280 pointsby Cider99867 hours ago27 comments
  • ahmedfromtunis5 hours ago
    I live in a country where the selection of available books, especially in English, is very limited. Buying online from foreign markets comes with a long list of administrative hurdles and limits.

    If it were not for Anna's Archive and Z-Library, I would've never been able to read the books that shaped who I am today, or keep my passion for learning alive.

    Thanks, AA and ZLib! (Also, thank you to the authors whose books and knowledge I consumed without being able to pay them back.)

    • whycomean hour ago
      So you’re saying your entire current life is because of the proceeds of crime?!

      I’m kidding. Knowledge should be free. It was never created in a vacuum. It belongs to us all

    • jvm___4 hours ago
      https://send.djazz.se/

      This is key for getting epubs to your Kobo.

      • ahmedfromtunis4 hours ago
        Thanks, but I don't use e-readers as they are not available here.

        I've been using MoonReader for many years now and settled on pretty good parameters that make the reading experience very comfortable on both my phone and my tablet.

        • subscribed2 hours ago
          Moon reader is amazing. I love mine so much I don't see a point of having a separate book reader.
      • Salgat2 hours ago
        This is a genius way to farm ebooks while providing a useful service. I personally just use Google drive though.
        • whycomean hour ago
          Lol it never occurred to me that they might just save every single upload
      • pull_my_finger4 hours ago
        I don't understand what this is doing. Can't you sideload any ebook onto a kobo anyway? Never had an issue on my Clara
        • TFNA3 hours ago
          I’ve noticed that people today often bristle at any suggestion that one connect a device to a phone or computer with a cable – on Reddit, one will often get downvoted for this. Apparently, a lot of younger people are hardly aware this is possible and it strikes them as overly complicated or for old people. People want to wirelessly transfer stuff, and what the OP linked to is a popular way to do that with Kobo.
        • sureglymopan hour ago
          Yes. You can literally ssh into a kobo. I usually just put my books on a WebDAV share that is mounted on the kobo.
      • Almondsetatan hour ago
        Or... just a USB cable?
        • elrostelperienan hour ago
          I agree that a USB cable is the most practical option. However, the aforementioned site is useful in a specific scenario: if your Kobo is very old, macOS won't recognize it.
      • andrepd3 hours ago
        Handy, but a book lover with an ereader probably already uses Calibre :)
      • Brian_K_White3 hours ago
        I don't recall ever needing anything special on my Aura H2O. It's one of the reasons I chose Kobo in the first place. Just copy any file onto it.

        If you mean stripping drm I used Calibre for that but mostly I just avoid buying books with drm where possible.

    • pipes3 hours ago
      Look, fair enough from your perspective. But a lot of those books probably wouldn't exist if the author couldn't make some money from their work.

      I can't find the post but years ago on Reddit an author posted stats showing when her book turned up pirates online, real sales for it collapsed.

      Because of this I make a point of buying books, programming books especially. Yes I download pdfs, I use them as previews. This has led to buying way more than I would have.

      Anyway, I appreciate this doesn't apply if you live somewhere that these books can't be purchased. But everyone praising these sorts of sites tends to look at them from only a positive perspective.

      • mahdi7d122 minutes ago
        I live in Iran and the administrative hurdles the op was mentioned are not an issue here because you just can't even buy intwrnationally to begin with so there is no hurdle you might need to circumvent. The few English books I have are largely illegal reprints of a pirated version or some old ancient printed version that have somehow gotten imported (no clue how or is there an actual legal way)

        I remember opening "thinking fast and slow" and noticing the weird paging. After checking the official version's page count and seeing how the version in my hand doesn't match, my best guess was that someone had printed a pirated epub version.

        • mahdi7d119 minutes ago
          I rambled so much that I forgot to say what I laid all that introduction for.

          I'm not part of the market for these products. I don't have access to them nor even if I buy some imported (probably illegally and by single persons) or printed version, am I going to benefit them since I'm disconnected monetarily from yhe author.

          Me reading pirated versions of these books has no negative effects on the earnings of the authors.

