rot.
I guess I feel two ways about this. It does could be a "working with the garage open" sort of thing where these are known to be unverified and they're just letting people see the sausage being made.
But that also means marketing and PR should shut the fuck up until things are verified. And they should more clearly annotate lack-of-verification status on their repo.
Whose names are on these papers?
It reads like it too. I don't fully understand why a sentence like "It's the signature that costs." makes it into machine output (It costs... what? What does it cost?) But here we are.
States could harmonize laws (if they wanted to).
That would have been a scathing criticism if anyone believed any of the American companies have the goal of helping humanity or open-source.
Notwithstanding that Chinese publishing methods actually does help both humanity and open-source.
I remember seeing some documentary about an art forger who donated his works to museums where the authorities were profoundly frustrated that they couldn't do anything precisely because he didn't get any gain.
Edit: the documentary is "Art and Craft" [1] about Mark Landis [2]
> It appears that in donating forgeries to art museums, Landis has not actually broken any laws, even though his activities were clearly deceitful. If he had sold the work to museums or taken a tax deduction on them, he might have fallen under federal art crime statutes. But the fact that he did not gain economically from his actions (apart from a few gifts from curators), and that he addressed his donations to specialists who had the expertise to detect his forgeries but did not, protected him in the eyes of the law. No legal action has been opened against him to date (as of 2014). As one art crimes expert put it: "Basically, you have a guy going around the country on his own nickel giving free stuff to museums."
[1] https://artandcraftfilm.com/
[2] https://en.wikipedia.org/wiki/Mark_Landis
That's a great option.
I just wonder whether that could be too annoying for Android Auto / Car Play. But to be fair wireless is an option.
The other downside I guess is that a thief would just plug a USB drive in to disable device tracking. But they could just wrap the damn thing in foil so dunno if that should matter.
A Book Apart shut down and says all copies have been sold. They suggest reaching out to Rachel Andrew directly to see if the book is available elsewhere. (Her website just links back to A Book Apart).
You can just take a picture of a pile of dice, a pile of rice, or a tree, patch of grass, etc, and compute a secure hash/whatever and base six it to get the rolls.
Why not automatically power down if any unknown USB device is attached?
Why hasn't that happened in Antarctica?
Governments shouldn't "want" anything. If people want more taxpayers, so be it.