While it may be impolitic now for technoprogressives to focus on uploading, for radical life extension advocates it is invaluable to have access to brief and compelling arguments in favor of the efficacy of such a process.
It is a process that will be necessary to enable people to live longer than an average of 150 years when they are increasingly likely to die from random accidents encountered in a normal lifestyle. In the past I have made arguments against the feasibility of some forms of uploading as a means of life extension. I would like to correct that error by providing a line of argument in favor of uploading as a life extension measure, though the life that is extended may change dramatically with a dilution of the self into a larger group of minds.
Is Consciousness and the Self Bound to Biology?
Multiple types of scientific evidence point by consistent correlation to matter, and particularly matter in neurophysiological states, as the basis or substrate of cognitive processes and conscious experience. While in philosophy of mind we cannot completely rule out the possibility that such correlation is illusory or that the substrate of thought and experience is something other than the observable brain, there seems to be no compelling evidence-based reason to doubt that our observations of brain activity are observations of the locus of thought and experience.
A decent materialist description of consciousness based on convergent lines of scientific evidence is that its basis or substrate is matter that is disposed to bear a meaningful relationship to something else based on its orientations to other matter in a cognitive system, orientations which allow data to be interpreted and computations to be processed by algorithms (a finite set of well-defined instructions for completing a task). The meaningful interpretations and calculations - whether unconscious or accessible in consciousness - apply either to a cognitive system’s observable environment or its models of internal system processes (“self-models”).
The meaning of such algorithmic “dispositional orientations” seems contextually determined by the neural networks where they occur and the weightings of connected neural networks through which signals are passed and interpretations of signals are performed. They also are transient in that brain activation cascades briefly across particular neural networks, that is, across matter in the system.
Clearly, even during states of wakefulness (ignoring the long lapses in consciousness during sleep), the subject of experience is not identifiable as specific matter that remains the physical locus of experience. The location and “physical identity” of the subject of experience appears to be constantly changing at the micro-biological level, without even considering the progressively mind-bending attributes of matter and energy at the molecular, atomic, and quantum levels of physical reality (as opposed to an unchanging set of atoms or quanta as the physical locus of experience).
The processing of the cognitive system as a whole is what enables holistic interpretations and perceptions to be experienced by transiently selected matter in the system. The actual mechanisms of conscious processing remain in dispute, and there are many questions for which we currently lack conclusive answers. However, the theories and evidence suggest that there exist no insurmountable obstacles to uploading relevant information, algorithms , and even neurophysiological states.
Some convergent lines of research point to co-activated areas of transiently stable oscillations. In connectionist terms from Gerard O’Brien and Jon Opie at the University of Adelaide, the kind of relationship of transiently stable neural networks to their environment has been described as a “second order resemblance relation.” Second order resemblance is when the modeled relations and characteristics of a represented thing are preserved but not in the same format, like a flat topographical map or a transiently reverberating neural network interpreting part of a visual scene or parsing a sentence. The virtue of their perspective is that it describes specific neurophysiological states that could be the basis for dispositional orientations or how neural interrelations are meaningful based only on the activity of the “vehicles of consciousness” or neural network activity. The problem with this perspective is that it currently seems impossible to test.
Dennett has criticized this “pure vehicle” view of consciousness for several reasons [1] including questioning the assumption that any transiently stable activation pattern must represent a perceptual experience. Yet even if we were forced to consider the whole brain at any given time as the subject of experience, the whole can still be reduced to a fluid mix of atoms or even quanta that “perceive” en masse based on their temporary existence within a cognitive system.
Whether or not the subject of experience can be located (transiently) as distinct areas of the brain, the “self” as an entity that emerges from the brain as a whole and over longer time spans than a moment of perception (and certainly not the result of information converging to one point, which would severely constrain information processing capacity). Daniel Dennett describes the self as a “narrative center of gravity,” reflecting its fuzzy nature. William Sims Bainbridge has even created a “personality capture system” [2] that models the personality absent of information on neural structure and connectivity. Martine Rothblatt has done work to establish the existence of “mindfiles” [3], “mindware” [4] and “mindclones” [5] also without neurophysiological data.
Given this potential for deconstruction of the self, maybe it isn’t so strange to think of radical life extension, with and without uploading, as keeping the world populated with people like ourselves [6] - the same motive for having children to pass on genes or writing books to pass on “memes” (or cultural artifacts including ideas). Still, most people considering radical life extension would prefer the closest resemblance to their mind as possible in an upload, and that will involve detailed brain emulation.
Beta-Tests for an Uploading System
An uploading system can be tested using the original to verify the process, inasmuch as such a process can be verified by observation. It isn’t perfect, but it is about as good a verification method as we could hope to find. Three component processes would be involved in the test: integration, growth, and functionality shifting. [7]
- Integration: connections are established to a simulated system that mimics brain structures and activities via a perfusion of wireless interface materials (the simulation might include things like hormone interactions or, alternatively, it might not attempt to model biology in a close way). In his book The Singularity is Near, Ray Kurzweil describes this step as a nano-neural network which could facilitate a robust virtual reality experience. A contemporary example of integration is the way current brain chips interface through biological material to neurons. Now imagine the chip as an interface to a functional simulation of the biological brain that is supported on what today is called a supercomputer (remember, the subject of conscious experience is just matter processing information in a cognitive system, regardless of whether or not the system is biological).
- Growth: expansion of synaptic circuits into the new system in a functional way, i.e. new connections between the brain and model that serve new or existing functions. Think of the way neural networks adapt to a brain chip, only imagine the growth continuing on the other side of that brain chip interface - that the biological brain “grows into” or alters the neural firing and wiring patterns in the detailed simulation of it and affects the functioning of the simulated brain.
