A few days ago, in a post featuring a series of links to stories about new and emerging technologies, I included a link to a review of Nick Bostrom’s new book, Superintelligence: Paths, Dangers, Strategies. Not long afterwards, I came across an essay adapted from Bostrom’s book on Slate’s “Future Tense” blog. The excerpt is given the cheerfully straightforward title, “You Should Be Terrified of Super Intelligent Machines.”
I’m not sure that Bostrom himself would put quite like that. I’ve long thought of Bostrom as one of the more enthusiastic proponents of a posthumanist vision of the future. Admittedly, I’ve not read a great deal of his work (including this latest book). I first came across Bostrom’s name in Cary Wolfe’s What Is Posthumanism?, which led me to Bostrom’s article, “A History of Transhumanist Thought.”
For his part, Wolfe sought to articulate a more persistently posthumanist vision for posthumanism, one which dispensed with humanist assumptions about human nature altogether. In Wolfe’s view, Bostrom was guilty of building his transhumanist vision on a thoroughly humanist understanding of the human being. The humanism in view here, it’s worth clarifying, is that which we ordinarily associate with the Renaissance or the Enlightenment, one which highlights autonomous individuality, agency, and rationality. It is also one which assumes a Platonic or Cartesian mind/body dualism. Wolfe, like N. Katherine Hayles before him, finds this to be misguided and misleading, but I digress.
Whether Bostrom would’ve chosen such an alarmist title or not, his piece does urge us to lay aside the facile assumption that super-intelligent machines will be super-intelligent in a predictably human way. This is an anthropomorphizing fallacy. Consequently, we should consider the possibility that super-intelligent machines will pursue goals that may, as an unintended side-effect, lead to human extinction. I suspect that in the later parts of his book, Bostrom might have a few suggestions about how we might escape such a fate. I also suspect that none of these suggestions include the prospect of halting or limiting the work being done to create super-intelligent machines. In fact, judging from the chapter titles and sub-titles, it seems that the answer Bostrom advocates involves figuring out how to instill appropriate values in super-intelligent machines. This brings us back to the line of criticism articulated by Wolfe and Hayles: the traditionally humanist project of rational control and mastery is still the underlying reality.
It does seem reasonable for Bostrom, who is quite enthusiastic about the possibilities of human enhancement, to be a bit wary about the creation of super-intelligent machines. It would be unfortunate indeed if, having finally figured out how to download our consciousness or perfect a cyborg platform for it, a clever machine of our making later came around, pursuing some utterly trivial goal, and decided, without a hint of malice, that it needed to eradicate these post-human humans as a step toward the fulfillment of its task. Unfortunate, and nihilistically comic.
It is interesting to consider that these two goals we rather blithely pursue–human enhancement and artificial intelligence–may ultimately be incompatible. Of course, that is a speculative consideration, and, to some degree, so is the prospect of ever achieving either of those two goals, at least as their most ardent proponents envision their fulfillment. But let us consider it for just a moment anyway for what it might tell us about some contemporary versions of the posthumanist hope.
Years ago, C.S. Lewis famously warned that the human pursuit of mastery over Nature would eventually amount to the human pursuit of mastery over Humanity, and what this would really mean is the mastery of some humans over others. This argument is all the more compelling now, some 70 or so years after Lewis made it in The Abolition of Man. It would seem, though, that an updated version of that argument would need to include the further possibility that the tools we develop to gain mastery over nature and then humanity might finally destroy us, whatever form the “us” at that unforeseeable juncture happens to take. Perhaps this is the tacit anxiety animating Bostrom’s new work.
And this brings us back, once again, to the kind of humanism at the heart of posthumanism. The posthumanist vision that banks on some sort of eternal consciousness–the same posthumanist vision that leads Ray Kurzweil to take 150 vitamins a day–that posthumanist vision is still the vision of someone who intends to live forever in some clearly self-identifiable form. It is, in this respect, a thoroughly Western religious project insofar as it envisions and longs for the immortality of the individuated self. We might even go so far as to call it, in an obviously provocative move, a Christian heresy.
Finally, our potentially incompatible technical aspirations reveal something of the irrationality, or a-rationality if you prefer, at the heart of our most rational project. Technology and technical systems assume rationality in their construction and their operation. Thinking about their potential risks and trying to prevent and mitigate them is also a supremely rational undertaking. But at the heart of all of this rational work there is a colossal unspoken absence: there is a black hole of knowledge that, beginning with the simple fact of our inability to foresee the full ramifications of anything that we do or make, subsequently sucks into its darkness our ability to expertly anticipate and plan and manage with anything like the confident certainty we project.
It is one thing to live with this relative risk and uncertainty when we are talking about simple tools and machines (hammers, bicycles, etc.). It is another thing when we are talking about complex technical systems (automotive transportation, power grids, etc.). It is altogether something else when we are talking about technical systems that may fundamentally alter our humanity or else eventuate in its annihilation. The fact that we don’t even know how seriously to take these potential threats, that we cannot comfortably distinguish between what is still science fiction and what will, in fact, materialize in our lifetimes, that’s a symptom of the problem, too.
I keep coming back to the realization that our thinking about technology is often inadequate or ineffectual because it is starting from the wrong place; or, to put it another way, it is already proceeding from assumptions grounded in the dynamics of technology and technical systems, so it bends back toward the technological solution. If we already tacitly value efficiency, for example, if efficiency is already an assumed good that no longer needs to be argued for, then we will tend to pursue it by whatever possible means under all possible circumstances. Whenever new technologies appear, we will judge them in light of this governing preference for efficiency. If the new technology affords us a more efficient way of doing something, we will tend to embrace it.
But the question remains, why is efficiency a value that is so pervasively taken for granted? If the answer seems commonsensical, then, I’d humbly suggest that we need to examine it all the more critically. Perhaps we will find that we value efficiency because this virtue native to the working of technical and instrumental systems has spilled over into what had previously been non-technical and non-instrumental realms of human experience. Our thinking is thus already shaped (to put it in the most neutral way possible) by the very technical systems we are trying to think about.
This is but one example of the dynamic. Our ability to think clearly about technology will depend in large measure on our ability to extricate our thinking from the criteria and logic native to technological systems. This is, I fully realize, a difficult task. I would never claim that I’ve achieved this clarity of thought myself, but I do believe that our thinking about technology depends on it.
There’s a lot more to be said, but I’ll leave it there for now. Your thoughts, as always, are welcome.