Opening page…
Knowledge
Who gets to call it solved?
An AI completes a coding task and passes the benchmark, but the software's maintainer says the result is unusable. Which approach should decide whether it is released?
Passing a test and satisfying the people who rely on the result can come apart.
This question is retired.
Nearby questions5
- You glance at a clock that stopped yesterday. By luck, it shows the correct time now. You believe it—and you are right. Do you know the time?
- How can you search for something if you don't already know what it is — and if you do know, what is left to find?
- You can sound exactly like you understand something without understanding it — and now the sounding is free while the understanding stays expensive. If no one can tell them apart, which one survives?
- When should you trust an answer you cannot check yourself — from an expert, a book or a machine?
- What could convince you that you are awake rather than dreaming? Could that same evidence appear inside a dream?