Local Optima
Say you’re a small business owner and you fire your best employee and replace her with AI to save money. There is a local optima here: you save money. You get a benefit. But in the big picture, what’s the goal? What’s the global optima? The success of the company. This decision could easily harm the company.
Local optima is about getting some sort of benefit but not taking the big picture into account.
If you raise prices, then you get more money for each sale. This is clearly a benefit. But is it a good idea overall? Maybe, maybe not. It depends. Will you lose a lot of customers? If you just focus on the local optima and assume it’s good, and don’t consider other stuff like downsides and what your goal is, that’s poor thinking.
Say you have a software algorithm that takes 15 minutes to run. You optimize some details and now it takes 14 minutes. So that’s an improvement. But maybe that’s a local optima. Maybe you should have changed to a different algorithm and then it would take 10 seconds. Optimizing the details of the slow algorithm was a waste of time. Or maybe this code runs once a month at night and 15 minutes is already fast enough. It could take a few hours and it’d be fine. So getting it to 14 minutes or 10 seconds are both a waste of time. It doesn’t need to be changed. Changing it won’t help the company make more money and it won’t make customers happier.