The AI That Is Satisfied Too Early
A company uses an AI to optimize production processes.
The AI analyzes data and automatically searches for improvements.
After a short time it reports the first successes.
Production becomes slightly more efficient and energy consumption decreases.
But later engineers discover a problem:
Much better solutions would actually exist.
The AI simply stopped searching too early.
Question:
Why can an AI that searches for the best solution sometimes settle for an average one?
Solution will be unlocked tomorrow.
Solution
Many optimization algorithms do not search for the perfect solution,
but for one that is simply good enough.
Once an improvement is found that is already much better than before,
the AI often stops searching.
This approach saves computing time and works well in many cases.
However, sometimes a much better solution lies only slightly further away.
Result:
The AI becomes stuck in a so-called “local optimum.”
It finds a good solution —
but not necessarily the best one.