Of all options, the safest way forward is actually to accelerate AI research as much as possible.
Since we underestimate the difficulty of building a super-AI, such “premature” research is likely to fail almost totally. This will reveal many unsuspected threats at a level they can be easily contained.
Of all options, the safest way forward is actually to accelerate AI research as much as possible.
Since we underestimate the difficulty of building a super-AI, such “premature” research is likely to fail almost totally. This will reveal many unsuspected threats at a level they can be easily contained.