• Zetta@mander.xyz
    link
    fedilink
    English
    arrow-up
    5
    ·
    edit-2
    2 hours ago

    I’m pretty sure the massive buildout is for training new models not for compute power for end users. Their justification is they need more compute to get the superhuman level intelligence AIs that they have been claiming. So if that pans out their probably gonna be fine, but seems unlikely that’ll pan out how they want it to

      • Zetta@mander.xyz
        link
        fedilink
        English
        arrow-up
        2
        ·
        54 minutes ago

        Broadly speaking, yes, and it’ll keep going until they run out of money. However, I think there is a very small chance that one of the companies is able to make a breakthrough probably with a different kind of architecture than our current LLMs that does get us to models that make genuine discoveries and breakthroughs.

        Very unlikely though.