      • bawolff3 hours ago
        > But a lot of those books probably wouldn't exist if the author couldn't make some money from their work.

        I think that's at least a bit debatable. People thought that about (normal) libraries back in the day, but it ended up having the opposite effect.

        Not to mention out of print books or academic books which is a big usage of sites like these, since lots of people prefer physical books and only reach for pdfs as a last resort.

        • dsizzle2 hours ago
          Libraries spend like $2B / year buying books https://www.imls.gov/sites/default/files/2021-08/fy19-pls-re..., which is like 10% of the total book market. So even if no one ever bought a book because they first encountered the book, author, or genre in the library that's already a signficant difference
        • j2kun2 hours ago
          I think I agree, the FAR bigger impact on my book's sales was Google search deciding not to surface it in search results. Presence on pirate websites had no effect, and eventually I switched to the PDF as "pay what you want."
        • brookst2 hours ago
          Can you imagine if we didn’t have libraries and someone tried to create them today? From publishers to right wingers, they would be painted as communist plots to destroy creativity.
          • toomuchtodo2 hours ago
            The Internet Archive tried, at great cost and peril, to defend its ability to lend books as an online library due to format shift (physical books get first sale doctrine, ebooks are licensed, you cannot own them), and were told no by the system, so “pirating” it is until copyright changes and becomes more reasonable. Disk is cheap, and the Internet global. Global distributed storage system durability and availability is the path to success until laws change imho.

            (Archiving culture alone is not the same as also enabling universal access to the culture and knowledge one is acting as custodian for and serving to global citizens)

            The Internet Archive has lost its appeal in Hachette vs. Internet Archive - https://news.ycombinator.com/item?id=41447758 - September 2024 (793 comments)

            https://archive.org/details/brewsterkahlelongnowfoundation

            Totally unrelated: Dweb camp 2026 is coming up for those interested: https://dwebcamp.org/

            (no affiliation with any person or entity mentioned in this comment)

      • 2 hours ago
        undefined
  • dr_dshiv4 hours ago
    https://SourceLibrary.org has about 16,000 rare books translated — most for the first time. 50,000 books archived (will be translated when we have $$ for it). More tokens than English Wikipedia and about .75 petabytes.

    Not sure if we will qualify for a bounty, but happy to share! Btw, we are looking for funding from small or large donors who want to help us translate the Renaissance…

    • wrsh074 hours ago
      Hey, this looks fascinating!

      I can't quickly tell what all you have archived^, but I have some friends who are academic historians who might be interested in certain categories of work (and could help verify some esoteric languages) - is it possible to search by region or language?

      Have you reached out to any types of historians WRT the project? It seems like some PhD students might be able to find some projects in this work etc

      ^ when I looked at the timeline https://sourcelibrary.org/timeline, I got an error

      • dr_dshiv3 hours ago
        Yes, this is designed with historians and librarians from the Embassy of the Free Mind (https://embassyofthefreemind.com) in Amsterdam, stewards of the collection of the Biblioteca Philosophica Hermetica

        Please share with historian friends. I’m not great at socials or fundraising but this was really designed to support humanists. It can give DOIs for the versions of the translated books, which means they can be quoted and cited in academic papers.

        Tip: Try it in Claude or Claude code (even better)! Just point it towards the source library. It can find quotes and evidence on any topic of interest. Or try the librarian — our source-grounded research agent https://sourcelibrary.org/librarian

        Thanks for the feedback, I’ll fix the timeline.

        • therealpygon2 hours ago
          Interesting site. I picked a random topic to listen to — flying chariots or something like that — and the conversation of one person talking and the other whispering was definitely not to my preference. I’ll have to take another look when I have more time.
    • sgc4 hours ago
      Curious as to what your budget was to get where you are today? That's a lot of tokens. I presume you are using gemini flash?
      • dr_dshiv3 hours ago
        All the models used are shown with each page of translation and each book has a whole data provenance treatment.

        You can add it up!

        • sgc3 hours ago
          I don't see raw token counts, just a list of steps and page counts. For example, what is the rough average token count per page in the ocr and in the translation steps for a Greek book?

