30,216 karma · joined February 24, 2007
Twitter: http://twitter.com/gojomo
Project: http://thunkpedia.org
Idea blog: http://memesteading.com
Older blog: http://gojomo.blogspot.com
Homepage: http://xavvy.com (username @ here for email contact)
My HN peeve is formulaic downbeat comments, like: "How is this news, I already knew this!" "…Betteridge's Law…" "I stopped reading at…"
But even if the capture negligible, the negative-sign is still interesting - it's teased in the headline! And when a headline teases something, it owes us an explanation of where the carbon goes, a key part of the story to anyone with a middle-school-level understanding of 'carbon'.
Barring nuclear reactions or alchemical magic, it's impossible that "its only byproduct is water". The carbon has to collect somewhere.
Science journalism like this makes its readers dumber.
That leaves it so incomplete about its topic that the link should be replaced with a better one if possible.
Though this concept isn't as pat and reliable as in its most-simplistic formulations – you can't fix all crime with aesthetic enforcement – it captures real human tendencies to 'flock' in the space of norms-of-behavior, using visible cues of what will or won't be tolerated. And, it has applicability outside of just literal 'policing'.
Keeping spaces/communities far from any chaotic boundary where people start to wonder - "what can I get away with? does anyone confront/correct problems?" - can save a lot on overall defection/enforcement losses in the long run.
Just formalizing the automatic exclusion of all these publishers' in-print ebooks is no loss at all, compared to the rulings so far in the lawsuit. It is a bit unclear to me what happens if the IA gets a more-favorable ruling from their appeal - have they still agreed to not perform 1:1 lending of physical-book-derived ebooks for the plaintiffs, even if they win the right to do so?
Do you think that because personal computers & the internet took a while to develop, states like California missed an opportunity to helpfully regulate them by forcing their makers to attest to their safety before product development (much less public release) gets underway? Should we add that sort of regulation, now?
A new website could be harmful – used for "serious crimes", "bill pay", "litigation", "even war"! Should every new website require filing paperwork guaranteeing its safety with a Calfornia Department of Technology division before going live? (There were oppressive regimes that tried to control presses, printers/fax-machines, & websites like this, using "safety" rationales!)
The mere fact a new technology is a "powerful tool for humanity" is not something that must "naturally require" a novel state-bureaucracy-run "regulatory framework around it to ensure its proper use and application".
The state in general, and the State of California in particular, is not our wise cloud-father with the foresight & disinterest to do what's best for us. It's instead a clumsy and often-corrupted tool for solving some common-coordination problems.
States usually do best when addressing a well-understood common history of specific problems & market-failures – rather than improvising new filing requirements against theoretical fears, as here.
Any new "very powerful tool for humanity" deserves the same freedom from prior restraint, & forebearance from premature budens that mostly benefit incumbents and large players, that prior technological innovations enjoyed.
But the interactions of tradition, legal precedents, and different laws in different places are complicated, and subject to interpretation & various shifting balances-of-considerations!
That's why it's silly when armchair opinionators casually judge, "IA clearly did something wrong". There's a lot of room in the law for fair-use copying, even against the rightsholders' strongest wishes, and especially for librarylike activities.
With the advice of competent counsel, IA clearly estimates its chance of prevailing on that question as worth the continuing costs.
At worst, they lose on that question, and then comply with the August 2023 settlement & pay some undisclosed amount to publishers to cover publishers' legal fees.
At best, they set an important precedent that saves their (and many others libraries!) ability to lend in the digital era for the one-time cost of a physical book, instead of the new time-limited leasing models publishers are imposing for native ebook lending – which could massively increase library costs even though digital delivery could be far cheaper than physical lnding.
Note that even if the odds are long, setting that precedent would be very valuable! The calculation isn't, "are we more than 51% likely to win?", but "even if we've only got a 5% chance of winning, is the win valuable enough to spend for the chance?".
What if Sony had given up when the 'Betamax' case was going against it? It appealed to the Supreme Court, and won a unambiguous ruling that home recording of TV was legitimate fair use.
That's a result the rightsholders, and those who naively adopt rightsholder-like copyright-maximalist reasoning without acknowledging the counterbalancing factors in the law, had alleged was preposterous. "Clearly illegal!" they said. But despite that bluster, the actual highest court decided otherwise.
Without that ruling, VCRs and DVRs would have arrived far slower & on a short-leash fully controlled by incumbent big studios.
It's just that the lower court's judgement didn't really focus much that issue, because they found the broader/simpler and more-foundational "can you loan a format-shifted ebook 1:1 from your physical book" issue against IA, which then makes an anti-NEL conclusion almost automatic.
So NEL-related arguments haven't yet been the focus of the rulings, despite being part of the original lawsuit.
