Research | Concrete problems in AI safety
Concrete problems in AI safety

Concrete problems in AI safety

25 July 2016

This paper explores practical research issues associated with accidents in machine learning and artificial intelligence (AI) systems, due to incorrect objectives, scalability, or choice of behaviour. The authors present five research problems in the field, suggesting ways to mitigate risks in modern machine learning systems.

 

Please login or join for free to read more.