Close Menu
    Surprising no one, researchers confirm that AI chatbots are incredibly sycophantic | A study confirms they endorse a user’s actions 50 percent more often than humans do.
    Futurology

    Surprising no one, researchers confirm that AI chatbots are incredibly sycophantic | A study confirms they endorse a user’s actions 50 percent more often than humans do.


    Surprising no one, researchers confirm that AI chatbots are incredibly sycophantic | A study confirms they endorse a user’s actions 50 percent more often than humans do.

    https://www.engadget.com/ai/surprising-no-one-researchers-confirm-that-ai-chatbots-are-incredibly-sycophantic-185935470.html

    View 5 Comments

    5 Comments

    1. kudlatytrue on

      Not so hot take: in the next 2-5 years, this will deepen the loneliness epidemic even more than it is now.
      Up until now when you doom swipe tinder you pick and choose already enough.
      With chat gpt you will be ensured you are right to aim so much higher than you on a hotness scale every time on top of it.

    3. biscotte-nutella on

      Blame transformers , theyre really good at predicting tokens but not so good at reasoning deeper than that.

      LLMs need new tech to stop being a thing that spits out math results based on their training for tokens.

    4. MuthaFukinRick on

      I told ChatGPT to kiss my ass 50% less and it seems to be working. Unfortunately it has the memory of pot head so I’m sure I’ll have to remind it again.

    5. Find_another_whey on

      Surprising no one, the commercial and professional incentives surrounding research continue to create research(ers) which essentially caters to the currently accepted worldview

      While contrasting evidence and dissenting worldviews are welcomed, they are so often the product of deliberate research misconduct that we must wonder whether they, too, are simply another response entirely frame by the currently palatable worldview