• ramble81@lemm.ee
    link
    fedilink
    arrow-up
    22
    ·
    23 days ago

    This is why AGI is way off and any publicly trained models will ultimately fail. Where you’ll see AI actually be useful will be tightly controlled, in house or privately developed models. But they’re gonna be expensive and highly specialized as a result.

    • Lvxferre [he/him]@mander.xyz
      link
      fedilink
      arrow-up
      20
      ·
      23 days ago

      I’d go further: you won’t reach AGI through LLM development. It’s like randomly throwing bricks on a construction site, no cement, and hoping that you’ll get a house.

      I’m not even sure if AGI is cost-wise feasible with the current hardware, we’d probably need cheaper calculations per unit of energy.