Commentary: the intentions of washing machines

Abstract

Hancock makes a range of claims but the most important is this: if a machine ‘learns,’ then, eventually, it will become ‘self-aware.’ It is self-awareness, he argues, that will distinguish machines that are merely autonomous (i.e., which work without human intervention, of which there are many) and those which do something else, which become, in the things they do, like us I cannot understand why one would think this move from learning to awareness would happen but Hancock is convinced. One might add that it is not his discipline that leads to this view – there is no human factors research that asserts or demonstrates that self-awareness emerges through learning, for example; or at least as far as I am aware of. Certainly, Hancock does not cite any. On the contrary, it seems that Hancock takes this idea from the AI community, though as it happens it is an argument that coat-tails on similar notions put forward by cognitive scientists. Some philosophers argue the same, too, such as Dennett (For the view from AI and computer science, see Russell, 2019; for the view of cognitive science, see Tallis, 2011; for a review of the philosophy see Harper et al, 2016). Be that as it may, let me focus on this claim and ask what ‘self-awareness’ might mean or how it might be measured. It seems to me that this is a question to do with anthropology. Hence, one way of approaching this is through imagining how people would act when self-awareness is at issue (Pihlström, 2003, pp. 259–286). Or, put another way, one can approach it by asking what someone might mean when they say they are ‘self-aware’? One might ask, too, why would they say it? I think they do so if they are ‘conscious’ of such things as their intentions. ‘I am about to do this’ they say when they are wanting some advice on that course of action. Intentions are a measure of selfawareness. So, is Hancock saying that autonomous machines would be conscious of their intentions and would that mean, too, that they would treat these intentions as accountable matters? Would that mean, say, that washing machines could have intentions of various kinds? And more, would it mean that these emerge from the learning that washine machines engage in? There are a number of thoughts that arise given this anthropological ‘vignette’ of washing machines and their intentions. How would these intentions be shown? Would these machines need to speak? Besides, when would these machines have these intentions? At what point during learning would they arise? After they have been working a while? One might presuppose some answers here – a machine might only ‘speak’ (if that is its mode of accountability) only once it is switched on. Moreover, one imagines a washing machine would not have any intentions when it was being assembled nor would it have any when it was being disassembled either (as it happens, Hancock refers to similar matters when he reminds the reader of one of his many phrases in earlier human factor articles: this time, the phrase ‘isles of autonomy.’ This is an allusion to how current machines are only autonomous at particular moments in their life, as elsewhere in that life they are subject to the control of and management by people. So, here: a washing machine might only have intentions once it has been made and when it is switched on).