Abstract

In the future, when we compare the welfare of a being of one substrate (say, a human) with the welfare of another (say, an artificial intelligence system), we will be making an intersubstrate welfare comparison. In this paper, we argue that intersubstrate welfare comparisons are important, difficult, and potentially tractable. The world might soon contain a vast number of sentient or otherwise significant beings of different substrates, and moral agents will need to be able to compare their welfare levels. However, this work will be difficult, because we lack the same kinds of commonalities across substrates that we have within them. Fortunately, we might be able to make at least some intersubstrate welfare comparisons responsibly in spite of these issues. We make the case for cautious optimism and call for more research.