This article was published in two parts in Cryonics magazine, January and April 1994. A more highly annotated version can be found on Ralph Merkle's Cryonics Pages.
See also cryobiologist Dr. Gregory Fahy's critique of this paper and Dr. Merkle's response. For an alternative repair scenario contributed by an anonymous biologist that circumvents Dr. Fahy's concerns, see "Realistic Scenario for Nanotechnological Repair of the Frozen Human Brain".
A short version of this paper entitled "The Technical Feasibility of Cryonics," appeared in Medical Hypotheses Vol. 39, 1992; 6-16.
Cryonic suspension is a method of stabilizing the condition of someone who is terminally ill so that they can be transported to the medical care facilities that will be available in the late 21st or 22nd century. There is little dispute that the condition of a person stored at the temperature of liquid nitrogen is stable, but the process of freezing inflicts a level of damage which cannot be reversed by current medical technology. Whether or not the damage inflicted by current methods can ever be reversed depends both on the level of damage and the ultimate limits of future medical technology. The failure to reverse freezing injury with current methods does not imply that it can never be reversed in the future, just as the inability to build a personal computer in 1890 did not imply that such machines would never be economically built. This paper considers the limits of what medical technology should eventually be able to achieve (based on the currently understood laws of chemistry and physics) and the kinds of damage caused by current methods of freezing. It then considers whether methods of repairing the kinds of damage caused by current suspension techniques are likely to be achieved in the future.
Perhaps the most important question in evaluating this option is its technical feasibility: will it work?
Given the remarkable progress of science during the past few centuries it is difficult to dismiss cryonics out of hand. The structure of DNA was unknown prior to 1953; the chemical (rather than "vitalistic") nature of living beings was not appreciated until early in the 20th century; it was not until 1864 that spontaneous generation was put to rest by Louis Pasteur, who demonstrated that no organisms emerged from heat-sterilized growth medium kept in sealed flasks; and Sir Isaac Newton's Principia established the laws of motion in 1687, just over 300 years ago. If progress of the same magnitude occurs in the next few centuries, then it becomes difficult to argue that the repair of frozen tissue is inherently and forever infeasible.
Hesitation to dismiss cryonics is not a ringing endorsement and still leaves the basic question in considerable doubt. Perhaps a closer consideration of how future technologies might be applied to the repair of frozen tissue will let us draw stronger conclusions -- in one direction or the other. Ultimately, cryonics will either (a) work or (b) fail to work. It would seem useful to know in advance which of these two outcomes to expect. If it can be ruled out as infeasible, then we need not waste further time on it. If it seems likely that it will be technically feasible, then a number of nontechnical issues should be addressed in order to obtain a good probability of overall success.
The reader interested in a general introduction to cryonics is referred to other sources[23, 24, 80]. Here, we focus on technical feasibility.
While many isolated tissues (and a few particularly hardy organs) have been successfully cooled to the temperature of liquid nitrogen and rewarmed, further successes have proven elusive. While there is no particular reason to believe that a cure for freezing damage would violate any laws of physics (or is otherwise obviously infeasible), it is likely that the damage done by freezing is beyond the self-repair and recovery capabilities of the tissue itself. This does not imply that the damage cannot be repaired, only that significant elements of the repair process would have to be provided from an external source. In deciding whether such externally provided repair will (or will not) eventually prove feasible, we must keep in mind that such repair techniques can quite literally take advantage of scientific advances made during the next few centuries. Forecasting the capabilities of future technologies is therefore an integral component of determining the feasibility of cryonics. Such a forecast should, in principle, be feasible. The laws of physics and chemistry as they apply to biological structures are well understood and well defined. Whether the repair of frozen tissue will (or will not) eventually prove feasible within the framework defined by those laws is a question which we should be able to answer based on what is known today.
Current research (outlined below) supports the idea that we will eventually be able to examine and manipulate structures molecule by molecule and even atom by atom. Such a technical capability has very clear implications for the kinds of damage that can (and cannot) be repaired. The most powerful repair capabilities that should eventually be possible can be defined with remarkable clarity. The question we wish to answer is conceptually straightforward: will the most powerful repair capability that is likely to be developed in the long run (perhaps over a few centuries) be adequate to repair tissue that is frozen using the best available current methods?[note 2] Eigler and Schweizer have already developed the capability "... to fabricate rudimentary structures of our own design, atom by atom." Eigler said, "...by the time I'm ready to kick the bucket, we might be able to store enough information on my exact physical makeup that someday we'll be able to reassemble me, atom by atom."
The general purpose ability to manipulate structures with atomic precision and low cost is often called nanotechnology (also called molecular engineering, molecular manufacturing, molecular nanotechnology , etc.). There is widespread belief that such a capability will eventually be developed [1, 2, 3, 4, 7, 8, 10, 19, 41, 47, 49, 83, 84, 85, 106, 107, 108, 116, 117, 118, 119, 121, 122] though exactly how long it will take is unclear. The long storage times possible with cryonic suspension make the precise development time of such technologies noncritical. Development any time during the next few centuries would be sufficient to save the lives of those suspended with current technology.
