4 comments

  • jdonaldson 5 minutes ago
    I think neuro-symbolic AI has a lot of potential here, since small models can handle a lot of conversational inputs, while relying on wired-in solvers for more complex symbolic math/computation needs. https://jjd.io/posts/swollm-bbh-leaderboard.html
  • tim-fan 1 hour ago
    Is anyone making LLM-in-a-box for emergency supply kits yet?

    I feel that would be handy in all sorts of situations when networks are down.

    • vessenes 0 minutes ago
      I've been mulling over a good use of a large philanthropy spend in the next decade, and I would love to build a bunch of hardware "oracles" that include an LLM. Ideally solid state, visual/audio, solar + usb-c, so, good in a lot of doomsday scenarios as well as just out hiking. It's a fun thought experiment. I imagine making like 1 million of them, they could be sold and genuinely useful, but also stored and put in an emergency box.
    • Terr_ 22 minutes ago
      > emergency supply kits

      I imagine that for most "emergency" scenarios, a device that stores large amounts of normal reference material will be dramatically cheaper, longer-lasting, and able to run usefully on limited power supplies.

      If someone will need something like a medical flowchart for first-aid, it's probably better to generate that kind of artifact in advance and then get it verified by an expert.

      • skybrian 13 minutes ago
        You will probably want a search engine though. Perhaps a small LLM would work well as a component for that?
    • SwellJoe 35 minutes ago
      This is couched in prepper nonsense, but it's got LLM, WikiPedia, maps, etc. A bunch of genuinely useful stuff to keep on a USB stick or whatever: https://www.projectnomad.us/

      But, the current model you really want for an emergency kit is Gemma 4 12B QAT 4-bit. At ~7GB on disk, it's small enough to run on a tablet or any modern computer, slowly if you don't have a GPU or modern Apple silicon, but exceedingly smart for its size, excellent vision capabilities, good tool user, surprisingly good reasoning.

    • cdnsteve 56 minutes ago
      Can you expand what you mean?
      • wahnfrieden 46 minutes ago
        They want to ask the iOS Foundation model (frontier on device intelligence for something small) for instance about emergency procedures and life-saving info. I wouldn’t trust that model with much at all though. More likely to find what you need from miniature survival guides.
  • bombcar 1 hour ago
    99% of the model "work" (meaning the connection to your computer) is just spinning a spinner - something that makes me want to wrap it with a mosh shell so I can just keep moving from network to network.
  • enoint 1 hour ago
    Fascinating to wonder whether the bigger model finds fewer or more counterfeits than the on-device one.