          I have seen Gemini costs change quite a bit when processing very similar books from the same series lately, mainly because thinking tokens have increased about 5x. Has that has happened to you as well?

          Edit: for ocr I am using about 15k-25k tokens per page, but I have a complex prompt.

        • mmargenot3 hours ago
          How do you handle the more densely written pages in script ? I did a very similar exercise OCRing works from this exact collection, but I stuck with the English books for the first pass.
        • efilife2 hours ago
          Can't you just tell him?
    • ziofill2 hours ago
      Wow this is amazing!
  • tangenter2 hours ago
    Anna’s came clutch for me yesterday. I spent a few days trying to find a zip file of a CD that came with an old book from early 2000s on programming. One of those Thomson Publishing slap jobs that I actually enjoyed. I checked used copies all of them said does not come with CD. I tried googling around, nothing. LLMs couldn’t find it. ChatGPT kept saying it is on the archive (no it isn’t you useless piece of shit). Anyway, on a whim I went to AA, lo and behold, zip files for both first and second edition. Godsend.
  • trilogic5 hours ago
    Who is behind Annas archive, there is a lot of english speakers involved in the team and forums! Anyway as long as buying isn´t owning no issues here.
    • Cider99864 hours ago
    • bcyean hour ago
      I’d reckon many books available on there are otherwise available DRM-free, you’d be surprised really how many authors don’t bother with DRM.

      And then you could obviously just buy it physically where buying is definitely owning, so I find that sentence a bit inappropriate for books

    • shevy-java3 hours ago
      I think the main source may be in Russia; or that was with libgen.

      But I could be wrong.

      I am more surprised to see that there are so few alternatives to it. Or perhaps I am unaware of them but after Facebook and co declared war on libgen, and libgen going down, there were surprisingly few alternatives. Anna was one of the few. I still don't know what happened with libgen, but since the attack it really is kind of semi-gone.

      • trilogic2 hours ago
        Libgen and similar are more alive than ever with an extended botnet growing weekly. The "googlers" indexed framework is shrinking everyday, so users wont find it in those search engines easily, also it is hard to keep up with a good storage considering price trend last 5 years so the botnet and torrents are some kind of solution I guess. (We for instance are considering to use the old taping system, cause is at least a viable alternative.
    • tumdum_2 hours ago
      If no issue there, then why would you ask who is behind it in a public forum?
  • hedora5 hours ago
    I wonder how long it will be before they offer bounties for internet scrapes.

    Cloudflare captchas have made the internet unusable for me, and I'm sure it will only get worse over time. I'd much rather just browse (or even torrent) a copy of archive.is or similar. The latter would be much better for privacy, and hey, I run ad blockers anyway.

  • DeepYogurt5 hours ago
    Anyone afraid of being laid off at google right now? Perhaps this is a backup :)
    • Cthulhu_5 hours ago
      I think if you get caught exfiltrating data they'll sue you for much more than $200K.
      • imhoguy4 hours ago
        I don't think anybody would do it purely for money. I would rather see someone who is terminally ill and decides to do some "good".
        • dlenski4 hours ago
          There are not too many mentally-sharp, fully-employed, terminally-ill people that I have met. Even fewer at tech companies.

          And even fewer who are single and childless. (Google would likely go after the estate of anyone who did this.)

          • bitmasher93 hours ago
            I wonder how hard they would press an estate. It’s bad PR to go after widows and surviving children, and the data has already escaped.

            This is something they’d want to settle quietly, so the family would have leverage.

