• macniel@feddit.org
    link
    fedilink
    arrow-up
    8
    ·
    13 hours ago

    Oh you haven’t noticed it yet? When it comes to improvement of AI or perceived value, its always that it will be better in 6 months. Just like fusion is only 2 years away. All the while the world is burning, hurrah for the shareholders!

      • Natanox@discuss.tchncs.de
        link
        fedilink
        English
        arrow-up
        2
        ·
        9 hours ago

        AI has not significantly improved in the last 6 months and won’t in the upcoming 6 months, it’s all marketing nonsense at this point. We’ve reached the mathematical limit of what transformer models based on language can achieve and there’s no new fundamental tech in the pipeline, all you see now are minor improvements in efficiency and implementation (harnesses and stuff). New domains it is used in doesn’t make it better overall, and it sure as hell doesn’t improve LLM’s ability of understanding structural design or suddenly add an ability of abstract adaptation (what would be necessady for “AGI”).

        Even the trust-me-bro benchmarks - if not faked - show them topping out at this point, not just scientific studies into the matter.

        The technology has reached its limit being financially pumped up without breaking even. To say it has improved massively in the last 6 months is just drinking the marketing Koolaid, only the implementations proceeded. The models behind it are stagnating.

        And for the love of god, stop with this “but China!!!” bullshit. The west literally moved all production capacity there and now blames China for the results. I dislike the CCP as well, but once taking production capacity, market of said products (i.e. western products get counted for western CO2 budgets) and population into account it looks completely different.