인공지능과 실존적 위험 ―비판적 검토―

초록

This paper examines the ‘existential risk’ related to the possible emergence of superintelligent artificial intelligence (AI) and argues that we should pay more attention to the ‘practical risks’ related to current or near-future AI rather than the ‘existential risk’ of future AI. The practical risks of AI are those already realized or highly probable in the near future, which include the risk of AI strengthening the pre-existing social biases and discrimination through non-statistically skewed training data or statistically representative training data of historically accumulated biases. I emphasize the need for an immediate response to these risks whether the controversial exponential growth of AI development is realistic or not. The study concludes that we should formulate our response in consideration of the respect of human values and the inclusive redefinition of them.

키워드