            • tumdum_2 hours ago
              They’ve made so many terrible decisions already. Going after widows wouldn’t change anything.
          • imhoguy3 hours ago
            But the one would be enough, especially in large organization. Surely they would need access to the exact data too.
      • merpkz4 hours ago
        Copy data into extra large capacity micro sdcard and hide it in your rubiks cube, nobody will suspect a thing
        • diab0lic4 hours ago
          It’s the “ Copy data into extra large capacity micro sdcard” step that gets you caught. Nobody is stopping you from leaving with an SD card or USB stick at Google.
        • takipsizad4 hours ago
          I wish an extra capacity SD card was enough, google books holds (probably) an insane numbers of books
          • stephenlf3 hours ago
            Comments on the source mention dataset sizes ranging between 1.5PB and 200PB
            • cydodon3 hours ago
              For 200PB one would need 25kg worth of 2TB microSD cards... that would be lots of Rubik's cubes =P
            • takipsizad3 hours ago
              my guess would be the 7PB mark
      • mmooss3 hours ago
        I'm sure they'd go after you, but hypothetically: What damages would they claim? They still have the data, which isn't their IP to begin with.
        • shevy-java3 hours ago
          Good point. But it would still be a breach of Google policy, most likely and they signed a pact with the devil so ...
      • the_real_cher5 hours ago
        If your money is in private crypto or offshore you have nothing to worry about.
        • zuzululu4 hours ago
          i'd strongly caution anybody foolish enough to go down this path

          financial watchdogs and international treaties make it impossible unless you are perhaps a multi billionaire who can afford to buy people at the political level

        • mock-possum5 hours ago
          Except perhaps jail time.

          Lying about your assets to avoid paying a lawful fine is criminal. Just because they can’t see your money doesn’t mean they can’t prove that you have it, and can’t jail you for hiding it to get out paying a fine.

    • shevy-java3 hours ago
      I think the problem is more that financial damage would result from this. So people would need to be prepared to relocate to another country probably.
  • alkyon2 hours ago
    Gemini should be trained on those books already, so in theory it could regurgitate some verbatim fragments (as NYT lawsuit against OpenAI showed some time ago).
  • bix66 hours ago
    Piracy / copyright predictions?

    The current situation feels untenable with renting. So many regular people I know have learned about VPN, NAS, etc.

    • codemog5 hours ago
      Hopefully the guillotines. Look up how much the authors and artists who create the actual work get paid.
      • 0x3f4 hours ago
        Quite a few textbook authors I know are paid well to be part of the whole scheme (kickbacks, forced yearly repurchase for the 'online' component of books, etc). So I think it varies a lot.
      • smashah2 hours ago
        All authors should have a pay + linktree type thing so pirates can pay them directly.

        Or something like thanks.dev

    • specproc5 hours ago
      It was never sustainable, just regulatory capture by large IP owners.

      Spotify, Netflix, Amazon etc provided OK value for a while, but now enshitification is biting, this is due a massive comeback.

  • wxw6 hours ago
    Some more interesting bounties they offer: https://software.annas-archive.gl/AnnaArchivist/annas-archiv...

    > Purchase all Library of Congress MARC datasets — $3,000 bounty

    > English Wikipedia pages about relevant institutions — up to $100 per new page

    > Internet Archive Digital Lending — $5000 per 1 million pdf files

    > Text version of our full library — $20,000

    ...

  • neilv5 hours ago
    The US should just find a way to quietly share literature access with the Russians, rather than letting piracy be promoted and facilitated for US consumers as freedom-fighter "archiving".

    Between all the piracy, and all the AI training and the purchase/visitor-circumventing AI services, the practice of writing and publishing genuinely good work is being wiped out.

    We're killing the goose that lays the eggs, for selfish gain.

    • TFNA5 hours ago
      This ship has sailed for academic publications, and academics define that term very liberally because we want to read everything, fiction included. The shadow libraries started off as a way for scholars in ex-Soviet countries in particular (but also India, SE Asia, etc.) to access literature that simply wasn’t available in their country. But the shadow libraries proved so successful and convenient that academics in all countries are using them now, even if they have access to official subscription services. I use AA several times a day and so do the researchers around me in my office; at conferences, if the presenter mentions an interesting publication, the whole room immediately opens AA on their laptops, etc.

      Even if projects like AA didn’t have nation-level support, academics would find a way to keep as much of it as possible going. After all, we’re the ones who compiled the bulk of pre-2020 material, and we’re the ones who do all the hard work of scanning from our institutional libraries stuff that doesn’t exist anywhere in digital form.

    • logicchains4 hours ago
      >the practice of writing and publishing genuinely good work is being wiped out.

