Software Immortals: Science or Faith?

Abstract

According to the early futurist Julian Huxley, human life as we know it is ‘a wretched makeshift, rooted in ignorance’. With modern science, however, ‘the present limitations and miserable frustrations of our existence could be in large measure surmounted’ and human life could be ‘transcended by a state of existence based on the illumination of knowledge’ (1957b, p. 16).

Francis Heylighen on Proudfoot’s “Software Immortals: Science or Faith?”

The Continuity of Embodied Identity

I enjoyed reading Diane Proudfoot’s essay on “technological supernaturalism”, i.e. the belief that human individuals will be resurrected as immortal software entities by some future, God-like artificial intelligence(s) (Proudfoot 2012). Proudfoot thoroughly deconstructs the many dubious assumptions underlying this philosophy, as propounded by authors such as Kurzweil, Bostrom, Moravec and Tipler.

I particularly liked her arguments showing that this purportedly scientific vision is almost wholly parallel to the traditional religious vision in which our souls are promised an eternal life in heavenly bliss after our mortal bodies have passed away. The “terror management” theory (Pyszczynski et al. 1999) that she refers to indeed provides a plausible explanation for why people, whether religiously or scientifically inspired, seem to be drawn so strongly to the idea that their personhood would somehow survive physical death. But we may not even need such a psychological explanation for this glaring similarity between technological and religious supernaturalism: to me it seems obvious that the former is directly inspired by the latter. For example, while Tipler initially presented his ideas as purely scientific inferences, in further writing (Tipler 2007) he made it clear that he is a devout Catholic who takes doctrine rather literally. The motivation to rationalize a pre-existing faith may be less obvious in the case of more humanistic thinkers, like Bostrom or Moravec. But even a staunch atheist cannot avoid being influenced by such a pervasive meme as the belief in an afterlife, and may be tempted to defuse its power to convert people to religion by reinterpreting it scientifically.

After pointing out where I agree with Proudfoot, let me now indicate where we part ways. In my view, her paper falls in the common trap of what may be called “analytic nitpicking”. Philosophers from the analytic tradition investigate issues by making fine-grained distinctions between the different possible meanings of a concept, and then applying logic to draw out the implications of each of these possible interpretations, in particular in order to show how a particular interpretation may lead to some inconsistency or counter-intuitive result. But these “technical distinctions”—to use Proudfoot’s phrase—are in general considered meaningful only by philosophers: scientists and practitioners typically do not care, because these distinctions tend to lack operational significance. A classic example is the zombie thought experiment about consciousness (Chalmers 1995): if a zombie by definition behaves indistinguishably from a normal human, then according to Leibniz’s principle of the identity of the indistinguishables, a zombie must be a human. The zombie argument therefore fails to clarify anything about consciousness.

Proudfoot applies the analytic method to the problem of personal identity: in how far can an “uploaded”, software personality be identical to the original flesh-and-blood person that it is supposed to resurrect? She argues that various interpretations of the identity concept all lead to problems—such as lack of transitivity or the apparently nonsensical conclusion that two independent software instantiations, A and B, are actually one person. I consider this nitpicking because the identity concept, like practically any concept used in real life, is essentially vague and fluid. The recurrent error made by analytic philosophers is to assume that distinctions are absolute and invariant, while in the complex reality that surrounds us distinctions tend to vary across times, observers and contexts (Gershenson and Heylighen 2005).

Apparently universal rules about the logical notion of identity (such as A = B, B = C, therefore A = C), hence, are unlikely to be applicable to the much more fluid notion of personal identity. Proudfoot is to some degree aware of these difficulties, and therefore considers the alternative model of fuzzy logic. But fuzzy logic is still a kind of logic, and therefore built on invariant (albeit fuzzy) distinctions. The nature of personal identity is precisely that it is not invariant. It is not only the case—as the authors cited by Proudfoot point out—that since I was born about every atom in my body has changed, but also that about every bit of knowledge, experience or emotion in my mind has changed. My personality is substantially different from the personality I had when I was born, or even when I was 5, 10, 15, or 20 years old….

The only thing that allows me to state that the Francis Heylighen of today is somehow still the same as the Francis Heylighen of 40 years ago is continuity: during that time, there was a continuing distinction between Francis Heylighen and the rest of the world, even while the nature of that distinction was changing. This continuity was not one of consciousness (which waxed and waned along with my sleep-wake cycle), but of the rough outline of my body and personality. This continuity is precisely what lacks in the resurrection scenarios of the technological supernaturalists. In such scenario, my body and personality break down at my biological death, while my personality (or at least a software equivalent of it) is recreated by a super-intelligent AI many decades later, in a completely different (non-physical) environment.

Proudfoot is right to question the claim that the resurrected personality would be identical to my original personality (together with the more outlandish claims that the AI would feel compelled to resurrect every person that ever lived, or that the information about all these personalities would have survived the inevitable thermodynamic dissipation). However, rather than wandering through “technical distinctions” about identity, she should better have focused on the most glaring difference: the resurrected personality would lack both my body and my environment. While she mentions the situated and embodied perspective on cognition merely in passing, for me it is crucial: the ability to interact with the environment via bodily sensors and effectors is a defining feature of the notions of person, mind, consciousness or intelligence. As I have developed this point in more depth in my criticism of the common view of the Singularity as the emergence of a disembodied super-intelligence (Heylighen 2012), I won’t go into further details here.

However, note that this philosophy does not deny the possibility of attaining some sort of technological immortality: continuity of identity can in principle be maintained by gradually replacing my different body parts by various electronic circuits—as long as these maintain (or augment) my ability to interact with the world via high-bandwidth sensors and effectors. But now we are entering the domain of practical implementation, leaving behind both the metaphysical speculations of the techno-supernaturalists and the Platonic nitpicking of the analytic philosophers…