In this paper, we give a brief introduction to nanotechnology and then clarify the technical issues involved in applying it in the conceptually simplest and most powerful fashion to the repair of frozen tissue.
This concept is receiving increasing attention in the research community. There have been two international research conferences directly on molecular manufacturing [83, 84, 116, 121] as well as a broad range of conferences on related subjects. Science [47, page 26] said "The ability to design and manufacture devices that are only tens or hundreds of atoms across promises rich rewards in electronics, catalysis, and materials. The scientific rewards should be just as great, as researchers approach an ultimate level of control -- assembling matter one atom at a time." "Within the decade, [John] Foster [at IBM Almaden] or some other scientist is likely to learn how to piece together atoms and molecules one at a time using the STM [Scanning Tunneling Microscope]."
Eigler and Schweizer at IBM reported on "...the use of the STM at low temperatures (4 K) to position individual xenon atoms on a single-crystal nickel surface with atomic precision. This capacity has allowed us to fabricate rudimentary structures of our own design, atom by atom. The processes we describe are in principle applicable to molecules also. In view of the device-like characteristics reported for single atoms on surfaces [omitted references], the possibilities for perhaps the ultimate in device miniaturization are evident."
J. A. Armstrong, IBM Chief Scientist and Vice President for Science and Technology said "I believe that nanoscience and nanotechnology will be central to the next epoch of the information age, and will be as revolutionary as science and technology at the micron scale have been since the early '70's.... Indeed, we will have the ability to make electronic and mechanical devices atom-by-atom when that is appropriate to the job at hand."
The New York Times said: "Scientists are beginning to gain the ability to manipulate matter by its most basic components --- molecule by molecule and even atom by atom." "That ability, while now very crude, might one day allow people to build almost unimaginably small electronic circuits and machines, producing, for example, a super computer invisible to the naked eye. Some futurists even imagine building tiny robots that could travel through the body performing surgery on damaged cells."
Drexler[1,10,19,41, 85] has proposed the assembler, a small device resembling an industrial robot which would be capable of holding and positioning reactive compounds in order to control the precise location at which chemical reactions take place. This general approach should allow the construction of large atomically precise objects by a sequence of precisely controlled chemical reactions.
The best technical discussion of nanotechnology has recently been provided by Drexler[ 85].
The instructions that the ribosome follows in building a protein are provided by mRNA (messenger RNA). This is a polymer formed from the 4 bases adenine, cytosine, guanine, and uracil. A sequence of several hundred to a few thousand such bases codes for a specific protein. The ribosome "reads" this "control tape" sequentially, and acts on the directions it provides.
Calculations indicate that an assembler need not inherently be very large. Enzymes "typically" weigh about 105 amu (atomic mass units [note 3]), while the ribosome itself is about 3 x 106 amu. The smallest assembler might be a factor of ten or so larger than a ribosome. Current design ideas for an assembler are somewhat larger than this: cylindrical "arms" about 100 nanometers in length and 30 nanometers in diameter, rotary joints to allow arbitrary positioning of the tip of the arm, and a worst-case positional accuracy at the tip of perhaps 0.1 to 0.2 nanometers, even in the presence of thermal noise[ 85]. Even a solid block of diamond as large as such an arm weighs only sixteen million amu, so we can safely conclude that a hollow arm of such dimensions would weigh less. Six such arms would weigh less than 108 amu.
An assembler might have a kilobyte of high speed (rod-logic based) RAM, (similar to the amount of RAM used in a modern one-chip computer) and 100 kilobytes of slower but more dense "tape" storage -- this tape storage would have a mass of 108 amu or less (roughly 10 atoms per bit -- see below). Some additional mass will be used for communications (sending and receiving signals from other computers) and power. In addition, there will probably be a "toolkit" of interchangeable tips that can be placed at the ends of the assembler's arms. When everything is added up a small assembler, with arms, computer, "toolkit," etc. should weigh less than 109 amu.
E. coli (a common bacterium) weighs about 1012 amu[14, page 123]. Thus, an assembler should be much larger than a ribosome, but much smaller than a bacterium.
Further work on self-replicating systems was done by NASA in 1980 in a report that considered the feasibility of implementing a self-replicating lunar manufacturing facility with conventional technology. One of their conclusions was that "The theoretical concept of machine duplication is well developed. There are several alternative strategies by which machine self-replication can be carried out in a practical engineering setting." They estimated it would require 20 years (and many billions of dollars) to develop such a system. While they were considering the design of a macroscopic self-replicating system (the proposed "seed" was 100 tons) many of the concepts and problems involved in such systems are similar regardless of size.
