It’s hard to credit that a computer program which confuses fruit-flavoured cocktails with soup is about to end life as we know it.
But how do we know that it hasn’t already?
ChatGPT, a chatbot that mimics human responses when asked typed questions, has lately gained much notoriety for its uncannily fluent replies to users’ queries.
It is part of the much-vaunted research effort to create so-called artificial intelligence, or AI.
This term has long been abused as a cheap marketing label by every company with even the slightest connection to technology. Most claims of AI are nonsense.
Even the most advanced data manipulation up till now has been a form of machine learning: spinning ever-larger quantities of information through massively powerful processors to recognise patterns and make predictions.
It seems likely that ChatGPT, which is developed by a company called OpenAI, is employing a particularly advanced form of machine learning, partly by drawing on the vast corpus of text now available online to ‘learn’ English and other languages.
But some of its latest versions have recently been giving beta users sleepless nights, thanks to conversations so weird that its human interlocutors have begun to spy a ghost in the machine.
