Bots are currently scraping the internet for LLM training data at unprecedented rates[1][2][3], driving up costs and destabilizing public-facing websites. I want to talk about how this has been particularly difficult for wikis, and has gotten much worse in the last few months.
  • tal@lemmy.todayEnglish
    15·
    3 hours ago

    Detect whether the user is a human, but instead of blocking the request (which is going to be obvious to the scraper operator and will just cause the bot developers to go work on better human emulation until they get the data), poison the response. Just as blocking scrapers is hard for website operators, so is separating useful data from not-useful data for people building AI training corpuses.

    https://www.cloudflare.com/learning/ai/data-poisoning/

    Data poisoning involves injecting malicious information into training datasets to manipulate an AI model’s behavior, compromising its accuracy, reliability, and the overall integrity of machine learning results.