"We believe there are at present no easy answers to the big questions, and it will take radical rethinking of the issues before people can be expected to reach a consensus about the meaning of the word "I". This book, then, is designed to provoke, disturb, and befuddle its readers, to make the obvious strange and, perhaps, to make the strange obvious."
-from the preface.
The Mind's I is not a continuous book, but rather a collection of essays, sci-fi scenarios and thought experiments from a range of sources, each accompanied by a set of reflections by either or both of the compilers, Douglas R. Hofstadter and Daniel C. Dennett. It would be impossible to do each selection justice here -they or the full sources from which they are drawn merit writeups of their own- so instead I intend to merely list and give an overview of them.
But can an overview of the book itself be given? As the quote suggests, this is not a book of answers so much as a book of questions. A basic thesis of the mind as program is explored, but not taken as an assumption. Two key ideas from the field of artificial intelligence, the Turing Test and Searle's Chinese Room experiment, are most central to this exploration. Within this the concept of intertwined levels of complexity and self-reference are considered as key to the emergence of a 'self' from material parts (although not to the extent of Hofstadter's work Gödel, Escher, Bach), and the importance of both distinguishing such levels, and the power of muddling them, is stressed. We are invited to consider fantasy, dreams, fiction, illusion, and perhaps above all, simulation- and to ask when these become of equal power to 'real' phenomena or experience.
Imagine a simulation of a hurricane, capturing the essence of such a storm within a computer. Is it somehow less real because our digital hurricane fails to tear apart the chips of the machine? This is a confusing of levels, however- a digital hurricane can no more devastate the computer than a real hurricane can tear apart the laws of physics. So the correct question to ask may be whether our hurricane is real within the simulation. If implemented in a game, would we do our best to keep a character under our control away from it, and if so, is it at some level just as meaningful as a real hurricane? Many would dismiss the environment of a computer game as a fantasty world, and the hurricane with it. But what of phenomena that could make the leap, and cross the boundaries between simulation and reality? Is a simulated mathematical proof any less of a proof? Is there such a thing as an artificial joke? Can we identify any part of human thought that could not conceivably be simulated in another medium? What of other aspects of human mental experience- are digitial emotions somehow less conceivable than digitial thoughts? Why and how would an intelligent machine differ from us- and what must stay the same to deserve the description of consciousness? Can the two even be separated?
I A Sense of Self
1. Borges and I - Jorge Luis Borges
-A version of the text appears here on E2.
2. On Having No Head -D.E. Harding
3. Rediscovering the Mind- Harold J. Morowitz
These first selections seek to ease us into the inquisitive nature of this book. You can read Borges and I on E2, although the translation chosen is different. It introduces the idea of self-representation: for Borges, the challenge was reconciling the internationally famous author with himself; we are led into a click of recognition, a mental leap, when a fairly general piece of descriptive text turns out to be about us, the reader. In On Having No Head, solipsism is introduced, and our willingness to project our own internal processes on others.
This point bears some more consideration, for it will be a key idea of later chapters- how intelligence or conciousness can be identified in another. We reject the notion of having no head as absurd, for clearly everyone else seems to have a head, and when we poke at the area above our neck we are aware there is a structure there- with a mirror we can confirm that it looks like everyone else's. So how about the notion of having no mind? This seems even less probable- as what could be entertaining the notion of having no mind whilst being so unequipped? But then comes the jump- we trusted that we have a head (a feature external to our inner self-awareness) based on the evidence of others. Can we invert this, and based on our innate awareness of a self, assume that all those people with heads are also people with minds, that truly exhibit consciousness?
The third selection by Morowitz indicates the strange role of the observer in quantum physics. That field cannot be formulated without reference to consciousness. This may lead us in a paradoxical circle in our attempt to explain that phenomenon- if a conscious mind can be explained in terms of a brain, then perhaps that brain can be described in terms of biology. But biology reduces to chemistry, which reduces to physics- which reduces to consciousness. How then to step outside the system?
II Soul Searching
4. Computing Machinery and Intelligence- A.M. Turing
-The version here on E2 is actually more complete than that included in The Minds I, which finishes with objection 9 (Argument from ESP) and thus excludes the final section 7, Learning Machines.