      Most of the best literature in the English language was written before modern IP law was even a thing. There's very little good literature written by authors primarily motivated by money.

      • Jtarii3 hours ago
        How much of that literature was written by wealthy landowners who already had little need for money?
        • mr_toad2 hours ago
          Well, you needed the means to get an education, since most of the poor in those days were illiterate, which is something of an impediment to becoming a successful writer.

          I can only think of one writer off hand who wasn’t a wealthy landowner, although it is a particularly notable example; that of William Shakespeare.

          Shakespeare wasn’t poor (his parents seem to be of upper middle class standing), he was able to get a basic (but not a university) education and then pursue an acting career (with perhaps a side hustle as a teacher). Whatever the case he certainly wasn’t independently wealthy before he started writing, he needed to earn a living.

          He did seem to be in it for the money (and fame) since he wasn’t just a writer he was an actor, theatre owner, and something of a celebrity, and he did make enough money to become a wealthy landowner by the time he died.

      • boca_honey4 hours ago
        That's just cultural elitism. I hope you meet someone in your life who finds absolute joy in reading young adult romance novels or D&D fantasy books so you can understand how irrelevant "good" literature is. I love Dostoevsky and Verne (and D&D novels, especially those written by R.A. Salvatore), but I would never judge the modern "IPs" that got my daughter into reading.

        > best literature

        What does that even mean?

        • OriginalPenguin39 minutes ago
          Everyone has their own opinion as to what the best literature is, just like what the best music is.

          But there is also some consensus. For music it would be Beethoven, Mozart, the Beatles, etc.

          For literature it would include Tolstoy, Shakespeare, Cervantes, Dickens, Austen, Tolkien, and many more. I would bet it will eventually include Stephen King but it's too early to make that call now.

    • mjburgess5 hours ago
      Possibly but this act of governmental self-harm is useful to The People. We live in a world where if your valuation is ~1T you can more or less just do what you like. And the work of The People is stolen from you and launderd.

      In such a world, isnt it useful that governments are stupid enough to give adversaries reasons to undermine it? When the government props up a corporate tyranny domestically, and racketeering, should we make a temporary alliance with all its enemies?

      (Eg., the provision to AI companies of all corporate secretes and competitive practices via prompts, eventually to be used against their capital interests and their labour interests).

      • LearnYouALisp4 hours ago
        So when will the American people form an "Incorporation" to lobby against business for them?
    • WarmWash4 hours ago
      >We're killing the goose that lays the eggs, for selfish gain

      We already did that when the internet collectively agreed decades ago that everything digital should be free for anyone.

      We're now 20 years downstream of ad-blocking being a virtuous good, and piracy being the ultimate show of liberty, and now suddenly everyone cares about the creator's revenue stream.

      The mask slipped and unsurprisingly the internet is a bunch of selfish morally stunted children. Some of them even pushing 50 years old.

      Yes, I am talking to you with the 4TB of pirated content, proud of not loading any ads in the last 15 years, and getting enraged over LLM training.

      • lelanthran4 hours ago
        > Yes, I am talking to you with the 4TB of pirated content, proud of not loading any ads in the last 15 years, and getting enraged over LLM training.

        That's oddly-specific :-)

        In any case, I have no pirated content that I know off, neither proud nor ashamed of blocking ads[1], but I still get annoyed that a bunch of VCs can use their invested-into companies to launder all the worlds IP, then sell it back to them.

        [1] Who feels proud of blocking ads? It's like feeling proud of tying your shoelaces: "Good job, well done, but that's the expectation, son".

    • 5 hours ago
      undefined
  • FerritMans6 hours ago
    So AA is a front for openai?
    • flexagoon4 hours ago
      No, but they openly make a lot of money from selling their library to AI companies. Fast enterprise access to Anna's Archive starts at $100.000
      • poly2it2 hours ago
        A lot? I would be kind of interested if there were any known figures. Do companies want to be implicated in AA-cooperation in any capacity?
        • flexagoonan hour ago
          No specific figures, but see, for example:

          https://annas-archive.gl/blog/ai-copyright.html

          > Virtually all major companies building LLMs contacted us to train on our data. Most (but not all!) US-based companies reconsidered once they realized the illegal nature of our work. By contrast, Chinese firms have enthusiastically embraced our collection, apparently untroubled by its legality.

