Erowid
 
 
Plants - Drugs Mind - Spirit Freedom - Law Arts - Culture Library  
Follow @Erowid on Instagram!
An Interview with Francis Jeffrey
by TRP
Issue 2, Winter 1997/98
Originally published in The Resonance Project
As a student at U.C. Berkeley in the early '70s, Jeffrey pushed the envelope of computational neuroscience by trying to model on early lab computers the now-famous "binding problem"‚ how the brain unites different pieces into a coherent percept‚ a topic not generally accepted for academic study until the roaring '90s. Jeffrey moved on to the NASA Ames Research Center in Palo Alto, where he wrote programs for ILIAC-IV, the world's first parallel supercomputer. Shortly thereafter he signed on with pioneering experiments in interspecies communication at John Lilly's Human-Dolphin Foundation.

What we need is a system that's completely LUCID while still being able to do all the practical things computers do.

TRP: From what I understand, you were pushing the boundaries of neural computing back in the days when computing was young. What first got you interested in parallel and networked computing?

FJ: Following the seminal experiments of Lawrence Stark (with David Noton) I was doing cognitive network models of context and attention shifting in visual perception. It was clear that this would be a massively parallel process, structured around the seriality of each thread of attention. Then I worked on the Illiac IV, the first "supercomputer," that had 64 parallel computational channels. This was exciting, but I could see that the concurrency we needed was not simply achieved by lock-step parallelism.

TRP: Did the study of the human brain lead you into programming, or was it programming that led you to explore the human brain?

FJ: I just picked up programming because it was useful in psychological research. My first insights were of a physiological character, but it was difficult to use computers in those days for modelling any significant complexity, which is the essence of physiology. So Larry Stark and I cooked up this idea of "neurophysiological computation," which later became known as "computational neurophysiology" and is now an interdepartmental program or major you can take at leading universities.

TRP: How did you come to be involved with John Lilly's experiments in human/dolphin communication?

FJ: By 1973 I was bored with the academic scene, so I took a consulting job with NASA and spent my spare time studying privately with Lilly, whom I'd discovered in my favorite bookstore. Larry cautioned me, "He's brilliant, but not very practical." From his second round of dolphin research in the early 1980s, I discovered that effectively sorting out dolphin vocalizations would require a massively concurrent multiprocessing network to dynamically work on context represented as a multilevel network of expectations. This was conceptually, but not practically, feasible at the time, so I started thinking seriously about designing a system for which this sort of complex activity would be easy, "second nature."

TRP: What do you think of Lilly's early theories on metaprogramming and the human biocomputer?

FJ: "Programming and Metaprogramming in the Human Biocomputer" is sketchy, psychological talk, but it has enormous breadth; I took it as a challenge to explain if and how the brain really might be modeled in terms of computers and programs.

TRP: What was your impetus for creating eLPHIN? What problems were you trying to solve? Where has it lead?

FJ: Well it seemed to me that contemporary reality is largely shaped by technology, so, given the rapid introduction of computers into our lives, it was urgent to get the analogies just right, or we would have a very confused culture. In a 1982 radio interview, Charlie Seraphim of KCBS in San Francisco asked me, "What will all these computers be used for?", and I replied, a bit like Dustin Hoffman's uncle in The Graduate, "Modeling.""Modeling WHAT?" "Everything," I said, and thousands of Silicon Valley commuters listening on their car radios must have simultaneously thought I was nuts, even though they themselves were already doing the practical work in this direction. Modeling each other is exactly what our brains do, which has driven our evolution to large-brained beings ã so if someone or something NEW is now entering this game, we had better grasp what the true relationships are, on the double. It seemed to me that what we needed was a system to make all this completely LUCID, while at the same time doing everything computers could do on a practical level, but doing it better than the Medieval constructs the industry was trapped in.

TRP: What is a "Virtual Neuronal Network"?

FJ: "Virtual Neuronal Network" implies that it is not physically real, and distinguishes it from contemporary "neural nets" which are much simpler abstractions of the interconnection characteristics of real neurons.

TRP: In exactly what sense is eLPHIN a relational processing system?

FJ: eLPHIN is a relational processing system in three senses: (1) the basic structure is flexible relational linkages; (2) relational structures are not only represented but also manipulated, moved and transplanted throughout the system; (3) everything is contextualized and interpreted in relation to users and other points of view.

TRP: If it mimics the neural processing of the human brain, can it make symbolic and associative as well as logical steps in it's processing?

