Global and Local Optima About problems and solutions

There are often solutions to problems. Sometimes the solutions are obvious, for simple problems. But in our world problems have become more complicated. The solutions are therefore not straight forward. Or maybe they are simple, but implementing them might is too difficult to achieve. And sometimes interventions are presented as solutions but they are mainly there to maintain the status quo (or worse). But I want to start from good intention here.

Developers sometimes write algorithms to find solutions. The ideal is to achieve a global optimum. A well-known example is route calculation. When you try to calculate a route from A to B, you want to get the shortest route. Calculating all routes is often impossible. So tricks are used. In the search for a solution you will come across optima, in this example routes that do just a little better than others. What you cannot always know is whether that outcome is the best possible (a global optimum) or a local optimum (‘this seems to be the best we have been able to find’). I often look at new developments (in society) with these glasses.

AI : given that it works - and I think the tool itself can provide useful things for someone who wants to get something done quickly that many others have already done; a lot of our work is simply not original. AI models are constantly being optimised, but are we working towards a local optimum, or is the route towards a higher global optimum reached in a totally different direction? Wouldn’t there be a solution to efficiently present information that consumes less energy, is more predictable and reliable, and not crashing when the complexity gets too high? Can this last problem ever be solved?

5% NATO demand: Suppose that with a stronger army unit, which will costs us 5% of our income every year, we indeed are able to deter the aggression of actors such as Russia, is that an optimal effort of the resources? Or is it wiser, from a win of safety per €, to weaken the enemy in a different way (e.g. European dismantling fossil and nuclear dependence)? And should we be increasing our spending when Russia and China start working together to crash giant swarms of autonomous drones on our European cities? Or have we optimised for the wrong thing?

Sometimes there is nothing wrong with accepting a local optimum. Whether you are 20 minutes cycling or 28 minutes, it is about half an hour and it is fine. Sometimes there is much hype around a certain local optimum that you can still make some nice profit for a few years by riding that hype.

Nevertheless, it seems wise to stay sharp, especially when there are indications that there are major disadvantages to certain type of solutions (local optima). Especially considering other dimensions in which clear disadvantages are being exposed. Because that 2D graph from above is just a simplification.