Abstract

This paper examines whether machines (algorithms/programs/ AI systems) are apt subjects for praise or blame for some actions or performances. I consider "Move 37" of AlphaGo as a case study. DeepMind’s AlphaGo is an AI algorithm developed to play the game of Go. The AlphaGo utilizes Deep Neural Networks. As AlphaGo is trained through reinforcement learning, the AI algorithm can improve itself over a period of time. Such AI models can go beyond the intended task and perform novel and unpredictable functions. There is a surprise element associated with "Move 37". "Move 37" not only surprises the Go players, the programmers, but also whoever is informed of this unpredicted move. Does someone or something deserve praise or blame for the surprise? If so, who or what deserves the praise or blame for "Move 37"? The programmer cannot be praised for "Move 37", which is either surprising or was not intended or imagined at all. At the same time, would we accept that neither the algorithm deserves praise for the unpredicted move that the algorithm allowed the program to make? From this, would we accept that since neither the programmer nor the algorithm/AI system deserves the praise, there is such a good or exciting move for which no one or nothing could be praised? Would we say this unpredictable move is a move for which no one deserves praise or blame? Wouldn’t there be at least a few who were surprised by the unpredictable move? Should we say that for this pleasant surprise, no one deserves praise? Nonetheless, for us, specifically regarding the particular unpredictable move, we firmly find it counterintuitive to say that there is an exciting move for which no one deserves praise. The surprise element is the result of the property that belongs to the algorithm. It seems quite difficult for us to accept that no one deserves praise for "Move 37" or for similar moves. Therefore, someone or something deserves praise which is a matter of scrutiny.