Three years of ChatGPT – from chattering parrot to sparring partner

Critics call ChatGPT a chattering parrot that understands nothing. Yet something remarkable has happened in three years: unexpected abilities have emerged from the mere prediction of words, such as arithmetic, reasoning, and programming. Far from perfect, but often surprisingly useful in practice, especially with human correction and supplementation.

Sometimes these systems produce convincing-sounding nonsense. But that doesn’t make them worthless, rather creative. Not creativity with a capital C – no Kafka or Einstein – but with a small c: surprising, playful, associative. This makes them excellent sources of inspiration and sparring partners.

ChatGPT and other language models are getting better and better at articulating their reasoning step by step and consulting external sources of knowledge. They still make mistakes, but it is often quicker to improve an AI answer than to search endlessly yourself.
Generative AI is not infallible, but neither are humans. Psychologists distinguish between some 200 forms of cognitive bias, ranging from groupthink to the tendency to weigh negative information more heavily than positive information.

The key to successful AI applications lies in hybrid intelligence: a collaboration in which human and artificial intelligence complement and reinforce each other. AI brings speed, precision, and scale; humans bring meaning, context, insight, humanity, and a moral compass. Be pragmatic: use AI for what it does well, be critical where it falls short, maintain your own skills, and learn to work with it.