We need to design AI to minimize impact of human fallibility and maximize for human brilliance,.

The more responsability AI is taking the more crucial is the human control over it.

The responsibility over AI has the manufacturer. The manufacturer must bear all residual risks of the product, as with every other product. As the manufacturer cannot always fully bear this, the AI has to do it itself. Therefore ethics musto be programed  within AI. What are the fundamentals for this ethics?

Monitoring of AI. Accountability of AI. 
AI Incident investigation or forensics must be possible.
AI cannot be a blackhole, therefore every decisions reasons must be traceable. 
Boundaries of AI will dissolve. Goals and tasks will be solved by people and AI together as a system.

Vision
AI that makes systems safer
AI that makes people safer
AI that makes the world safer

AI ethics
nothing new to be invented, only existing ethics/human factors to be applied to AI/computing.
Isaac Asimovs laws must be included
Traceability to understand decisions
AI/Computing maintaining itself!
Probably big topics missing like fairness, privacy, etc.

Inspirierd by Rod Muttram „human factors and the challenge of automation“ in his inaugural presidential lecture and Prof Sarah Sharples.

*Under AI/Computing all automatics is meant, from AI, algorithms, computing up to robots and autonomous systems.