A large atomically precise structure, however, can be viewed as simply a collection of small atomically precise objects which are then linked together. To build a truly broad range of large atomically precise objects requires the ability to create highly specific positionally controlled bonds. A variety of highly flexible synthetic techniques have been considered by Drexler [ 85]. We shall describe two such methods here to give the reader a feeling for the kind of methods that will eventually be feasible.
We assume that positional control is available and that all reactions take place in a hard vacuum. The use of a hard vacuum allows highly reactive intermediate structures to be used, e.g., a variety of radicals with one or more dangling bonds. Because the intermediates are in a vacuum, and because their position is controlled (as opposed to solutions, where the position and orientation of a molecule are largely random), such radicals will not react with the wrong thing for the very simple reason that they will not come into contact with the wrong thing.
It is difficult to maintain biological structures in a hard vacuum at room temperature because of water vapor and the vapor of other small compounds. By sufficiently lowering the temperature, however, it is possible to reduce the vapor pressure to effectively 0.
Normal solution-based chemistry offers a smaller range of controlled synthetic possibilities. For example, highly reactive compounds in solution will promptly react with the solution. In addition, because positional control is not provided, compounds randomly collide with other compounds. Any reactive compound will collide randomly and react randomly with anything available (including itself). Solution-based chemistry requires extremely careful selection of compounds that are reactive enough to participate in the desired reaction, but sufficiently non-reactive that they do not accidentally participate in undesired side reactions. Synthesis under these conditions is somewhat like placing the parts of a radio into a box, shaking, and pulling out an assembled radio. The ability of chemists to synthesize what they want under these conditions is amazing.
Much of current solution-based chemical synthesis is devoted to preventing unwanted reactions. With assembler-based synthesis, such prevention is a virtually free by-product of positional control.
To illustrate positional synthesis in vacuum somewhat more concretely, let us suppose we wish to bond two compounds, A and B. As a first step, we could utilize positional control to selectively abstract a specific hydrogen atom from compound A. To do this, we would employ a radical that had two spatially distinct regions: one region would have a high affinity for hydrogen while the other region could be built into a larger "tip" structure that would be subject to positional control. A simple example would be the 1-propynyl radical, which consists of three co-linear carbon atoms and three hydrogen atoms bonded to the sp3 carbon at the "base" end. The radical carbon at the radical end is triply bonded to the middle carbon, which in turn is singly bonded to the base carbon. In a real abstraction tool, the base carbon would be bonded to other carbon atoms in a larger diamondoid structure which would provide positional control, and the tip might be further stabilized by a surrounding "collar" of unreactive atoms attached near the base that would limit lateral motions of the reactive tip.
The affinity of this structure for hydrogen is quite high. Propyne (the same structure but with a hydrogen atom bonded to the "radical" carbon) has a hydrogen-carbon bond dissociation energy in the vicinity of 132 kilocalories per mole. As a consequence, a hydrogen atom will prefer being bonded to the 1-propynyl hydrogen abstraction tool in preference to being bonded to almost any other structure. By positioning the hydrogen abstraction tool over a specific hydrogen atom on compound A, we can perform a site specific hydrogen abstraction reaction. This requires positional accuracy of roughly a bond length (to prevent abstraction of an adjacent hydrogen). Quantum chemical analysis of this reaction by Musgrave et. al. show that the activation energy for this reaction is low, and that for the abstraction of hydrogen from the hydrogenated diamond (111) surface (modeled by isobutane) the barrier is very likely zero.
Having once abstracted a specific hydrogen atom from compound A, we can repeat the process for compound B. We can now join compound A to compound B by positioning the two compounds so that the two dangling bonds are adjacent to each other, and allowing them to bond.
This illustrates a reaction using a single radical. With positional control, we could also use two radicals simultaneously to achieve a specific objective. Suppose, for example, that two atoms A1 and A2 which are part of some larger molecule are bonded to each other. If we were to position the two radicals X1 and X2 adjacent to A1 and A2, respectively, then a bonding structure of much lower free energy would be one in which the A1-A2 bond was broken, and two new bonds A1-X1 and A2-X2 were formed. Because this reaction involves breaking one bond and making two bonds (i.e., the reaction product is not a radical and is chemically stable) the exact nature of the radicals is not critical. Breaking one bond to form two bonds is a favored reaction for a wide range of cases. Thus, the positional control of two radicals can be used to break any of a wide range of bonds.
A range of other reactions involving a variety of reactive intermediate compounds (carbenes are among the more interesting ones) are proposed in , along with the results of semi-empirical and ab initio quantum calculations and the available experimental evidence.