          > We have given high-speed access to about 30 companies. Most of them are LLM companies, and some are data brokers, who will resell our collection. Most are Chinese, though we’ve also worked with companies from the US, Europe, Russia, South Korea, and Japan. DeepSeek admitted that an earlier version was trained on part of our collection, though they’re tight-lipped about their latest model (probably also trained on our data though).

          It's at least 30 companies, each of which paid hundreds of thousands of dollars.

        • Cider99862 hours ago
          They likely use intermediary companies, but NVIDIA might have purchased from them directly, I don't remember the full story.
      • shevy-java3 hours ago
        Interesting. But AI companies drive the RAM prices, which costs me more. So someone makes me pay more here ... :(
    • 650REDHAIR5 hours ago
      How did you come to that conclusion?
    • awakeasleep5 hours ago
      the bounty would be a bit higher with openAI money behind it
  • hereme8884 hours ago
    The link sort of reads like people who have very easy access to the requested material. Almost like they're Google employees.
  • stephenlf3 hours ago
    Anna’s archive rocks
  • 5 hours ago
    undefined
  • thenthenthen3 hours ago
    There was a time where you would get a random page preview, some artists found a way to extract full books that way (F.A.T lab?).
  • stephenlf3 hours ago
    The only legal hurdle keeping Anna’s Archive away from its noble goal (piracy laws) has been shown to mean zilch in the age of AI.
  • cubefoxan hour ago
    HN logic:

    Training on copyrighted material

    --> bad

    Actually distributing copyrighted material

    --> good

    Needless to say, this is backwards. Any copyright holder will be much more worried about the latter.

    • sublimefire12 minutes ago
      The problem is that it is quite difficult to access the published papers is you are not in academia or some company that pays for the access, so AA sort of serves that niche to transfer the knowledge. Training on the other hand is a commercial activity to later rent the model, if this would be purely for open weights I suspect everyone was cool with it.
    • johndoughan hour ago
      This could be another fine example of the Goomba fallacy: https://en.wiktionary.org/wiki/Goomba_fallacy

      Another explanation might be a general dislike of big establishments like AI companies and publishers (which glosses over individual authors, but they probably make up a negligible portion of total sales anyway).

    • Cider9986an hour ago
      At least Anna's archive is consistent.

      Copyright reform is necessary for national security

      https://annas-archive.gl/blog/ai-copyright.html

  • mmooss3 hours ago
    How is Anna's Archive funded? I see they have memberships, but it's hard to believe that can fund all these bounties - some going into six figures. Ask any FOSS project about funding by that method.

    It seems like there are some deep pockets funding them.

    • atemerev2 hours ago
      Chinese (and some other) AI companies buying fast access to their dataset.
      • mmooss2 hours ago
        So Anna's Archive is in some ways a front for AI companies, gathering the sources they can't get themselves?
        • Cider9986an hour ago
          They could get access for free through torrents.
  • anyaya14 hours ago
    Does Anna's Archive use a completely different "source repository" from LibGen?
    • TFNA3 hours ago
      AA compiles from everywhere; LibGen and Z-Lib served as the major sources of books. This has unfortunately led to search results for a particular book containing multiple versions of that book, and it is not readily clear which one is the highest-quality version. A real library would have librarian staff who carefully curate everything, but in the pirate world this isn’t realistic so it just gets all thrown together.

      LibGen is now more or less a dead project. The servers of the original version were reportedly seized a couple of years ago already, and other sites under the LibGen name were notorious for piggybacking the original collection and just plastering it with ads. If one wants to upload stuff, better now to upload it to Z-Lib (not a perfect site, but still) and it will then get picked up by AA in a few months.