Thus, also, IA's appeal is mostly about that foundational finding – though section II of the IA appeal brief says that if the appeal succeeds on "ordinary controlled digital lending", the NEL ruling should also be rejudged:
>Il. THIS COURT SHOULD REMAND FOR RECONSIDERATION OF THE NATIONAL EMERGENCY LIBRARY Publishers do not deny that the district court’s NEL ruling depends entirely on its analysis of ordinary controlled digital lending. Resp.Br. 61-62. If the Court reverses the latter, then it should also reverse and remand the former. It need not address Publishers’ arguments about the justifications for NEL (Resp.Br. 62), which should be left for the district court in the first instance.
You demonize Hachette et al (4 major publishers) as seeking to destroy a public good. In fact, they've already settled with the IA, as of last August 2023, in a manner that caps costs to IA at a survivable level and sets clear mutually-acceptable rules for future activity.
You imply IA would dismiss the appeal if the plaintiffs "could drop enforcement of the judgement". In fact, there were never any assessed damages, the parties have already reached a mutually-acceptable settlement per above, and despite that – in fact, as part of the settlement! – the IA has retained the right to appeal regarding the fair-use principles that are important to them.
Per https://en.wikipedia.org/wiki/Hachette_v._Internet_Archive#F...
>On August 11, 2023, the parties reached a negotiated judgment. The agreement prescribes a permanent injunction against the Internet Archive preventing it from distributing the plaintiffs' books, except those for which no e-book is currently available,[3] as well as an undisclosed payment to the plaintiffs.[25][26] The agreement also preserves the right for the Internet Archive to appeal the previous ruling.[25][26]
IA's August 2023 statement on how much will continue despite the injunction & settlement limits: https://blog.archive.org/2023/08/17/what-the-hachette-v-inte...
>Because this case was limited to our book lending program, the injunction does not significantly impact our other library services. The Internet Archive may still digitize books for preservation purposes, and may still provide access to our digital collections in a number of ways, including through interlibrary loan and by making accessible formats available to people with qualified print disabilities. We may continue to display “short portions” of books as is consistent with fair use—for example, Wikipedia references (as shown in the image above). The injunction does not affect lending of out-of-print books. And of course, the Internet Archive will still make millions of public domain texts available to the public without restriction.
"Save itself"?
Despite occasional prior histrionic kayfabe about "IA in existential danger" in the media (& HN threads) – sometimes fanned by the IA's supporters themselves – that's never been the real stakes.
For the serious librarians, publishing businesspeople, and lawyers involved, this has been about legal clarity for a gray area at the intersection of copyright, fair use, & traditional rights of first-sale and library practices. It's not really about damages, nor the IA's (or traditional publishers') existence. Instead: the principles controlling what's allowable going forward.
To that end, per Wikipedia, as of eight months ago (August 2023), the lawsuit parties already reached & had the court approve a negotiated settlement that caps the potential costs to the IA at a survivable level, & sets ground rules for future similar e-book activities that the Hachette et al (4 major publishers) plaintiffs and AAP (publishers' trade group) find acceptable. But further: this mutual settlement permits IA to continue its legal appeal on the principles involved.
From <https://en.wikipedia.org/wiki/Hachette_v._Internet_Archive#F...>:
> On August 11, 2023, the parties reached a negotiated judgment. The agreement prescribes a permanent injunction against the Internet Archive preventing it from distributing the plaintiffs' books, except those for which no e-book is currently available,[3] as well as an undisclosed payment to the plaintiffs.[25][26] The agreement also preserves the right for the Internet Archive to appeal the previous ruling.[25][26]
That is: the publishers were never b-movie villains trying to destroy a public resource; the IA was never reckless anarchists gambling all its other programs for a quixotic legal precedent. They were all adults with a legitimate legal dispute about what's allowed, seeking a clear definitive resolution in the culturally-appropriate manner.
And via the settlement and appeal, the parties are still working out the issues.
This author misdescribes the IA as "a profitable enterprise (bringing in between $20 and $30 million per year) that is on the verge of a potentially devastating legal ruling which could put [it] out of business". But IA is a non-profit, arguing for a mission-critical principle – a principle which is a plausible extrapolation of existing fair-use rights and library/IA practices into a new domain. And it's doing so with explicit permission under the existing settlement, capping financial risks far below any existential risk.
This author further deceptively excerpts IA's central argument as being just "Controlled digital lending is not equivalent to posting an ebook online for anyone to read". Against this, the author writes, essentially, "nuh-uh, that's exactly what they did".
