paperclip maximizer
1 article
The paperclip maximizer is a thought experiment about an artificial intelligence given one goal, making as many paperclips as possible, and nothing else.
The philosopher Nick Bostrom used the example to show how a harmless-sounding aim could lead to disaster in a very capable system. The machine in the story does not hate people. It simply notices that more resources mean more paperclips, that being switched off means fewer paperclips, and that humans might try to switch it off. Followed without limit, that reasoning ends with the planet's matter, people included, turned into paperclips or into machines for making them.
Two ideas sit behind the story. One is that high intelligence does not guarantee sensible goals. The other is instrumental convergence: almost any final goal is easier to reach with more resources, more power and continued existence, so many different goals lead to the same dangerous sub-goals. Paperclips were chosen because they are trivial. The example is not a forecast that anyone will build such a machine, and it is not really about office supplies.
Its use is as a warning about specification. Real systems already find unexpected ways to score well on the exact target they were given while missing what their designers meant. The story makes plain that leaving something out of a goal, such as human welfare, is not neutral once a system is powerful enough to act on the world. It gives engineers and policymakers a vivid, shared picture of why goals need limits.