FJ: It is not clear to what extent "it mimics the neural processing of the human brain," simply because little is currently known about how the brain "processes" if it does so at all; so that is just a metaphor someone used in trying to explain eLPHIN. In the design, I incorporated insights I had into the workings of the brain, as well as being generally informed by a study of neuroscience. As to whether it can make symbolic and associative steps in its processing, it is primarily associative; while "symbolic" is difficult to define, I imagine it relates to being holistic and structural, to preserving a gestalt from step to step, which eLPHIN does quite naturally since it is context-preserving and does not throw away meaning like conventional databasing and programming systems do.

TRP: Can it learn?

FJ: The answer is basically YES. Contemporary software is built up, like a great city, out of "bricks," etc., so eventually you have all these terrific structures, but it is extremely expensive to add just a little flexibility at the top, a little relationship, because we have carried to the surface the non-organic principles used in construction all along the way. "Structured programming" is just a representation, at the surface, of the structural principles out of which structured programming was constructed: it is isomorphic to machines and bureaucracies. eLPHIN begins with relationship and flexibility, and then works downward and inward to the "metal." In this respect it is modeled on a natural nervous system -- as far as we understand a nervous system today (and maybe a little farther). It can easily represent patterns of relationship and rapidly assimilate patterns of conduct in relation to the environment. I call it, generically, a Conscious Network System. It is an "omniscient system" in that it is lucid, with every part knowing all its relationships, the whole aware of all its parts, and all of this information available to an observer at every point or interface. I imagine people can find uses for that sort of capability. eLPHIN is designed, and uniquely suited, to assimilate structural information, including structures of conduct through interaction with users and other systems, in the course of time; that seems to be a reasonable definition of "learning."

TRP: Can you envision the Internet evolving into a self-organizing, self-aware entity? If so, how could this happen? Is this something we want to happen?

FJ: Seymour Pappert defined an intelligent system as a network including intelligent and unintelligent entities. Self-awareness, in a limited and objective sense, may be simply "lucidity" as I defined it in relation to eLPHIN's "CNS" (Conscious Network System); but in a broader sense, I think of consciousness as inseparable from intelligence, and vice versa. The intelligence-consciousness that creates the world is present everywhere-everywhen. The trick is to NOT loose touch of it by substituting false icons -- like the worship of supercomputers, systems, internets, "the media" and its McLuhanesque gods.

TRP: If a network could develop an "intelligence," how would that intelligence manifest itself? How would this mesh with the current Sci-Fi model of AI.

FJ: Usually SiFi is about future technology, but its idea of AI is right out of the past. Like Pappert, I define an intelligent system as a network including intelligent beings; so the best a technical network can do is to interfere as little as possible. Not surprisingly, today there seems to be a lot of technological Satanism, with strong elements of popular masochism. By that I mean the worship of the DEFECTS in the system, a system which has exhausted human beings, by being both fast and inflexible. The defects are the most obvious apparition of the unseen, like the thunderbolts of Zeus, so they attract devotion. Just as television has largely replaced the cathedral as popular virtual reality, religious sentiment has transferred onto technology. A really bad scenario would be-- like with nuclear energy--if another deep penetration into nature were to be harnessed for political and economic power before it's widely understood and integrated in the context of culture. Another Bad Idea: we know from history, both recent and ancient, that certain cultural patterns can make people awfully stupid; and we know from industry that economic expediency can sow confusion amplified by immersive technology. Since computers (including their software) are a lot dumber than we are (most of the time), their rising power could be a force for declining intelligence in our world.

TRP: Will dynamic, self-modifying operating systems blur the hierarchical model [distinction] between user, programmer, and program?

FJ: Not just BLUR, but abolish it.

TRP: What is your vision of the future of the Internet?

FJ: It will certainly grow; what I'm rooting for is, like I said in the 1995 patent document [for eLPHIN], "In the ideal expansion of this concept, the entire telecommunications network assumes the capabilities of a general-purpose programmable computer serving its many users."

TRP: Are there any eLPHIN developer's kits, documentation, or sample apps that can be accessed online?

FJ: You can order it on-line. We offer one unit packaging, under the "ANGEL" TM, that can be shipped on a Jaz or Zip disk, or pre-configured on a NeXT workstation, intended for personal use and experimentation. The system is modular, including programs that maintain the virtual cellular space, interfaces, and "Clumps" and "Clumplets" that plug together in ways that are far more flexible and powerful than is possible with Java

(eLPHIN and "Conscious Networks System" are trademarks of ELFNET, INC., "ELFNET" is a registered servicemark.) Please see: http://elfnet.com

©Copyright Resonant Media, 1998. All rights reserved.