Paths to Singularity: Intelligence Amplification

_Other Paths to the Singularity: Intelligence Amplification_

              When people speak of creating superhumanly intelligent beings,
         they are usually imagining an AI project. But as I noted at the
         beginning of this paper, there are other paths to superhumanity.
         Computer networks and human-computer interfaces seem more mundane than
         AI, and yet they could lead to the Singularity. I call this
         contrasting approach Intelligence Amplification (IA). IA is something
         that is proceeding very naturally, in most cases not even recognized
         by its developers for what it is. But every time our ability to access
         information and to communicate it to others is improved, in some sense
         we have achieved an increase over natural intelligence. Even now, the
         team of a PhD human and good computer workstation (even an off-net
         workstation!) could probably max any written intelligence test in
         existence.

              And it's very likely that IA is a much easier road to the
         achievement of superhumanity than pure AI. In humans, the hardest
         development problems have already been solved. Building up from within
         ourselves ought to be easier than figuring out first what we really
         are and then building machines that are all of that. And there is at
         least conjectural precedent for this approach.  Cairns-Smith [5] has
         speculated that biological life may have begun as an adjunct to still
         more primitive life based on crystalline growth.  Lynn Margulis [14]
         has made strong arguments for the view that mutualism is the great
         driving force in evolution.

              Note that I am not proposing that AI research be ignored or less
         funded. What goes on with AI will often have applications in IA, and
         vice versa.  I am suggesting that we recognize that in network and
         interface research there is something as profound (and potential wild)
         as Artificial Intelligence. With that insight, we may see projects
         that are not as directly applicable as conventional interface and
         network design work, but which serve to advance us toward the
         Singularity along the IA path.

              Here are some possible projects that take on special
         significance, given the IA point of view:
            o Human/computer team automation: Take problems that are normally
              considered for purely machine solution (like hill-climbing
              problems), and design programs and interfaces that take a
              advantage of humans' intuition and available computer hardware.
              Considering all the bizarreness of higher dimensional
              hill-climbing problems (and the neat algorithms that have been
              devised for their solution), there could be some very interesting
              displays and control tools provided to the human team member.
            o Develop human/computer symbiosis in art: Combine the graphic
              generation capability of modern machines and the esthetic
              sensibility of humans. Of course, there has been an enormous
              amount of research in designing computer aids for artists, as
              labor saving tools.  I'm suggesting that we explicitly aim for a
              greater merging of competence, that we explicitly recognize the
              cooperative approach that is possible. Karl Sims [22] has done
              wonderful work in this direction.
            o Allow human/computer teams at chess tournaments. We already
              have programs that can play better than almost all humans. But
              how much work has been done on how this power could be used by a
              human, to get something even better? If such teams were allowed
              in at least some chess tournaments, it could have the positive
              effect on IA research that allowing computers in tournaments had
              for the corresponding niche in AI.
            o Develop interfaces that allow computer and network access without
              requiring the human to be tied to one spot, sitting in front of a
              computer. (This is an aspect of IA that fits so well with known
              economic advantages that lots of effort is already being spent on
              it.)
            o Develop more symmetrical decision support systems. A popular
              research/product area in recent years has been decision support
              systems. This is a form of IA, but may be too focussed on
              systems that are oracular. As much as the program giving the user
              information, there must be the idea of the user giving the
              program guidance.
            o Use local area nets to make human teams that really work (ie,
              are more effective than their component members). This is
              generally the area of "groupware", already a very popular
              commercial pursuit. The change in viewpoint here would be to
              regard the group activity as a combination organism. In one
              sense, this suggestion might be regarded as the goal of inventing
              a "Rules of Order" for such combination operations. For instance,
              group focus might be more easily maintained than in classical
              meetings. Expertise of individual human members could be isolated
              from ego issues such that the contribution of different members
              is focussed on the team project. And of course shared data bases
              could be used much more conveniently than in conventional
              committee operations. (Note that this suggestion is aimed at team
              operations rather than political meetings. In a political
              setting, the automation described above would simply enforce the
              power of the persons making the rules!)
            o Exploit the worldwide Internet as a combination human/machine
              tool. Of all the items on the list, progress in this is
              proceeding the fastest and may run us into the Singularity before
              anything else. The power and influence of even the present-day
              Internet is vastly underestimated. For instance, I think our
              contemporary computer systems would break under the weight of
              their own complexity if it weren't for the edge that the USENET
              "group mind" gives the system administration and support people!)
              The very anarchy of the worldwide net development is evidence of
              its potential. As connectivity and bandwidth and archive size and
              computer speed all increase, we are seeing something like Lynn
              Margulis' [14] vision of the biosphere as data processor
              recapitulated, but at a million times greater speed and with
              millions of humanly intelligent agents (ourselves).

