Needless to say, Fujifilm is citing generative AI and other AI-related services as major drivers behind the growing demand for magnetic tape

LTO-10 tapes were released earlier this year, providing a native capacity of 40 terabytes.

you are viewing a single comment's thread
view the rest of the comments
[–] 26 points 1 week ago (23 children)

I don't get it, why bother archiving anything AI produces? You can just ask the next AI to reproduce it.

  • source
  • hideshow 23 child comments
  • [–] 37 points 1 week ago* (7 children)

    As some comments here have pointed out, it's probably more like data that goes into the training corpus.

    I'd also suggest that some demand won't be from AI companies, but rather from other consumers of data storage, even if a boom in AI-driven spending is at the root of things. Like, maybe your ideal storage medium is SSD, but AI-related demand has driven prices up to 5x what they were earlier. Okay, so some users get priced out and fall back to rotational drives, because those are good enough for their use case. Prices on those go up by 2x. Well, the next fallback for some of those users who get priced out, who maybe used hard drives to store backups, say, is gonna be magnetic tape storage.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 7 points 1 week ago (6 children)

    I buy the follow on effects rational, but I'm not convinced that tapes are a suitable alternative to HDDs. The random IO basically doesnt exist. But I guess if there are companies using HDDs for "cold" storage then using tape makes some sense for them.

    My personal theory for this announcement is that its pure stock price manipulation. AI investing is full on fomo driven, so announce that your company is investing in X that will help AI, and watch the chuds hit buy in robinhood.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 5 points 1 week ago (4 children)

    I’d love consumerish tape storage for cold backups. I don’t really need random IO for archival stuff.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 3 points 1 week ago (3 children)

    Yeah, same, but every time I look at second hand stuff its crazy expensive, or completely obsolete.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 2 points 1 week ago* (last edited 1 week ago) (2 children)

    And if something dies, good luck on getting replacements lol.

    I'd love tape but >5k for a drive is a bit outside my budget.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 1 week ago (1 child)

    Yeah, the longevity for tape backups feels entirely theoretical when the tape drives themselves are out of production, so you could end up with a perfect tape, and nothing can read it...

  • source
  • parent
  • hideshow 1 child comment
  • [–] 13 points 1 week ago (6 children)

    Archival of everything being scraped before its pushed off the internet and replaced with AI produced garbage.

  • source
  • parent
  • hideshow 6 child comments
  • [–] -1 points 1 week ago (5 children)

    The ship has sailed on the front. But even if its for the corpus of training data, surely that can't be stored on tape, the access time would be way too slow.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 9 points 1 week ago (4 children)

    It’s purpose is archival, not access. Tape density, cost, and reliability is much better than other mediums for that purpose.

  • source
  • parent
  • hideshow 4 child comments
  • [–] -1 points 1 week ago (3 children)

    Thats exactly my point, why bother archiving when you need it accessible for training? Surely they would be putting the training data on a faster medium.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 3 points 1 week ago (1 child)
  • [–] 0 points 1 week ago

    True in general, but I am not convinced its true for OpenAI or anthropic.

    Their training data is probably already replicated across multiple datacenters, so tape archiving is less critical. And if tape archiving is critical, then surely they already have sufficient backups already right? And its not like the organic content of the internet is growing anymore, so there isn't going to be an explosion of new organic data they need to store.

    So I don't buy that they need this. I think this is fujifilm trying to get in on the bubble, and nothing more.

  • source
  • parent
  • [–] 6 points 1 week ago (2 children)

    They don’t need to archive AI output silly, they need to archive the cumulative works of the internet because they keep forcing hosts to shut down with what are among the most aggresive web-crawlers ever deployed outside of DoS attacks.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 11 points 1 week ago (1 child)

    This feels like a brain dead comment.

    Why bother archiving anything when a computer can produce it?

    Because sometimes it's faster or cheaper to compute something and store the result then to recompute it every time. Nothing about AI changes that or is special in any way shape or form.

  • source
  • parent
  • hideshow 1 child comment