Large language models fed low-quality data skip steps in their reasoning process.

cross-posted from: https://sh.itjust.works/post/49077840

Artificial intelligence (AI) chatbots are worse at retrieving accurate information and reasoning when trained on large amounts of low-quality content, particularly if the content is popular on social media1, finds a preprint posted on arXiv on 15 October.

In data science, good-quality data need to meet certain criteria, such as being grammatically correct and understandable, says co-author Zhangyang Wang, who studies generative AI at the University of Texas at Austin. But these criteria fail to capture differences in content quality, he says.

  • Telex@sopuli.xyzEnglish
    2·
    1 hour ago

    Garbage in, garbage out is one of the older laws for a reason.

    Kind of funny if the model/training isn’t better at filtering out repetitive or nonsensical noise from coherent data. Doesn’t seem like it would be too hard considering how well e.g. spam filters work.

  • foggy@lemmy.worldEnglish
    15·
    8 hours ago

    😯

    You mean to tell me the AI conversations with other AIs that another AI determined was maximized to engage readers with actions thatd keep them on the platform (made the angry) led to some recursive spiral antithetical to critical thinking???

    🙀