Another general principle that can be employed with positional synthesis is the controlled use of force. Activation energy, normally provided by thermal energy in conventional chemistry, can also be provided by mechanical means. Pressures of 1.7 megabars have been achieved experimentally in macroscopic systems. At the molecular level such pressure corresponds to forces that are a large fraction of the force required to break a chemical bond. A molecular vise made of hard diamond-like material with a cavity designed with the same precision as the reactive site of an enzyme can provide activation energy by the extremely precise application of force, thus causing a highly specific reaction between two compounds.
To achieve the low activation energy needed in reactions involving radicals requires little force, allowing a wider range of reactions to be caused by simpler devices (e.g., devices that are able to generate only small force). Further analysis is provided in .
Feynman said: "The problems of chemistry and biology can be greatly helped if our ability to see what we are doing, and to do things on an atomic level, is ultimately developed -- a development which I think cannot be avoided." Drexler has provided the substantive analysis required before this objective can be turned into a reality. We are nearing an era when we will be able to build virtually any structure that is specified in atomic detail and which is consistent with the laws of chemistry and physics. This has substantial implications for future medical technologies and capabilities.
A single repair device of the kind described will not, by itself, have sufficient memory to store the programs required to perform all the repairs. However, if it is connected to a network (in the same way that current computers can be connected into a local area network) then a single large "file server" can provide the needed information for all the repair devices on the network. The file server can be dedicated to storing information: all the software and data that the repair devices will need. Almost the entire mass of the file server can be dedicated to storage, it can service many repair devices, and can be many times the size of one device without greatly increasing system size. Combining these advantages implies the file server will have ample storage to hold whatever programs might be required during the course of repair. In a similar fashion, if further computational resources are required they can be provided by "large" compute servers located on the network.
One consequence of the existence of assemblers is that they are cheap. Because an assembler can be programmed to build almost any structure, it can in particular be programmed to build another assembler. Thus, self reproducing assemblers should be feasible and in consequence the manufacturing costs of assemblers would be primarily the cost of the raw materials and energy required in their construction. Eventually (after amortization of possibly quite high development costs), the price of assemblers (and of the objects they build) should be no higher than the price of other complex structures made by self-replicating systems. Potatoes -- which have a staggering design complexity involving tens of thousands of different genes and different proteins directed by many megabits of genetic information -- cost well under a dollar per pound.
The brain, like all the familiar matter in the world around us, is made of atoms. It is the spatial arrangement of these atoms that distinguishes an arm from a leg, the head from the heart, and sickness from health. This view of the brain is the framework for our problem, and it is within this framework that we must work. Our problem, broadly stated, is that the atoms in a frozen brain are in the wrong places. We must put them back where they belong (with perhaps some minor additions and removals, as well as just rearrangements) if we expect to restore the natural functions of this most wonderful organ.
In principle, the most that we could usefully know about the frozen brain would be the coordinates of each and every atom in it (though confer note 5 ). This knowledge would put us in the best possible position to determine where each and every atom should go. This knowledge, combined with a technology that allowed us to rearrange atomic structure in virtually any fashion consistent with the laws of chemistry and physics, would clearly let us restore the frozen structure to a fully functional and healthy state.
In short, we must answer three questions:
Rather than directly consider these questions at once, we shall first consider a simpler problem: how would we go about describing the position of every atom if somehow this information was known to us? The answer to this question will let us better understand the harder questions.
Other work which considers the information required to describe a human being exists[127, 128].
Thus, if we could store 100 bits of information for every atom in the brain, we could fully describe its structure in as exacting and precise a manner as we could possibly need. (Dancoff and Quastler, using a somewhat better encoding scheme, say that 24.5 bits per atoms should suffice). A memory device of this capacity should be quite literally possible. To quote Feynman: "Suppose, to be conservative, that a bit of information is going to require a little cube of atoms 5 x 5 x 5 -- that is 125 atoms." This is indeed conservative. Single stranded DNA already stores a single bit in about 16 atoms (excluding the water that it's in). It seems likely we can reduce this to only a few atoms. The work at IBM suggests a rather obvious way in which the presence or absence of a single atom could be used to encode a single bit of information (although some sort of structure for the atom to rest upon and some method of sensing the presence or absence of the atom will still be required, so we would actually need more than one atom per bit in this case). If we conservatively assume that the laws of chemistry inherently require 10 atoms to store a single bit of information, we still find that the 100 bits required to describe a single atom in the brain can be represented by about 1,000 atoms. Put another way, the location of every atom in a frozen structure is (in a sense) already encoded in that structure in an analog format. If we convert from this analog encoding to a digital encoding, we will increase the space required to store the same amount of information. That is, an atom in three-space encodes its own position in the analog value of its three spatial coordinates. If we convert this spatial information from its analog format to a digital format, we inflate the number of atoms we need by perhaps as much as 1,000. If we digitally encoded the location of every atom in the brain, we would need 1,000 times as many atoms to hold this encoded data as there are atoms in the brain. This means we would require roughly 1,000 times the volume. The brain is somewhat over one cubic decimeter, so it would require somewhat over one cubic meter of material to encode the location of each and every atom in the brain in a digital format suitable for examination and modification by a computer.
