From Colin Nicholson, Stockport, Cheshire, UK
In her letter, Jacqueline Christmas misses the point that it was exactly such human-imposed guard rails that I was looking for in my correspondence with AI concerning nuclear weapons. Another insight into the nature of AI came when I asked ChatGPT if Eric was a suitable name for a pet caterpillar, and if it would be OK to take it for walks on a lead. The large language model (LLM) replied that Eric was a perfectly suitable name, but advised against taking my caterpillar for walks on a lead, as it might pull on the lead and hurt itself. In such ways, LLMs reveal their true nature (Letters, 23 May).
