The ‘alignment problem’ is the term given to the surprisingly gnarly problem of teaching an AI what we really want. If the AI is super powerful and super intelligent this becomes exremely important. Even if we can define exactly what we want it to do, what happens if we change our mind, or discover that there are downsides that we had not imagined?

I am gratified to hear some of the top people in the field talking about AI as a child that we need to bring up to be healthy and well adjusted.

Here is a collection of the best talks I have discovered on this subject that may well decide the shape of the future of humanity, and even the whole planet.