This contribution argues that autonomous weapon systems may have advantages from the perspective of ensuring better respect for international humanitarian law (IHL), if they are one day capable of perceiving the information necessary to comply with IHL, then to apply IHL to that information, and if it can be ensured that they will not do what humans who created them did not want them to do. In view of the author, targeting decisions do not imply subjective value judgements, a machine would be unable to make. However, to respect IHL by using such autonomous systems, agreement must be found on how to interpret certain IHL rules discussed in this contribution properly when a machine executes autonomously attacks according to parameters established by human beings. In particular, the accountability of those who produce and program such systems must be clarified; the sytem must be constantly updated with information on the military advantage to enable it to determine what is a military objective and to apply the proportionality principle; parameters must be established for comparing the performance of autonomous weapon systems with that of human beings in carrying out attacks; and the feasibility of precautions must be redefined in relation with attacks using such systems.