One day Artificial Intelligence might become more intelligent than humans. Let’s define it as the day AI becomes better at everything that humans try. Let’s call that AI “superintelligence”. It is not clear whether superintelligence will ever really come about. However, we can say that there is definitely a chance that superintelligence will come into being in the future. If superintelligence occurs, there is a risk of uncontrolled automation doing unexpected things, thereby inflicting a lot of harm. This may be the result of superintelligence trying to make sure it reaches its programmed goal, long after we still want it to. In order to make sure that we do find ourselves in a situation with harmful automation, research has to be done about how to make superintelligence safe and regulating mechanisms have to be implemented to facilitate that safety.

Today we are probably still many years away from creating superintelligence. However, it is even less clear how far we are from creating safe superintelligence. Therefore, there is a chance that we are closer to creating superintelligence than safe superintelligence. Looking at the considerable risk superintelligence brings with it, it is therefore of vital importance that we make a plan to mitigate that risk.

This organisation was founded to be a foundational discussion community that can support the creation of a roadmap towards the implementation of safe superintelligence before the creation of unsafe superintelligence. In order to reach this goal, solutions will have to be excogitated in value alignment, explainability, accountability, bias, limitation of nudging by AI, human agency preservation and many other aspects.