Ethical robots: the future can heed us

The paper originally appeared in Wired as (Joy 2000), and is available online: http://​www.​wired.​com/​wired/​archive/​8.​04/​joy.​html. I quote in this paper from the online version, and therefore don’t use page numbers. The quotes are of course instantly findable with search over the online version.

The presentation can be found without videos at http://www.kryten.mm.rpi.edu/PRES/CAPOSU0805/sb_robotsfreedom.pdf. Those able to view keynote, which has the videos of PERI in action embedded, can go to http://www.kryten.mm.rpi.edu/PRES/CAPOSU0805/sb_robotsfreedom.key.tar.gz. A full account of PERI and his exploits, which have not until recently had anything to do with autonomy (PERI has been built to match human intelligence in various domains; see e.g., Bringsjord and Schimanski 2003, 2004) can be found at http://www.cogsci.rpi.edu/research/rair/pai.

This is as good a place as any to point out that, as the parentheticals associated with a number of the propositions on the list just given indicate, by the lights of some computationalists we are not pure software, but are embodied creatures. Joy and Moravec (and Hillis) assume that human persons are in the end software that can be attached to this or that body. That seems like a pretty big assumption.

The most recent one appeared in Theoretical Computer Science (Bringsjord and Arkoudas 2004). For a formal list that was up-to-date as of 2003, and reached back to my What Robots Can and Cannot Be (1992), see my Superminds (2003).

However, it does occur to me that it would perhaps be nice if a new argument against computationalism could be introduced in the present paper. Accordingly, here is one such argument, one that happens to be in line with the themes we are reflecting upon herein: Argument No. 4. 1. If computationalism is true, then concerted, global efforts undertaken by the world’s best relevant scientists and engineers to build computing machines with the cognitive power of human beings will succeed after n years of effort—the “clock” having been started in 1950. 2. Concerted, global efforts undertaken by the world’s best relevant scientists and engineers to build computing machines with the cognitive power of human beings have not succeeded after n years of effort (the clock). 3. Computationalism is false. Obviously, my case against Joy hinges not a bit on this argument. But I do think this argument should give pause to today’s computationalists. I have made it a point to ask a number of such strong “believers” how many years of failure would suffice to throw the truth of computationalism into doubt in their minds—and have never received back a number. But clearly, there must exist some n for which Argument No. 4 becomes sound. It seems to me that 50 is large enough, especially given that we have not produced a machine able to converse at the level of a sharp toddler.

Let me point out here that it is entirely possible to do some first-rate thinking predicated on the supposition that human-level robots will eventually arrive. At the conference where I presented the keynote lecture centered around an ancestor of the present paper, such thinking was carried out by Torrance (2005) and Moor (2005).

Any kind of reassurance would require that that which it feels like to be me had been reduced to some kind of third-person specification—which many have said is impossible. I have alluded above to the fact that today’s smartest machines cannot verbally out-duel a sharp toddler. But at least we do have computers that can understand some language, and we continue to press on. But we are really and truly nowhere in an attempt to understand consciousness in machine terms.

Of course, some philosophers (e.g., Parfit 1986) have championed views of personal identity that seem to entail the real possibility of such downloading. But this is quite beside the point on the table, which is whether you would, in my thought-experiment, take the plunge. It is easy enough to describe thought-experiments in which even conservative folks would take the plunge. For example, if you knew that you were going to die in one hour, because an atom bomb is going to be detonated directly below your feet, you might well, out of desperation, give the downloading a shot. But this is a different thought-experiment.