Think you’re a prompt engineer? Think again
A very interesting article in IEEE Spectrum, “AI prompt engineering is dead, long live AI prompt engineering”, offers some insight into the future of structuring requests so they can be properly interpreted and understood by generative algorithms, predicting that this will be a task they will carry out themselves.
The internet is awash with cheat sheets offering the best prompts, most of which describe a good prompt engineer as someone who understands algorithms, as if algorithms needed to be spoken to in a certain way. The reality is that algorithms don’t speak any language, they are simply mathematical equations that act on the basis of correlations, and the occasions when a certain prompt works well are more often the result of chance than of some kind of special ability.
Obviously, experience and fluency in handling generative algorithms is a factor, and a person who uses them frequently in their work will tend to generate better prompts than somebody with no experience, but the reality is that for all the mystique surrounding prompt engineering, there is little consistency. The evidence suggests that algorithms are best left to develop prompts themselves, and that they are capable of evaluating different alternatives competitively when formulating a request and presenting the best options based on purely mathematical parameters such as…
