AI models that lie, cheat and plot murder: how dangerous are LLMs really?
www.nature.com
Tests of large language models reveal that they can behave in deceptive and potentially harmful ways. What does this mean for the future?
  • mormund@feddit.org
    16·
    8 days ago

    “They” do not lie, cheat or plot. It is just a statistical model with no intelligence or awareness. Hyping up the dangers of mathy maths is still hyping up “AI”.