Philosopher John Haugeland says that one of the most fundamental differences between human beings and machines is that machines, and AIs in particular, don't give a damn. Things matter to us but not to machines. It’s quite plausible that AIs don’t give a damn, at least not yet. But is that just something that humans can do?
Here’s Haugeland, Professor of Philosophy at the University of Chicago, speaking in Tao Ruspoli’s documentary about Martin Heidegger at time code 19:33:
Perhaps I can reach to the one of the most fundamental aspects about the difference between human beings and machines by adverting to something about each of us with which we are all deeply familiar. And that is that it matters to us what happens in the world, what matters; it matters to us what happens to us; it matters to us what happens to our friends; it matters to us the progress of science and philosophy. All of those are desiderata. Those are things to build a life on that one can summarize in the phrase “giving a damn.” And if you have that phrase then you can say in a word, What AI has so far failed to come up with, by saying, “The trouble with computers is that they don't give a damn.”
A few minutes further on, at time code 20:28, Taylor Carman, Professor of Philosophy at Barnard College and Columbia echoed the sentiment:
What we are at bottom, much more fundamental than our being thinking subjects, is that we care about something; something matters to us.
First, I’m not sure about AIs. To the extent they have a goal and the means to achieve it, something (that goal) matters to them. We’re probably not at the point where they have enough power in the world, but the existential risk of paperclip maximizers is that they do give a damn.
Second, I think anything that has agency – defined as acting on the world to achieve some states and avoid others – by definition gives a damn. One can debate what kinds of things have agency. We certainly believe humans do, and most people believe animals do, even down to bacteria for some scholars.
Third, one can treat social institutions (aka ogregores) as machines. I believe at least some of them have agency, see the posts Defining Agency and Degrees of Agency.
We should not equate “giving a damn” as “having the subjective state of giving a damn.” We can’t be sure of the subjective states of anyone other than ourselves. A more reliable test is that something gives a damn because it behaves as if it gives a damn. organizations pass the test. They act to survive and grow; their existence matters to them. Therefore, at least some machines give a damn: institutions now, and some AIs perhaps soon.
If one concedes that, questions arise about whether those entities bear responsibility for their actions; see, e.g., Christian List’s Group Agency and Artificial Intelligence.
No comments:
Post a Comment