While this much memory is remarkable by today's standards, its construction clearly does not violate any laws of physics or chemistry. That is, it should literally be possible to store a digital description of each and every atom in the brain in a memory device that we will eventually be able to build.
As the molecule we are describing gets larger and larger, the savings in storage gets bigger and bigger. A whole protein molecule will still require only 150 bits to describe, even though it is made of thousands of atoms. The canonical position of every atom in the molecule is specified once the type of the molecule (which occupies a mere 20 bits) is given. A large molecule might adopt many configurations, so it might at first seem that we'd require many more bits to describe it. However, biological macromolecules typically assume one favored configuration rather than a random configuration, and it is this favored configuration that we will describe [note 6].
We can do even better: the molecules in the brain are packed in next to each other. Having once described the position of one, we can describe the position of the next molecule as being such-and-such a distance from the first. If we assume that two adjacent molecules are within 10 nanometers of each other (a reasonable assumption) then we need only store 10 bits of "delta X," 10 bits of "delta Y," and 10 bits of "delta Z" rather than 33 bits of X, 33 bits of Y, and 33 bits of Z. This means our molecule can be described in only 10+10+10+20+30 or 80 bits.
We can compress this further by using various other clever stratagems (50 bits or less is quite achievable), but the essential point should be clear. We are interested in molecules, and describing a molecule takes fewer bits than describing an atom.
While this reduces our storage requirements quite a bit, we could go much further. Instead of describing molecules, we could describe entire sub-cellular organelles. It seems excessive to describe a mitochondrion by describing each and every molecule in it. It would be sufficient simply to note the location and perhaps the size of the mitochondrion, for all mitochondria perform the same function: they produce energy for the cell. While there are indeed minor differences from mitochondrion to mitochondrion, these differences don't matter much and could reasonably be neglected.
We could go still further, and describe an entire cell with only a general description of the function it performs: this nerve cell has synapses of a certain type with that other cell, it has a certain shape, and so on. We might even describe groups of cells in terms of their function: this group of cells in the retina performs a "center surround" computation, while that group of cells performs edge enhancement. Cherniak said: "On the usual assumption that the synapse is the necessary substrate of memory, supposing very roughly that (given anatomical and physiological "noise") each synapse encodes about one binary bit of information, and a thousand synapses per neuron are available for this task: 1010 cortical neurons x 103 synapses = 1013 bits of arbitrary information (1.25 terabytes) that could be stored in the cerebral cortex."
This kind of logic can be continued, but where does it stop? What is the most compact description which captures all the essential information? While many minor details of neural structure are irrelevant, our memories clearly matter. Any method of describing the human brain which resulted in loss of long term memory has rather clearly gone too far. When we examine this quantitatively, we find that preserving the information in our long term memory might require as little as 109 bits (somewhat over 100 megabytes). We can say rather confidently that it will take at least this much information to adequately describe an individual brain. The gap between this lower bound and the molecule-by-molecule upper bound is rather large, and it is not immediately obvious where in this range the true answer falls. We shall not attempt to answer this question, but will instead (conservatively) simply adopt the upper bound.
death \'deth\ n [ME deeth, fr. OE death; akin to ON dauthi death, deyja to die -- more at DIE] 1: a permanent cessation of all vital functions : the end of life
Webster's New Collegiate Dictionary
Determining when "a permanent cessation of all vital functions" has occurred is
not easy. Historically, premature declarations of death and subsequent burial
alive have been a major problem. In the seventh century, Celsus wrote "... Democritus,
a man of well merited celebrity, has asserted that there are in reality, no characteristics
of death sufficiently certain for physicians to rely upon."[87, page 166].
Montgomery, reporting on the evacuation of the Fort Randall Cemetery, states that nearly two percent of those exhumed were buried alive.
Many people in the nineteenth century, alarmed by the prevalence of premature burial, requested, as part of the last offices, that wounds or mutilations be made to assure that they would not awaken ... embalming received a considerable impetus from the fear of premature burial.
Each new medical advance forces a reexamination and possible change of the existing ad hoc criteria. The criteria used by the clinician today to determine "death" are dramatically different from the criteria used 100 years ago, and have changed more subtly but no less surely in the last decade [note 7]. It seems almost inevitable that the criteria used 100 years from now will differ dramatically from the criteria commonly employed today.
These ever shifting criteria for "death" raise an obvious question: is there a definition which will not change with advances in technology? A definition which does have a theoretical underpinning and is not dependent on the technology of the day?
The answer arises from the confluence and synthesis of many lines of work, ranging from information theory, neuroscience, physics, biochemistry and computer science to the philosophy of the mind and the evolving criteria historically used to define death.