              The above examples illustrate research that can be done within
         the context of contemporary computer science departments. There are
         other paradigms. For example, much of the work in Artificial
         Intelligence and neural nets would benefit from a closer connection
         with biological life. Instead of simply trying to model and understand
         biological life with computers, research could be directed toward the
         creation of composite systems that rely on biological life for
         guidance or for the providing features we don't understand well enough
         yet to implement in hardware. A long-time dream of science-fiction has
         been direct brain to computer interfaces [2] [28]. In fact, there is
         concrete work that can be done (and has been done) in this area:
            o Limb prosthetics is a topic of direct commercial applicability.
              Nerve to silicon transducers can be made [13].  This is an
              exciting, near-term step toward direct communcation.
            o Similar direct links into brains may be feasible, if the bit
              rate is low: given human learning flexibility, the actual
              brain neuron targets might not have to be precisely selected.
              Even 100 bits per second would be of great use to stroke
              victims who would otherwise be confined to menu-driven
              interfaces.
            o Plugging in to the optic trunk has the potential for bandwidths
              of 1 Mbit/second or so. But for this, we need to know the
              fine-scale architecture of vision, and we need to place an
              enormous web of electrodes with exquisite precision.  If we want
              our high bandwidth connection to be _in addition_ to what paths
              are already present in the brain, the problem becomes vastly more
              intractable. Just sticking a grid of high-bandwidth receivers
              into a brain certainly won't do it.  But suppose that the
              high-bandwidth grid were present while the brain structure was
              actually setting up, as the embryo develops.  That suggests:
            o Animal embryo experiments. I wouldn't expect any IA success
              in the first years of such research, but giving developing brains
              access to complex simulated neural structures might be very
              interesting to the people who study how the embryonic brain
              develops.  In the long run, such experiments might produce
              animals with additional sense paths and interesting intellectual
              abilities.
              
              Originally, I had hoped that this discussion of IA would yield
         some clearly safer approaches to the Singularity. (After all, IA
         allows our participation in a kind of transcendance.) Alas, looking
         back over these IA proposals, about all I am sure of is that they
         should be considered, that they may give us more options. But as for
         safety ...  well, some of the suggestions are a little scarey on their
         face. One of my informal reviewers pointed out that IA for individual
         humans creates a rather sinister elite. We humans have millions of
         years of evolutionary baggage that makes us regard competition in a
         deadly light. Much of that deadliness may not be necessary in today's
         world, one where losers take on the winners' tricks and are coopted
         into the winners' enterprises. A creature that was built _de novo_
         might possibly be a much more benign entity than one with a kernel
         based on fang and talon. And even the egalitarian view of an Internet
         that wakes up along with all mankind can be viewed as a nightmare
         [25].

              The problem is not that the Singularity represents simply the
         passing of humankind from center stange, but that it contradicts some
         of our most deeply held notions of being. I think a closer look at the
         notion of strong superhumanity can show why that is.


         _Strong Superhumanity and the Best We Can Ask for_

              Suppose we could tailor the Singularity. Suppose we could attain
         our most extravagant hopes. What then would we ask for:
         That humans themselves would become their own successors, that
         whatever injustice occurs would be tempered by our knowledge of our
         roots. For those who remained unaltered, the goal would be benign
         treatment (perhaps even giving the stay-behinds the appearance of
         being masters of godlike slaves).  It could be a golden age that also
         involved progress (overleaping Stent's barrier). Immortality (or at
         least a lifetime as long as we can make the universe survive [9]
         [3]) would be achievable.  

              But in this brightest and kindest world, the philosophical
         problems themselves become intimidating. A mind that stays at the same
         capacity cannot live forever; after a few thousand years it would look
         more like a repeating tape loop than a person.  (The most chilling
         picture I have seen of this is in [17].)  To live indefinitely long,
         the mind itself must grow ... and when it becomes great enough, and
         looks back ... what fellow-feeling can it have with the soul that it
         was originally?  Certainly the later being would be everything the
         original was, but so much vastly more. And so even for the individual,
         the Cairns-Smith (or Lynn Margulis) notion of new life growing
         incrementally out of the old must still be valid.

              This "problem" about immortality comes up in much more direct
         ways.  The notion of ego and self-awareness has been the bedrock of
         the hardheaded rationalism of the last few centuries. Yet now the
         notion of self-awareness is under attack from the Artificial
         Intelligence people ("self-awareness and other delusions").
         Intelligence Amplification undercuts the importance of ego from
         another direction.  The post-Singularity world will involve extremely
         high-bandwidth networking. A central feature of strongly superhuman
         entities will likely be their ability to communicate at variable
         bandwidths, including ones far higher than speech or written messages.
         What happens when pieces of ego can be copied and merged, when the
         size of a selfawareness can grow or shrink to fit the nature of the
         problems under consideration?  These are essential features of strong
         superhumanity and the Singularity. Thinking about them, one begins to
         feel how essentially strange and different the Post-Human era will be
         -- _no matter how cleverly and benignly it is brought to be_.

         
              From one angle, the vision fits many of our happiest dreams:
         a place unending, where we can truly know one another and understand
         the deepest mysteries.  From another angle, it's a lot like the worst
         case scenario I imagined earlier in this paper.

              Which is the valid viewpoint? In fact, I think the new era is
         simply too different to fit into the classical frame of good and
         evil. That frame is based on the idea of isolated, immutable minds
         connected by tenuous, low-bandwith links. But the post-Singularity
         world _does_ fit with the larger tradition of change and cooperation
         that started long ago (perhaps even before the rise of biological
         life). I think there _are_ notions of ethics that would apply in such
         an era. Research into IA and high-bandwidth communications should
         improve this understanding.  I see just the glimmerings of this now,
         in Good's Meta-Golden Rule, perhaps in rules for distinguishing self
         from others on the basis of bandwidth of connection. And while mind
         and self will be vastly more labile than in the past, much of what we
         value (knowledge, memory, thought) need never be lost. I think
         Freeman Dyson has it right when he says [8]: "God is what mind becomes
         when it has passed beyond the scale of our comprehension."
Vernor Vinge www-rohan.sdsu.edu/faculty/vin…