5. The Turing Test- A Coffeehouse Conversation - Douglas R. Hofstadter
6. The Princess Ineffabelle- Stanislaw Lem
7. The Soul of Martha, a Beast - Terrel Miedaner
8. The Soul of the Mark III Beast- Terrel Miedaner
The Turing Test is probably the most discussed topic within AI, and such discussion is well represented both within this book and on E2, so I won't add too much here, having already touched on this topic in other writeups. It becomes apparent from reading the text that Turing's work is from a different age and should be viewed within that context. Whilst the original intention of the test could be considered the holy grail of strong AI- a convincingly human-level intelligence within a machine- I do not feel that it actually tests for that. Turing operated on the assumption that to give a convincing discussion of topics from chess to shakespeare would require an understanding of those topics- that is, of the breadth of human intellectual experience. Yet rather like IQ testing, it seems now that all the Turing Test assesses is being good at the Turing Test. The rafts of chatbots (from the classic Eliza to the most recent and heavily disputed claims of Jim Wightman's Nanniebots) don't try to pass by understanding the topics- their challenge is at the level of language manipulation: of convincing, not understanding.
Hofstadter's own following text points out that the Turing Test is a variation on an 'imitation game', the value of which is debateable from the onset. In the original setup, an interrogator is trying to determine which of two candidates, one male, one female, is indeed the female. If the man were to win at this game, we wouldn't want to conclude that he was indeed a woman- but that he can convincingly impersonate one. It is unclear whether moving to intelligence from gender radically alters this: whether a simulation of intelligence is actual intelligence.
Nonetheless, with current efforts being on the simulation of intelligent conversation (not the same goal as simulating intelligence itself!), I fear that even if the Turing Test was passed by some objective metric (e.g. 70% success rate under the setup Turing described) it would be considered invalid on these grounds. As is often the case for AI, once progress has been made against some goal, the goal itself may be rejected as a measure of 'genuine' (i.e., human) intelligence.
Other writings in this section are concerned with this willingness to credit intelligent-looking behaviour with an underlying intelligence: from a primate hooked up to a machine that enables it to speak a limited subset of english, to a robot programmed to avoid destruction. Both ultimately suffer such a fate, yet our emotional reaction to each is rather different- whilst empathy for the obviously living (although less obviously thinking) animal is understandable, why are we so prone to anthropomorphise 'inanimate' machines?
III From Hardware to Software
9. Spirit- Allen Wheelis
10. Selfish Genes and Selfish Memes- Richard Dawkins
-from The Selfish Gene
11. Prelude... Ant Fugue - Douglas R. Hofstadter
-from Gödel, Escher, Bach: An Eternal Golden Braid
12. The Story of a Brain- Arnold Zuboff
Perhaps all this questioning of the nature of mind and consciousness is premature- it may be that the explanation can be derived from the processes which created the primary example we are aware of- us. Dawkins' thoughts on the human body as a survival machine for our genes, and the memetic exchange of ideas as a survival mechanism for us, are offered as a way of generating and preserving order out of chaos. Whilst the first text is an appeal to a sense of soul or spirit guiding our development towards consciousness, in defiance of a more mechanical explanation of human evolution, the concept of the selfish gene is perhaps even more wondrous.
In a delightful example of the dialogues found in Gödel, Escher, Bach the idea of consciousness developing in a system as a whole from essentially mechanical parts is offered in the context of an anthill. Of course, Hofstadter's work is as much about the form as the content, as he takes the opportunity to blend in fugue structures and reference Escher drawings to help this explication of signals, symbols, and meaning emerging from unwitting lower layers (groups of ants carrying out tasks in a deterministic fashion).
Having offered a way to build up thought from unthinking pieces, the final story tries to pull apart such a system, and determine just how much can be taken away before we descend into meaningless. The reductio ad absurdum is to a zen-like encapsulation of all thought as the single firing of a lone neuron- a destruction of the ideas offered so far. However, the methodology used to reach this conclusion is called into question, for although not stated it implies a level of simulation only within the grasp of something like Descartes' infinitely powerful and all-deceiving demon. Further examples are offered that illustrate that it is not merely the components that matter, but their structures and relationships. Thus whilst it is true that books are composed of letters and music of notes, a single character or note cannot on its own define the piece.
