.TYPICAL software program responds predictably to instructions. "Generative" artificial-intelligence (AI) models, such as that made use of by ChatGPT, are various: they react to demands filled in day-to-day language, and can easily produce unusual outcomes. On the face of it, composing reliable prompts for AI is actually much less complex than, for example, learning a shows foreign language. But as artificial intelligence designs have actually become much more competent, making the most of the algorithms within these dark boxes has become harder. "Cue engineering", as this ability is actually understood, has been actually likened to directing a dancing companion or stabbing a creature to view exactly how it is going to answer. What does it involve?