- Functionality Shifting: a gradual shift toward reliance on new network portions in the model for exercising an increasing variety of activities and capacities. Visualize the biological brain gradually losing its status as the primary source for brain activity and it becoming increasingly controlled by the simulated brain, to the point of shared control or even domination by the simulated brain. Domination would make sense given that the simulated brain could be altered to enhance processing speed and bandwidth, short- and long-term memory capacity, general intelligence potential, and augmentation via new modules that process software and literally download and transmit complex information including knowledge or crystallized intelligence.
Sensorial, motor, and eventually higher faculty functions could be demonstrated to be effectively supported via a simulation through double-blind tests of observation, movement, and intellectual processing. In the first case, access to a visual scene in a different room could be the proof as long as processing was primarily done with the simulated brain. In the second case, movement of a mechanized arm in another room could be proof with the same condition. Last, higher mental function processing could be demonstrated to occur primarily in the copy brain yet the conclusions reached would be accessible to the original brain.
Once a prototype system is demonstrated to work successfully there would be no rational basis for claiming that hard-uploading or destructive uploading using the same process would not “work” to transfer one’s information theoretic and subjective identity to a new substrate.
The second parallel processing or dominating simulation of a brain would ensure that if the original brain were to die, an enhanced copy would live on from the point of death forward. Moreover, multiple parallel processing units could be operating at various locations to protect against threats such as natural disasters and on a regular basis static back-up copies also could be saved and stored in highly secure locations. Additionally, as minds become increasingly like software and less like neural networks, components of a person’s mind and not just her ideas could be stored in the minds of a multitude of other augmented uploads.
What Kind of Existence for Uploaded Minds?
In Robin Hanson’s “If Uploads Come First,” a 1994 paper with an in-depth analysis of uploading economics and culture, he explores potential dynamics of uploading if uploads are individualistic entities in a highly competitive environment as well as if they have an employee status to the person in the original body. [8] Yet with sufficiently advanced neurological implants it should become possible to communicate directly from one mind to another mind using signals from internal vocalization and auditory processing areas of the brain. In the case of uploads there would be fewer barriers to mental telepathy, especially given the relative ease of testing and adaptation. As minds communicate more fluidly and freely, the emergence of collective minds may result.
With the emergence of collective minds, there is the possibility that a “hive mind” or “collective consciousness” might appear in which the values and opinions of the group trump those of the individual (only more thoroughly than in collectivistic cultures). In such a scenario personal identity could largely dissipate into group identity. Since such a pattern of association reduces the likelihood of vigorous debate, it may pose a challenge to the continued emergence of rational thought and some types of progress - at least from within any given hive mind. [9] However, hive minds might provide some measure of collective security against reckless or brutish behavior by super-powerful uploads if they cannot effectively be constrained by laws and weapons used against human bodies. The Borg of the Star Trek universe represented a dystopian cyborg version of a hive mind. However, as Hanson has pointed out on the “Overcoming Bias” blog, as people become more knowledgeable and intelligent we should expect to see them reach agreement on an increasing number of topics that have a more factually substantive basis than taste alone.
Finally, one or more hive mind that was sufficiently large and powerful could become a god-like collective. Such a scenario seemed to be explored with the creature toward the end of the film Star Trek V: The Final Frontier and with the “star-child” at the very end of the film 2001: A Space Odyssey. If and when such a being emerges, we can hope it will incorporate - or choose to adopt - a “friendliness” utility function (an elemental motivational drive, like pleasure and pain and emotional biases) for the sake of non-uploaded people and for individualistic (and likely less powerful) uploads. The Singularity Institute for Artificial Intelligence is considering basing this feature on “coherent extrapolated volition” or an accurate estimation of what people would want if they had access to all relevant information and used rational methods to make assessments [10], though its utility function is being designed primarily for non-human artificial general intelligence systems that can bootstrap themselves to a super-powerful status. A friendliness utility function will not be an easy goal to achieve, but it is an eminently worthy one for humanity to strive to realize, and one which may determine the nature of god-like minds of the future.
Men rarely (if ever) manage to dream up a god superior to themselves. Most gods have the manners and morals of a spoiled child. - Robert A. Heinlein
1. Dennett, D.C., and C.F. Westbury, Stability Is Not Intrinsic: Commentary on O’Brien and Opie, A Connectionist theory of Phenomenal Experience.
2. Bainbridge, William Sims. Self-Analysis and Preservation Software for Windows. 23 Mar. 2010.
3. Rothblatt, Martine. What are Mindfiles?. 2 April 2009. 11 December 2010.
4. Rothblatt, Martine. What is Mindware?. 9 April 2009. 11 December 2010.
5. Rothblatt, Martine. What are Mindclones?. 4 May 2009. 11 December 2010.
6. Thanks to Robin Hanson for that perspective.
7. Hyink, Ben. Toward Non-fatal Uploading: A New Framework. 8 August 2004. Online PowerPoint. Uploading and Immortality. 23 March 2010.
8. Hanson, Robin. If Uploads Come First: The Crack of a Future Dawn. 8 March 1994. Reprinted with permission from Extropy 6:2 (1994). 23 March 2010.
9. Wikipedia. Hive mind 11 December 2010
10. Singularity Institute for Artificial Intelligence. Research Area 3: AGI Ethical Issues. 2010. 23 March 2010.
IEET Blog
Great article Ben, thanks.
Re “it may be impolitic now for technoprogressives to focus on uploading”: we don’t focus on uploading here—it is just one of the many interesting things discussed on the IEET blog.