IV Mind as Program
13. Where Am I? - Daniel C. Dennett
14. Where Was I?- David Hawley Sanford
15. Beyond Rejection- Justin Leiber
16. Software- Rudy Rucker
17. The Riddle of the Universe and Its Solution- Christopher Cherniak
This collection of more overtly sci-fi work is included to tackle the issue of what would constitute the relocation or copying of a mind. Dennett offers the scenario of preserving a brain in a vat whilst maintaining each original link to the nervous system of its former body by means of radio transmitters. Setting aside the technological implications (besides the scale of such an endeavour, there are issues with the artificial lengthening of nerves and the introduction of lag), the philosophical concerns revolve around the question of the title- where exactly is Dennett? The point of view is assumed to stay with the body- that is, Dennett feels himself to be somewhere behind the eyes, not floating in a tank. Subsequent developments (failure of the radio link, the introduction of a spare body, and ultimately of a spare, electronic, brain) influence this thought process, twisting the parameters in various ways to induce different conclusions. Sanford's narrative works on the same theme, but inverted: leaving the brain in the body, but obtaining the entire sensory perception of other (robotic) bodies. Again, a displacement of point of view is experienced, and the nature of self questioned as a result.
Beyond Rejection and Software contemplate the extreme of the mind as a program- backup and restoration. In the first we consider the challenge of matching a human mind to a different human brain; whilst in the second we start with a robotic mind. This second scenario is more plausible in technical terms, but still calls into question the nature of consciousness- to what extent does the original robot die: are we merely copying, rather than reincarnating? Could we recreate exactly the same (to outward observers) machine minus an internal 'self-symbol', or is that vital to consciousness?
As a tangent, anyone particularly interested in these kind of questions would be well served by the cyberpunk genre of science fiction (the obvious classic is Gibson's Neuromancer, with its sentient AIs, ROM-constructs and sim-stim , but more recently John Courtenay Grimwood offers some interesting ideas such as pacifist buddhist AIs or the meaning of guilt for a machine), as even if the philosophical ramifications aren't always explored they provide much raw material for your own thought experiments.
The final story is perhaps an indulgence for the authors, but is highly thought provoking. It suggests a "Gödel sentence for the human Turing machine"- that is, a pattern of text capable of locking a human observer into an unbreakable trance, despite normal subconscious activity. The power of self-referential systems, even at the price of closed loops of paradox ("This sentence is false", or in a level-muddling vein, "thiss sentence contains threee errors") is used as a jumping point for the emergence of a self from inanimate matter.
V Created Selves and Free Will
18. The Seventh Sally or How Trurl's Own Perfection Led to No Good- Stanislaw Lem
19. Non Serviam- Stanislaw Lem
20. Is God a Taoist?- Raymond M. Smullyan
21. The Circular Ruins- Jorge Luis Borges
22. Minds, Brains, and Programs - John R. Searle
23. An Unfortunate Dualist- Raymond M. Smullyan
A variety of themes are elaborated upon in this section. Lem's work explores the ramifications of perfect simulation of intelligence, and the question of whether there is any difference between such a simulation and a real intelligence. In the more fantastic seventh sally not only is the suffering of the perfect subjects Trurl creates for an exiled king considered; they end up escaping their simulated world, an impossible level jump that makes for great sci-fi. Non Serviam (ingeniously written as a review of a non-existent book revolving around an imaginary field)
meanwhile looks at the possibility of creating purely mathematical beings within a computer, and equipping them with a grasp of english to monitor their progress through the (simulated) generations. What emerges is a group fascinated by the concept of a creator, who due to their grounding in logic cut a particularly analytical path through ideas such as pascal's wager.