    • takipsizad4 hours ago
      annas archive is practically a compilation from all sources possible (including libgen afaik)
  • rvbaan hour ago
    Comment by Borja is a great example of eternal September.
  • leoc3 hours ago
    Just do it and be legends, Larry. ;)
    • Cider99863 hours ago
      Apple won't even help Asahi linux even though it would help hardware sales and give them a ton of goodwill.
      • leoc27 minutes ago
        Oh, I have no expectation that Page or Brin would do something like this, let alone do it openly. But Page did seem to care about Books access at one point, and I find the image of him secretly leaking the Google Books corpus amusing.
  • TZubiri3 hours ago
    I think this would cross the line from civil copyright claims into criminal activity

    https://chatgpt.com/share/6a4970e8-7fe8-83e9-8f81-3aefd76b6b...

    On another note, if Google's cybersecurity were always one rogue employee away from a massive leak, then it wouldn't be Google. What was the last Google leak you remember, defense in depth people.

  • ThrowawayTestr6 hours ago
    One of my hopes is that when the AI bubble bursts, some brave person will sneak out a copy of the last frontier model.
    • Aboutplants6 hours ago
      Not worried about that, you will only have to wait 3-6 months and get a Chinese model just as good.
      • sulam5 hours ago
        That’s misunderstanding why these models are behind. A large part of why they’re behind is they aren’t able to do the reinforcement learning post-training steps that takes a pre-trained model and turns it into a frontier model like GPT 5 or Opus. Instead they do their best to recreate these models using distillation.

        Fundamentally, you can never distill your way to being the teacher, so these approaches will not advance the frontier.

        [edit, after thinking about it I think my phrasing is unfair. It's not necessarily that aren't able to do it, but they haven't yet shown that they are willing to do it.]

        • computerex4 hours ago
          That’s not remotely true. They did distillation as a cheap solution to the cold start problem. You need data/trajectories to hill climb to higher capabilities. All large Chinese labs do RLAIF.
          • sulam4 hours ago
            Oh yes, not remotely true. Which is why the frontier labs all have invested heavily in trying to identify and thwart distillers, using known company names / domains to drive their exclusion lists.

            /s

            • logicchains4 hours ago
              It's cheaper to distill than to do reinforcement learning, so of course they prefer that, but if it wasn't an option they could just pay up and spend more GPU time on RL.
        • DANmode2 hours ago
          > you can never distill your way to being the teacher

          Are you sure?

          What if you distill from 10 teachers?

          • poly2it2 hours ago
            In this case all teachers have also learned from each other.
        • FpUser4 hours ago
          >"they aren’t able to do the reinforcement learning post-training steps"

          Not yet.

          If there is a need someone will come and fulfill. Personally for me now I do not even want to use top models. Professionally I use AI to help with the coding using Junie agent that comes with IDEs from JetBrains. Junie is told to use Gemini Flash and works fine for what I ("I" being an emphasis here) ask it to do. I tried more advanced models and different vendors only to discover credits going down the toilet without any extra benefit.

          • sulam4 hours ago
            I'll agree I guess and clarify that the better phrasing is probably something like "haven't yet shown the capability to."
      • yorwba6 hours ago
        Chinese companies giving away expensive models for free is a symptom of the AI bubble, too. It's not a law of nature that they'll always be able to scrounge up the money for yet another training run.
        • gpm5 hours ago
          Shaping the tool that does the thinking is quite valuable when you're in the business of changing how people think - I think we can expect propaganda agencies to be subsidizing model creation forever.

          This doesn't strike me as a symptom of a bubble - except in so far as the bubble pushes the competitors models forwards and thus they need to invest more to stay competitive.

          • rvnx5 hours ago
            All the models, have to respect their local laws, and most of all, pressure from users and the employees.

            They all carry political weights, because humans behind defend their interests, and are promoting some social values.

            https://pastebin.com/hjhvsBFg

            This answer from Claude is so biased that it is ridiculous

        • jnwatson4 hours ago
          As long as it is in the CCP's national interest to have a frontier model, Chinese companies will have the resources for another training run.
        • nextos5 hours ago
          I think it's a deliberate business strategy of commoditization of their complement.

          China acts like an entire bloc, not as single companies, and they want to monetize hardware.

