Over just a few months, ChatGPT went from correctly answering a simple math problem 98% of the time to just 2%, study finds. Researchers found wild fluctuations—called drift—in the technology’s abi…::ChatGPT went from answering a simple math correctly 98% of the time to just 2%, over the course of a few months.

  • miaow@lemmy.ca
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    2
    ·
    1 year ago

    The “find around and fuck up” a.k.a “wade deeply through digital exhaust (cough POOP cough) to find answers to life, the universe, and everything” model:

    ...  : [Garbage IN] : [Garbage PROCESSING] :  (*) [ Garbage OUT] (@) <----> :  [Garbage IN] ... 
    
    (@) you think you are here
    (*)  you are here (lol but only probably)
    

    Not to diss the tech/math behind AI, but it could be that OpenAI is trying to tame/sanitize it down to the extent that it is now handicapped enough to fail its intended purpose in the name of “safety” and “security.” Tee hee. It is somewhat schadenfreude to see it collapse under its own weight, but oddly entertaining to see this go down.

    I’mma bring out the popcorn when when it starts fighting (copies of?) itself though, like AI generated ads fighting AI ad-blockers. On your phones. What an utter waste of energy and resources that will be, especially with FAANG+M trying desperately to lock people out of that very AI hardware on “AI-enabled” phones that they splurged on… the mind boggles.

    PSA/Reminder: OpenAI is just as “open” as the (Microsoft) windows letting sunlight through to where the sun don’t shine.