From here the question of free will is tackled by Smullyan, exploring the topic through a discussion between God and a mortal who wishes (at first) to reject his free will. This dilemna is similar to the question of how consciousness arises from unconscious parts- where, if at all, the mechanical rules of physics and biology give way to a sense of free will or the ability to exercise our own decision making. In particular, the internal conflict over the desire to sin is picked up to disprove the notion of an individual as a unitary organisation guided by a single will. Instead, the compilers advocate an interpretation based upon multiple sub-personalities, each level simpler and hence more certain of its goals, which in combination and conflict yield the often contradictory or confused nature of what we consider a single person.
The piece by Borges is largely supplemented by Dennett's comments. In the story we are asked to consider a wizard who dreams a person into being; in the reflections the question is whether we are in some sense our body's dream, a person emerging from the activity of unknowing parts.
Two very different interpretations of the core theme are then presented. Searle's chinese room experiment is intended to refute the view of consciousness emerging from lower level uncomprehending activities, but the compilers see it instead as an argument for such a view, at the level of the system. The text presented in The Mind's I consists only of Searle's article, not the large number of responses that accompanied it, but a flavour of many of the arguments can be found in this writeup so I won't attempt to cover it here. Finally, the section is closed with a very short thought experiment- a miracle drug that can annihilate the soul but leave the body operating just as before. This train of thought rapidly derails, and Smullyan claims dualism as the casualty. It is included as a differing way of attacking Searle's argument- rather than trying to resolve the issue of the chinese room, it goes for the throat of dualism. The story essentially disposes of am epiphenomenal stance, but the interactionist flavour of dualism requires more argument, and may in fact be impossible to resolve as it tries to place the mystery of thought in a realm off-limits to scientific investigation.
VI The Inner Eye
24. What Is It Like to Be a Bat?- Thomas Nagel
25. An Epistemological Nightmare- Raymond M. Smullyan
26. A Conversaton with Einstein's Brain- Douglas R. Hofstadter
27. Fiction- Robert Nozick
Thus far a variety of views on how consciousness arises from material parts have been presented. This last section sees a shift in focus to the experience of consciousness, furthering the ideas presented earlier through Who am I? and Where was I?. Nagel asks the question, what is it like to be a bat?- and it is important to be clear on what he means by this. He is not interested merely in experiencing the sensory perception of a bat (utterly alien to our own) or sharing its goals- that would be a case of what is it like for Nagel to be a bat? Rather, the complete mapping of one consciousness to another is being considered- an objective description of subjective experience.
Hofstadter develops this into a theme of "be-able things" and asks what the criteria might be for having such "BAT-itude", that is, identifying where this mapping of consciousness is conceivable, and also examines the way in which we use language to reflect our sense of 'I'. In his conversation with Einstein's Brain a mapping of a human mind onto a (colossal) book is considered (to keep computers out of the equation). This is another Achilles-Tortoise dialogue, and ties into the idea of copying a mind by copying the underlying brain structure in its entirety. The issue becomes to what extent the book (or more accurately, the book and the processing of it) becomes the person so copied- and how to consider multiple copies, damaged copies, or the pace at which the books are processed.
Smullyan's third contribution is used to revive the concepts of level crossings and self-referentialism, by examining an experimental epistemologist: who devises a machine to read his thought processes and becomes dependant upon it to know his own state of mind. Finally, Fiction, a short piece by Nozick (or is it?) considers the way in which writing is a reflection of an author and the authored- the extent to which a fictional first person is real. It is the only piece without accompanying reflections- the challenge is implicit to draw your own. To that end, a very comprehensive set of further reading for each of the six sections is offered.
I first read this book five years ago, and on this re-reading was amazed by the extent to which it had shaped my beliefs and studies over that time. In fact, a large number of my opinions and examples all come from this book, rather than separate sources as I had imagined. At a first reading I was entertained by each of the outlandish scenarios on their own merits, but now find myself in the equally rewarding experience of trying to piece them together as a cohesive whole, to formulate my own ideas. The job description of a mathematician (which I hope to one day be described as!) could simply be, thinking. But then cognitive science gets to raise the ante- to an extent it is thinking about thinking, and this book provides a perfect playground for the imagination in this regard. I hope the above questions have captured your interest- and that I can track down related ideas here on E2 through the softlinking. If you node any of the musings included (or the full text from which they are drawn), let me know!