What do you call it when an economic bubble stops growing? In February, stock analysts TD Cowen spotted that Microsoft had cancelled leases for new data centres — 200 megawatts in the US, and one g…
  • 𝔗𝔢𝔯 𝔐𝔞𝔵𝔦𝔪𝔞@jlai.luEnglish
    1·
    10 days ago

    The future for the (very few) legitimate uses of AI is small local models, like we’ve been using for a decade without the buzz.

    Those huge models are a nice novelty, and that’s about it.

    • ☂️-@lemmy.ml
      0·
      10 days ago

      there are good applications in medicine and research for larger models.

      also local, running on their own servers.

      • ℍ𝕂-𝟞𝟝@sopuli.xyzEnglish
        1·
        10 days ago

        Large-scale AI or specifically LLMs?

        All that has been around for a long time now, there is no “medical AI boom” that would explain the need for new DCs.

        • ☂️-@lemmy.ml
          1·
          10 days ago

          LLM stands for large language models

          its what i imagine the op i was responding to means when he said “huge models”, and what i meant with my response.

          “AI” is indeed a loaded term.

        • snooggums@lemmy.worldEnglish
          0·
          10 days ago

          Pattern recognition for assisting with diagnosis, treated similarly to lab work and other techniques that still require a human to verify. Also for developing new medicines and treatments, all with human confirmation but as a tool to help identify what to focus on.

  • bipedalsheep@programming.dev
    0·
    10 days ago

    We’ve all had our fun now with watching companies and investors throwing billions on these LLMs but after all that we have something which does not solve billion dollar problems… They’ve already scrubbed all the data out there and let’s be frank here, the LLMs still suck. It feels like chatting with these LLMs is like overcoming an obstacle, I end up doing web searches myself most of the time anyway. These LLMs can’t think or reason, let’s stop trying to fake it and start using these models for something useful. Medical is the obvious one. Surveillance and military will probably be where the shift will be to primarily. There will be interesting things with pattern recognition for sound and images, but that’s about it.

    • Roopappy@lemmy.world
      1·
      10 days ago

      I’m about to resign from a company where the CEO is still saying “We need to utilize AI more to achieve our goals”.

      We don’t. We need to figure out how to do basic budgeting and accounting.