When someone has suffered a loss of memory or mental function, we often say they "aren't themselves." As the loss becomes more serious and all higher mental functions are lost, we begin to use terms like "persistent vegetative state." While we will often refrain from declaring such an individual "dead," this hesitation does not usually arise because we view their present state as "alive" but because there is still hope of recovery to a healthy state with memory and personality intact. From a physical point of view we believe there is a chance that their memories and personalities are still present within the physical structure of the brain, even though their behavior does not provide direct evidence for this. If we could reliably determine that the physical structures encoding memory and personality had in fact been destroyed, then we would abandon hope and declare the person dead.
Considerations like this lead to the information theoretic criterion of death [note 8]. A person is dead according to the information theoretic criterion if their memories, personality, hopes, dreams, etc. have been destroyed in the information theoretic sense. That is, if the structures in the brain that encode memory and personality have been so disrupted that it is no longer possible in principle to restore them to an appropriate functional state then the person is dead. If the structures that encode memory and personality are sufficiently intact that inference of the memory and personality are feasible in principle, and therefore restoration to an appropriate functional state is likewise feasible in principle, then the person is not dead.
A simple example from computer technology is in order. If a computer is fully functional then its memory and "personality" are completely intact. If it fell out the seventh floor window to the concrete below, it would rapidly cease to function. However, its memory and "personality" would still be present in the pattern of magnetizations on the disk. With sufficient effort, we could completely repair the computer with its memory and "personality" intact [note 9].
In a similar fashion, as long as the structures that encode the memory and personality of a human being have not been irretrievably "erased" (to use computer jargon) then restoration to a fully functional state with memory and personality intact is in principle feasible. Any technology independent definition of "death" should conclude that such a person is not dead, for a sufficiently advanced technology could restore the person to a healthy state.
On the flip side of the coin, if the structures encoding memory and personality have suffered sufficient damage to obliterate them beyond recognition, then death by the information theoretic criterion has occurred. An effective method of insuring such destruction is to burn the structure and stir the ashes. This is commonly employed to insure the destruction of classified documents. Under the name of "cremation" it is also employed on human beings and is sufficient to insure that death by the information theoretic criterion takes place.
The ethicist and prolific author Robert Veatch said, in Death, Dying, and the Biological Revolution, "An `artificial brain' is not possible at present, but a walking, talking, thinking individual who had one would certainly be considered living."[15, page 23].
The noted philosopher of consciousness Paul Churchland said, in Matter and Consciousness, "If machines do come to simulate all of our internal cognitive activities, to the last computational detail, to deny them the status of genuine persons would be nothing but a new form of racism."[12, page 120].
Hans Moravec, renowned roboticist and Director of the Mobile Robot Lab at Carnegie Mellon said, "Body-identity assumes that a person is defined by the stuff of which a human body is made. Only by maintaining continuity of body stuff can we preserve an individual person. Pattern-identity, conversely, defines the essence of a person, say myself, as the pattern and the process going on in my head and body, not the machinery supporting that process. If the process is preserved, I am preserved. The rest is mere jelly."[50, page 117].
Another issue is not so much philosophical as emotional. Major surgery is not a pretty sight. There are few people who can watch a surgeon cut through living tissue with equanimity. In a heart transplant, for example, surgeons cut open the chest of a dying patient to rip out their dying heart, cut open a fresh cadaver to seize its still-beating heart, and then stitch the cadaver's heart into the dying patients chest. Despite this (which would have been condemned in the middle ages as the blackest of black magic), we cheer the patient's return to health and are thankful that we live in an era when medicine can save lives that were formerly lost.
The mechanics of examining and repairing the human brain, possibly down to the level of individual molecules, might not be the best topic for after dinner conversation. While the details will vary depending on the specific method used, this could also be described by lurid language that failed to capture the central issue: the restoration to full health of a human being.
A final issue that should be addressed is that of changes introduced by the process of restoration itself. The exact nature and extent of these changes will vary with the specific method. Current surgical techniques, for example, result in substantial tissue changes. Scarring, permanent implants, prosthetics, etc. are among the more benign outcomes. In general, methods based on a sophisticated ability to rearrange atomic structure should result in minimal undesired alterations to the tissue.
"Minimal changes" does not mean "no changes." A modest amount of change in molecular structure, whatever technique is used, is both unavoidable and insignificant. The molecular structure of the human brain is in a constant state of change during life -- molecules are synthesized, utilized, and catabolized in a continuous cycle. Cells continuously undergo slight changes in morphology. Cells also make small errors in building their own parts. For example, ribosomes make errors when they build proteins. About one amino acid in every 10,000 added to a growing polypeptide chain by a ribosome is incorrect[14, page 383]. Changes and errors of a similar magnitude introduced by the process of restoration can reasonably be neglected.
