To achieve so called 'strong-AI', or even better 'self-conscious AI', we need to build a machine that is a representation of our own capacity to experience, to understand and to reason. The traditional AI-ideas that are based on the premise of adding more and more code, are clearly not the solution.
One of the most important 'hard problems in AI-research is the so called 'symbol grounding problem'; the hypothesis states that as long as you can only use symbols to describe other symbols, there can never be real 'understanding'. To leverage these symbols to the level of 'understanding' we need to 'ground' these symbols in reality. So what we need is a machine that is capable of 'experiencing reality'.