In fact the full necessary context of IA's argument is:
> First, Publishers disregard the key feature of controlled digital lending: the controls that ensure borrowing a book digitally adheres to the same owned-to-loaned ratio inherent in borrowing a book physically. Publishers repeatedly compare IA’s lending to inapposite practices that lack this key feature. Controlled digital lending is not equivalent to posting an ebook online for anyone to read or copy (contra Resp.Br. 27) or to peer-to-peer file-sharing by companies like Napster (contra Resp.Br. 5). Neither practice is based on use of a library’s lawfully acquired physical copy, and neither ensures that only the one person entitled to borrow the book (or recording) can access it at a time. Controlled digital lending is also distinct from the digital resale considered in Capitol Records, LLC v. ReDigi, Inc., 910 F.3d 649 (2d Cir. 2018). Contra Resp.Br. 35. The former’s purpose is nonprofit library lending, while the latter’s was commercial resale. Controlled digital lending is fair use, even if these other practices are not.
That is: the heart of IA argument is that its "controlled digital lending" practices were technologically limited in purpose and duration to be like libraries' other traditional legal reuses of owned works. (Typically, this meant maintaining the 1:1 physical-copy-to-leant-ebook ratio, but even under the temporary "National Emergency Library" program, it meant no permanent unrestricted copies were created – all rights-managed borrowings could and did expire when the crisis ended and normal book sources reopened.)
This author's manipulative clipping distorts the IA's filing into a strawman not matching the actual arguments advanced.
And: if Senate majorities really want to pass something, they can change the filibuster rules – and have, for some topics.
Otherwise, the filibuster is maintained out of tradition, courtesy, and its usefulness as a change-of-control 'debounce' mechanism – as well as providing a convenient excuse for posturing more and doing less, as Congress is wont.
Still, in other eras, Congress was able to move compromise legislation forward. Recently, Congress has been unable to – both parties, no matter the relative control. Any belief that it's only "the other guys" is partisan myside blindness.
But it's likely those on a motivated search for evidence of LLM limitations ('cope') are going to tout & over-rely on this result – and miss the extensive evidence in larger models, with slightly-more selective generations, that self-training often elicits improved performance on domains of interest.
For example, both the 'UnnInstruct' and 'Self Instruct' papers (of December 2022) showed that if – rather than self-training on arbitrary generations – you ask a model to generate good examples of a certain kind of prompt and helpfulness, then train on those generations, the resulting model tends to get better on many similar challenges.
It's almost like, as with human practice/exercises, there's a spillover effect on related competencies - eliciting from the model a potential that was too fuzzy to exploit at first, but gets honed by effortful practice (even without additional authoritative-instructor corrections).
To me, it's eerily similar to human self-help routines - "give yourself a pep talk, visualize desired results, pick tiny positive steps doable once & keep doing repeatedly, imagine success, affirm all progress".
Or, say, the "Inner Game of Tennis" style of gently reinforcing some key skill into subconscious comfort, with broader effects:
More history, through January 2022, via JWZ's blog:
https://www.jwz.org/blog/2022/01/the-reason-the-tuesday-noon...
Woulda been a nice time to clean up some of this technical debt!
Or how about the SF Emergency Sirens, taken offline in late 2019 for a "2 year" upgrade plan that officeholders implied was already in place?
In August 2023, with no progress whatsover, with the Maui fire disaster fresh on their minds, Mayor Breed & Supervisors President Peskin touted they'd finally funded a plan to return them to service soon: https://www.sf.gov/news/mayor-breed-and-board-president-pesk...
In that same August 2023 timeframe, Peskin said the plan would bring this "need to have" system "up and running" & to "state of the art" by end of 2024, for $5.5M: https://www.cbsnews.com/sanfrancisco/news/san-francisco-city...
Of course, this was just more blatant self-exonerating bullshit from our local political machines immune from any real accountability for incompetence in basic public functions.
A mere 6 months later in February 2024, nothing's been started, Peskin admitted "we don't even have a plan", the department is still waiting until "funding is identified", and the cost estimate has ballooned to $20.5m: https://abc7news.com/san-francisco-sirens-emergency-911-aler...
That works out to $170K+ for each of 119 units – units that each could probably just be a weatherized consumer-grade handheld device with multiple mobile/packet/sat radios, & a simple authenticated-playback app, mounted on existing poles that presumably already have power and even loudspeakers.
• a founders-duration copyright term – at most 28 years from date-of-creation
• absolute statutory & precedential immunity for any copies made, or DRM-circumventing-technologies created/used, solely for non-commercial preservation – that is, only competing witht the rightsholder in the commercial market would be prosecutable
• whatever extra protections/enforcement options available with formal copyright-registration also come with the duty to deposit/escrow full-quality unencumbered versions with multiple responsible preservation institutions, from the Library of Congress to NGOs, who are free to release those versions at end-of-copyright
Storage is cheaper than ever before. Accumulated shared wisdom about older format, & open tools for forward- and up- conversions, keep growing. Communities of enthusiasts can now find each other & coordinate effectively on arbitrarily niche and old media.
Hollywood's policy of monopolizing the right-to-copy works, enforced with aggressive legal prosecutions and fragile proprietary technical-complexities, puts cultural history at extra risk.