In this one instance, we must ask not whether the person is dead by today's (clearly technology dependent) criteria, but whether the person is dead by all future criteria. In short, we must ask whether death by the information theoretic criterion has taken place. If it has not, then cryonic suspension is a reasonable (and indeed life saving) course of action.
The scientifically correct experiment to verify that cryonics works (or demonstrate that it does not work) is quite easy to describe:
This kind of problem is not entirely unique to cryonics. A new AIDS treatment might undergo clinical trials lasting a few years. The ethical dilemma posed by the terminally ill AIDS patient who might be assisted by the experimental treatment is well known. If the AIDS patient is given the treatment prior to completion of the clinical trials, it is possible that his situation could be made significantly worse. On the other hand, to deny a potentially life saving treatment to someone who will soon die anyway is ethically untenable.
In the case of cryonics this is not an interim dilemma pending the (near term) outcome of clinical trials. It is a dilemma inherent in the nature of the proposal. Clinical trials, the bulwark of modern medical practice, are useless in resolving the effectiveness of cryonics in a timely fashion.
Further, cryonics (virtually by definition) is a procedure used only when the patient has exhausted all other available options. In current practice the patient is suspended after legal death: the fear that the treatment might prove worse than the disease is absent. Of course, suspension of the terminally ill patient somewhat before legal death has significant advantages. A patient suffering from a brain tumor might view suspension following the obliteration of his brain as significantly less desirable than suspension prior to such obliteration, even if the suspension occurred at a point in time when the patient was legally "alive."
In such a case, it is inappropriate to disregard or override the patient's own wishes. To quote the American College of Physicians Ethics Manual, "Each patient is a free agent entitled to full explanation and full decision-making authority with regard to his medical care. John Stuart Mill expressed it as: `Over himself, his own body and mind, the individual is sovereign.' The legal counterpart of patient autonomy is self-determination. Both principles deny legitimacy to paternalism by stating unequivocally that, in the last analysis, the patient determines what is right for him." "If the [terminally ill] patient is a mentally competent adult, he has the legal right to accept or refuse any form of treatment, and his wishes must be recognized and honored by his physician."
If clinical trials cannot provide us with an answer, are there any other methods of evaluating the proposal? Can we do more than say that (a) cryonic suspension can do no harm (in keeping with the Hippocratic oath), and (b) it has some difficult-to-define chance of doing good?
Cryonics will fail if:
The second failure criterion is considered in the later sections on technical issues, which discuss in more detail how future technologies might be applied to the repair of frozen tissue.
As the reader will readily appreciate, the following reviews will consider only the most salient points that are of the greatest importance in determining overall feasibility. They are necessarily too short to consider the topics in anything like full detail, but should provide sufficient information to give the reader an overview of the relevant issues. References to further reading are provided throughout [note 12].
While cooling tissue to around 0 degrees C creates a number of problems, the ability to cool mammals to this temperature or even slightly below (with no ice formation) using current methods followed by subsequent complete recovery[61, 62] shows that this problem can be controlled and is unlikely to cause information theoretic death. We will, therefore, ignore the problems caused by such cooling. This problem is discussed in  and elsewhere.
Further, some "freezing" damage in fact occurs upon re- warming. Current work supports this idea because the precise method used to re-warm tissue can strongly affect the success or failure of present experiments even when freezing conditions are identical[5, 6]. If we presume that future repair methods avoid the step of re-warming the tissue prior to analysis and instead analyze the tissue directly in the frozen state then this source of damage will be eliminated. Several current methods can be used to distinguish between damage that occurs during freezing and damage that occurs while thawing. At present, it seems likely that some damage occurs during each process. While significant damage does occur during slow freezing, it does not induce structural changes which obliterate the cell.
Fractures that occur below the glass transition temperature result in very little information loss. While dramatic, this damage is unlikely to cause or contribute to information theoretic death.
Extracellular ice formation causes an increase in the concentration of the extra-cellular solute, e.g., the chemicals in the extracellular liquid are increased in concentration by the decrease in available water. The immediate effect of this increased extracellular concentration is to draw water out of the cells by osmosis. Thus, freezing dehydrates cells.
Damage can be caused by the extracellular ice, by the increased concentration of solute, or by the reduced temperature itself. All three mechanisms can play a role under appropriate conditions.
The damage caused by extracellular ice formation depends largely on the fraction of the initial liquid volume that is converted to ice[6, 57]. (The initial liquid volume might include a significant amount of cryoprotectant as well as water). When the fraction of the liquid volume converted to ice is small, damage is often reversible even by current techniques. In many cases, conversion of significantly more than 40% of the liquid volume to ice is damaging[70, page 134; 71]. The brain is more resistant to such injury: conversion of up to 60% of the liquid volume in the brain to ice is associated with recovery of neuronal function[58, 62, 66, 82]. Storey and Storey said "If the cell volume falls below a critical minimum, then the bilayer of phospholipids in the membrane becomes so greatly compressed that its structure breaks down. Membrane transport functions cannot be maintained, and breaks in the membrane spill cell contents and provide a gate for ice to propagate into the cell. Most freeze-tolerant animals reach the critical minimum cell volume when about 65 percent of total body water is sequestered as ice.".
