5 comments

  • Schlagbohrer16 minutes ago
    Amazing that they are trying to solve this with hardware rather than with a new software architecture but I suppose the current technology underlying LLM software must be far and away the best theoretically or most established, or the time taken to seek a new model isn&#x27;t worth it for the big companies.<p>I know Yann LeCun is trying to do a completely different architecture and I think that&#x27;s expected to take 2-3 years before showing commercial results, right? Is that why they&#x27;re finding it quicker to change the hardware?
    • AntiUSAbah4 minutes ago
      Nvidia has so much money, it would be a waste if they wouldn&#x27;t attack current problems on multiply points at once.<p>People, Researcher, Investor etc. probably also want to see what would be possible and someone has to do it.<p>I can also imagine, that an inferencing optimized system like this could split the context for different requests if it doesn&#x27;t need to use the full context.<p>Could also be that they have internal use cases which require this amount of context.
  • Schlagbohrer18 minutes ago
    What does this mean: &quot;In addition, because most AI models are not trained uniformly across their maximum context length, their reasoning quality tends to degrade gradually near the limit rather than fail abruptly.&quot;<p>Models aren&#x27;t trained across their context, their context is their short term memory at runtime, right? Nothing to do with training. They are trained on a static dataset.
    • andai8 minutes ago
      Not sure how it is now, but a while back most of the training data was short interactions.<p>I noticed that the longer a chat gets, the more unpredictable the models behavior becomes (and I think that&#x27;s still a common jailbreak technique too).<p>(I think it might also have something to do with RoPE, but that&#x27;s beyond me.)
  • __alexs37 minutes ago
    Does having 1 billion tokens mean more total tokens in the context window are actually good quality, or do we just get more dumb tokens?
    • RugnirViking32 minutes ago
      the article is almost entirely about this, yes.<p>Current approaches require fancy tricks to fit tokens into memory, and spread attention thinner over larger numbers of tokens. The new approach tries to find a way to keep everything in a single shared memory, and process the tokens in parallel using multiple GPUs
  • schnitzelstoat59 minutes ago
    Is such a large context window even desirable? It seems like it would consume an awful lot of tokens and, unless one was very careful to curate the context, could even result in worse performance.
    • AureliusMA17 minutes ago
      I remember when a large context was 8k! Nowadays that would seem extremely small, because we have new use-cases that require much larger context sizes. Maybe in the future, we will invent ways to use inference on very large contexts that we cannot even imagine today.
    • withinboredom53 minutes ago
      For larger codebases ... maybe it will cut down on &quot;let me create a random number wrapper for the 15th time&quot; type problems.
      • Weryj22 minutes ago
        You should already have skills which mention these utilities.<p>But maybe that’s enough tokens to feed an entire lifetime of user behaviour in for the digital twin dystopia?
        • withinboredom19 minutes ago
          &quot;type problems&quot; was doing the heavy lifting there, not literally &quot;this utility&quot;.
  • AureliusMA24 minutes ago
    How large would a 1 billion token kv even be ?!