Appendix A
Imagine a contemporary (2000) multimedia (audio-visual) personal computer system. We will call the system AV-1. Such a system consists of a computer containing a processor, various memories, hardware support for audio inputs and outputs, and video inputs and outputs including MPEG-2. AV-1 has Input-Output devices such as a keyboard, mouse, printer, scanner, modem, speakers, microphones, video cameras and various internal sensors. This is very much like some kind of creature that happens to be immobile and without a soul. Given the proper software it can see (capture and analyze images). A computer can read music and it can play that music approximately as though performed by any instrument or ensemble of instruments by synthesizing the equivalent sounds. A computer can be made to imagine creatures, common or strange, running around and show us what it is imagining on its display monitor, as was done in the movies Star Wars, The Phantom Menace and Jurassic Park. It can hear by capturing and recording sounds from its microphone and by performing speech recognition and music recognition. It can save and recall information from a database. Of course, in the year 2000 all of these computer capabilities are still very primitive. However, what a computer can do becomes more and more extensive with every passing year. Many of the things done routinely today, such as speech recognition or character recognition, were once thought far too difficult for an inexpensive computer.
It is possible to give an informational description of the hardware that is complete and exact, given that we only care about the informational point of view of the hardware. For example, an informational description of the processor is basically an interpreter program, SAV-1 written for some standard computer that can faithfully run any program written for AV-1. For the sake of being definite, we will call the standard computer that runs all simulations “Phi”. While absolutely any computer (that had enough memory) would do, there is a great advantage in using a concrete example so we arbitrarily pick an ordinary PC.
Imagine an AV-1 program speaking (converting text to speech), using loudspeakers and listening to its own speech using microphones (and processing the sound with speech recognition software to recover the original text). We could say that AV-1 is aware of the fact that it is speaking. On the other hand AV-1 could simulate all the transformations that occur starting from the digital textual information representing the words it sent to the speakers to the digital textual information representing the words heard from the microphones. Let’s look at all the steps in a sophisticated end to end simulation of the process:
1. Distortion and errors in the speech synthesis program that converts ASCII text, and other information regarding the emotional content of what is being said and that modulates the speech sounds, into a digital representation of the sound waveform, then
2. Distortion in the conversion from digital numbers which represent a sound waveform into electrical signals sent to a speaker, then
3. Distortion in the conversion of the electrical signals into sound waves produced by the speakers, then
4. Distortions that occur as the sound wave travels to the microphone and those due to extraneous external sounds picked up by the microphone
5. Distortions in the conversion by the microphone into electrical signals
6. Distortions in the conversion back into digital numbers
7. Finally, the processing of the digital numbers by a speech recognition program back into, hopefully, the same ASCII text along with the recovery of the emotional information that modulated the speech sounds.
Nevertheless, while not exact, such a simulation can still be faithful to high level interpretations. When doing this, AV-1 can be said to be consciously thinking about the audio aspects of what it is about to say. A similar process could model the reactions of a person who hears what AV-1 says, who thinks about it and then generates one of a number of plausible replies. If AV-1 is doing all these things, including analyzing high level consequences of other facts that come into its sensory apparatus while it is speaking to a nearby person, it can be said that AV-1 is conscious of what it is doing. Consciousness need not be a mystery.
The technology of simulating one computer with another is such that one can be confident that the results of the simulated computer are, aside from timing, substantively identical to the results of the original computer for any non-pathological program. A pathological program is one that is unnecessarily sensitive to the timing vagaries of various devices such as disk-drive seek times.
Let us imagine that, somehow, we have captured a static soul on a magnetic tape. This means that we have the informational content of Q. We also need P or at least a good approximation to P. Even if we had the Q part of a human soul recorded on a digital tape, we do not yet know how to program an interpreter for P. It is simply the case that we do not yet have sufficient knowledge and understanding as to what is required; there is no intrinsic reason that will prevent us from being able to do so in the future. This would mean that a computer such as Phi, could be engaged in a conversation wherein those conversing with it would feel that the essence of the captured soul had come to life.
To be very precise, consider P to be an interpreter program that runs in Phi. When Phi runs P, that interpreter processes the current state of Q in order to produce the future state of Q. In other words, the operation of P in Phi causes the temporal evolution of Q. If P and Q represent the informational aspects of a soul, then the operation of P in Phi which causes Q to evolve is the process by which a static soul becomes dynamic (which means that the soul can be said to be thinking and acting). Strangely enough, while we need to know more than we do to enable a disembodied soul to think, we do not have to understand how it is that that soul can think. We do not have to understand thinking or how to create thinking de nova. We can clone thousands of copies of a single tree from cells of one tree while it is not necessary for us to be able to create a tree from basic chemical synthesis. I see no need to imagine that we will be forever unable to know how to create a tree from scratch or how to create something that thinks from scratch. It’s that much can be done with partial knowledge.
It is not necessary for the disembodied soul to be cut off from the real world and the people in the real world. With video input, with audio input and output, a disembodied soul operating in Phi will be able to see and recognize people and carry on normal conversations with them. In such a conversation, if you were to ask the disembodied soul whether or not he or she felt that he or she existed, the answer might well be that Descartes’ Cogito applies. If the soul can think, the soul is.
The disembodied soul can be preserved for eternity in the state corresponding to its last living moment, and/or additionally it can be allowed to evolve as time goes by.
The PC Analogy
Say we observe an actively used PC (Personal Computer). Aside from use in the normal fashion, it might also be working simultaneously as a network server and as a bulletin board system. If we wish to characterize what we see and wish to understand it in general terms, we can divide it into the following entities:
1. The physical computer -- the microprocessor, the random access memories (RAM, ROM, and cache), the hard disks, floppy drive, modems, network interfaces, backup tape drive, CD, etc.
2. The software and data that are available to the PC -- this would include the operating system (e.g. Window NT), utilities, applications, games, network, BBS, etc.
3. The nature of what the PC system is connected to -- power, the network, the Internet, phone lines, ISDN, keyboard, mouse, monitor, audio, video, and indirectly other computers and various people. Here, we are more interested in the nature of the information exchanged rather than with hardware details.
Item 1 has a precise definition in terms of the static information that defines an interpreter that can exactly mimic the operation of the PC system. In a formal sense, the interpreter could be a very simple Turing machine. It would make more sense to define it in terms of a very simple canonical computer that is organized much like a simple RISC architecture computer.
