How do you ask a good AI question?

I have completed over 100 hours of machine learning/AI online courses and I now get to put that learning into practice – I get to pose three questions to a powerful AI program.

The core technique is Natural Language Programming (NLP) – mathematical analysis of text, mainly using Long Short Term Memory – but with a powerful interface and a set of tools which uses NLP to answers to question like:

• What could I ask the program on a particular topic?
• What could the answer(s) be?
• How could I reframe the questions to push my own thinking through my natural biases?

It is a powerful system.

So how should I frame/build my questions for this system to maximize my learning?

The power of machine learning/AI, in my view, is the still the sheer computing power behind the pattern recognition programming. We have not built a judgement engine that mimics the human experience but a powerful data sorter.

So what data do I want to sorted and in what way?

Guidance on good questions says that we should frame it as a vision (not a performance goal).

I have three vision questions:

• what does a world look like with shifting water availability?
• what does a world look like with unlimited electricity energy?
• what does a country with abundant fossil fuels do now?

I want to know, for each vision question, one thing that has the potential to change everything. It might be punching a hole through a key assumption (i.e. what we knew to be true may not be?) Or it could be a link to another problem that we didn’t know.

I think about an AI problem being the ultimate generalist. It is so general that whole sciences are reduced to a numerical array.

So how can I refine these questions to maximize my learning?