Just One Thing

From time to time I post questions on Facebook about the kind of strange question that keeps me up at night. Usually the kind of question that will either one day form the kernel of a thriller should I have my fictional hat on, or a piece of speculative futurology or creative research should I be wearing that jaunty trilby that day. The questions aren’t unique or particularly original but they are two things:

• fun. At least if your brain is all beaten out of shape the way mine is then this is what you find fun.
• great for getting you to think in different ways. In particular in forcing you to make a decision (do nothing is never an option), just like citizens and governments are forced to every day and then think through the consequences and how those consequences lead to conflicting outcomes, off the wall consequences that only emerge much later, and always disappointing if not downright hurting not only some people but parts of your own psyche that are integral to who you are and how you see and value the world. Whatever the answers, this trains you to develop new ways of balancing things that sit very ill at ease with each other

This particular question should probably be called the Columbo Dilemma, in honour of everyone’s favourite shabby detective who always wanted to know “just one thing”. As always with my questions, there’s no trick. The object is to think as deeply as possible about every possible aspect of the question and where it might take you.

As a very good programmer with a team of very good programmers, and a huge but finite wealth to draw on, you realise that you are in a position to tackle once and for all one of the world’s most pressing problems. But which one? Fortunately, you don’t have to answer that. You are, after all, a very good programmer, so you decide to create an algorithm to answer the question for you – which of the world’s problems should I solve?

1. How do you begin writing such an algorithm? What parameters do you set? What do you tell it to do and with what portions of which data?

2. Would having a more sophisticated algorithm and better processing power enable you to get a better answer but do you worry that at a certain point the algorithm will make a leap such that it incorporates its own self-interest into its calculations, which may be at odds with your goals, and if so do you scale back its capacity, knowing that the only reliable effect will be a less effective answer?