When it comes to Projects, we are eager to get a project plan in place, budgets and milestones, a blue-print, a design or a mockup, schedule meetings, setup a PMO and intiate the change workstream. Regardless whether Waterfall, Agile or something new is driving this, there is a method in place and we will follow it. Now, this may be the best option in a structured situation. But are you sure this is a structured situation, when it comes to Intelligence Engineering. Could it be that the method is in fact a cage that creates more constraints than value?
Imagine a family, a couple and a new-born child: how would you feel about feeding the child only selected information, books, research articles, loads of structured data and some rules. Total non-sense you say? Why would you not do that? Easy. You would not do that, because you know that – in order to develop its own brain – the child needs to start with simple things, take time to absorb everything, process it and gradually pick up more until eventually it is capable of recognizing speech, gesture and even emotions. The learning process of a child is nothing that you would want to dominate with intellectual ideas, ideologies or dogma – rather, you want the child be free in its choice and preferences.
Now imagine an algorithm that is being fed only those data-sets that you deem appropriate or relevant. Could it be, that you may actually not think of everything and thereby limit the algorithms potential to learn something new?
When it comes to developing AIs we need to recognize that developing an AI is not about developing a set of function, but rather we want to create a learning space where the AI can find insights. It is not about a specification that needs to be met, but rather about letting yourself be surprised about the outcome.
As humans we need to let go of the idea, that we know what the AI will come up with. We need a mindset that puts ourself in second place and accept that we actually want to create an intelligence that is smarter than we are. But if we do, we need to let go of all the constraints that a typical SW development project entails. We need to start thinking out-of-the-box.