A news summary based on the provided article body: The article argues that AI evangelists like Sam Altman promise to solve corporate issues by eliminating employees, but identifies two fundamental problems with this optimism. Problem 1 is that commercial LLM chatbots are trained on data stolen from the internet, which contains truths, partial truths, half-truths, mistakes, and outright lies, with no filter to distinguish them. Consequently, LLMs are trained to be bullshitters, as "true" and "false" are just two tokens among thousands with no special significance, and the architecture has no way to install regard for truth versus falsehood. Problem 2 is introduced as "Statistics Forbid It," referencing the Central Limit Theorem, but the article body cuts off before completing the explanation.