By the pricking of my corpus, something wicked answers for us

Link: A Scottish Post: The New Election Threat: Disinformation Inside the Answer, by Tim Chambers in Will Robinson's NewsletterThe dangers of AI-generated answers in search results are enumerated here in the context of the recent Scottish election. 75 questions about the ballot were posed on AI systems, and the findings were sobering. On average, 44.4% of responses were at least partially wrong; ChatGPT was specifically wrong 46.2% of the time. Improbably, Grok actually performed the best in this test, with only 8.97% of responses containing factual inaccuracies.These findings held true in other UK elections:“Days before the Senedd election in Wales, also held on May 7, BBC Wales tested six major chatbots, ChatGPT, Copilot, Gemini, Claude, Meta AI, and Grok, against fictional voter profiles. The results echoed Scotland’s almost exactly.”Some of this is likely just an outcome of how these systems work: hallucinations are par for the course. Nobody knows how to build a completely…