          • yorwba3 hours ago
            If you think Chinese companies always act as a bloc, your mental model needs to get about a billion times more detailed. But in this case just a few details may be enough: There are Chinese AI companies that have released LLMs without publishing the weights.

            ByteDance is going the direct-to-consumer route with their Doubao chatbot (the most popular in China, probably thanks to their social media prowess). iFlyTek seems to be angling for enterprise and government use cases, where they already have an in.

            The companies that have released weights have in common that they didn't have a monetization channel lined up and their models weren't good enough to make people pay attention with just API access. (You can see with Qwen Max that the calculus can change towards not releasing weights for better models.)

            And who exactly among the investors is having their complement commoditized? When Nvidia releases Nemotron, the story is clear, but it's less obvious for say Z.ai's GLM.

            • nextos2 hours ago
              I have never said they always act as a bloc, but their industry has a strong component of long-term strategic government planning behind them.
    • fastball5 hours ago
      If it's a bubble, why do you care about frontier models?
      • emdash2 hours ago
        If we had the dotcom bubble, why are you still on the Internet?
        • fastballan hour ago
          I never wanted the IP from dotcom bubble companies.
      • FpUser4 hours ago
        Internet was a bubble, so was telecom etc. at some point. Being bubble does not mean that when 90% of investments go down the drain the remains are not useful.
        • fastballan hour ago
          "The Internet" was not a bubble. Companies with no long-term business model / sufficient product-market fit that were riding hype were the "dotcom bubble". But when those companies crashed, nobody said "I really want to get my hands on their IP", because it wasn't valuable – an important pre-requisite to the the bubble popping. Seems to be a different case here if people actually want the SOTA models.
        • mr_toadan hour ago
          Railways were both a bubble and, eventually, one of the most significant technological innovations in history.
    • zuzululu4 hours ago
      which will be very difficult to run unless you have a large budget to operate your own mini datacenter
      • lelanthran4 hours ago
        In a crash the hardware will go for pennies on the dollar, if not for fractions of pennies on the dollar.

        Lots of companies will pick them up for scrap metal prices and host them for fractions of what we are paying today.

        That's the nature of bubbles.

    • thx675 hours ago
      Prediction markets can solve this.
  • 5 hours ago
    undefined
  • b1126 hours ago
    [dead]
  • OrangeDelonge5 hours ago
    Curious as to how you would approach this. I have no experience in this area, anyone on this forum willing to share their expertise?
    • 0x3f4 hours ago
      If it works as AA seems to theorize, you'd need to:

        (a) work out how Google books exposes fragments of books, and see if there's a systematic way of using this to get whole books.  For example, a naive approach might be to find any fragment of the book by searching some exact phrase.  Then, you can search for an exact phrase from the start or end of the fragment it gave you, hoping it will show you the previous or next part of the book.  You can then just loop that to get the whole book.
      
        (b) once you have (a), you need a way of bypassing Google's bot detection/rate limiting.  I don't know what current state of the art is, but there may be a solution for sale out there.  E.g. you pay to receive a cookie or browser state, and use that to fetch the URLs from (a).  Or if you're good/already in the scene, you could do this part yourself.
      • mr_toadan hour ago
        (c) avoid being hounded to death by a zealous district attorney.
      • takipsizad4 hours ago
        That way definitely will work with the current access google provides however its an extremely inconvenient way to scrape google books
  • tolerance2 hours ago

        If you shouldn't be able to copyright GRAPES...you shouldn't be able to copyright BOOKS.
    • ajcpan hour ago
      And why's that? I don't see how either of those two things relate to one another. A grape is a naturally occurring fruit, while the creation of a book is wholly a human creative endeavor.
      • tolerancean hour ago
        When I buy grapes, yeah, I get that someone worked for me to get 'em. But I don't get all metaphysical and spooky about it. I'm buying the grapes as objects. Like a book. Once I take ownership of the object, yes, thanks for your work. I paid you and I wish you well.

        As far as what I ought to be able to do with the thing? Beat it. Of course there are exceptions. You say, "Don't scan the book and upload it to object storage and place it behind a CDN so Ahmed from Tounis can access it". I say, "Don't make wine".