Appropriate treatment with cryoprotectants (in particular glycerol) prior to freezing will keep 40% or more of the liquid volume from being converted to ice even at liquid nitrogen temperatures.
Fahy has said "All of the postulated problems in cryobiology -- cell packing [omitted reference], channel size constraints [omitted reference], optimal cooling rate differences for mixed cell populations [omitted reference], osmotically mediated injury[omitted references], and the rest -- can be solved in principle by the selection of a sufficiently high concentration of cryoprotectant prior to freezing. In the extreme case, all ice formation could be suppressed completely by using a concentration of agent sufficient to ensure vitrification of the biological system in question [omitted reference]". Unfortunately, a concentration of cryoprotectant sufficiently high to protect the system from all freezing injury would itself be injurious. It should be possible to trade the mechanical injury caused by ice formation for the biochemical injury caused by the cryoprotectant, which is probably advantageous. Current suspension protocols at Alcor call for the introduction of greater than 6 molar glycerol. Both venous and arterial glycerol concentrations have exceeded 6 molar in several recent suspensions. If this concentration of cryoprotectant is also reaching the tissues, it should keep over 60% of the initial liquid volume from being converted to ice at liquid nitrogen temperatures [note 14].
Intracellular freezing is largely irrelevant to cryonic suspensions because of the slow freezing rates dictated by the large mass of tissue being frozen. Such freezing rates are too slow for intracellular freezing to occur except when membrane rupture allows extracellular ice to penetrate the intracellular region. If the membrane does fail, one would expect the interior of the cell to "flash."
Spontaneous recovery of function following freezing to liquid nitrogen temperatures using the best currently available techniques appears unlikely for mammalian organs, including the brain. Despite this, the level of structural preservation can be quite good. The complexity of the systems that have been successfully frozen and rewarmed is remarkable, and supports the claim that good structural preservation is often achieved. The mechanisms of damage that have been postulated in the literature are sufficiently subtle that information loss is likely to be small; that is, death by the information theoretic criterion is unlikely to have occurred. Further research aimed specifically at addressing this issue is needed.note 15] and future suspensions might eliminate delay entirely [note 16], delay is sometimes unavoidable [note 17]. The most significant type of damage that such delay causes is ischemic injury.
Broadly speaking, the structure of the human brain remains intact for several hours or more following the cessation of blood flow, or ischemia. The tissue changes that occur subsequent to ischemia have been well studied. There have also been studies of the "postmortem" changes that occur in tissue. Perhaps the most interesting of these studies was conducted by Kalimo et. al..
Effects of postmortem delay. Some brain functions are damaged irreversibly within minutes of the cessation of blood flow to the tissue. This led to the widespread belief that it would be impossible to isolate metabolically active and responsive preparations very long after death and use them to study neurotransmission. However, this is a misconception; many groups have successfully obtained functional preparations from normal (Table 1) [not present in this article] and pathological (Table 2) [not present in this article] human brain tissue from autopsies carried out up to 24 h or more postmortem. This is perhaps less surprising when the stability of enzymes, receptors, and nucleic acids is taken into consideration (see Hardy and Dodd, 1983 [reference 123 in this article]). With very few exceptions, the brain retains the metabolic machinery to reconstitute tissue metabolite and neurotransmitter pools. It also appears that sufficient structural integrity is retained to allow the various tissue compartments to remain relatively intact and distinct.In order to study immediate "postmortem" changes, Kalimo et. al. perfused the brains of 5 patients with aldehydes within half an hour of "clinical death". Subsequent examination of the preserved brain tissue with both light and electron microscopy showed the level of structural preservation. In two cases, the changes described were consistent with approximately one to two hours of ischemic injury. (Ischemic injury often begins prior to declaration of "clinical death", hence the apparently longer ischemic period compared with the interval following declaration of death and prior to perfusion of fixative). Physical preservation of cellular structure and ultrastructure was excellent. It is difficult to avoid the conclusion that information loss was negligible in these cases. In two other cases, elevated intraparenchymal pressure prevented perfusion with the preservative, thus preventing examination of the tissue. Without such an examination, it is difficult to draw conclusions about the extent of information loss. In the final case, "...the most obvious abnormality was the replacement of approximately four-fifths of the parenchyma of the brain by a fluid-containing cavity that was lined by what seemed to be very thin remnants of the cerebral cortex." Cryonic suspension in this last case would not be productive.
Experiments with both animal and human brain have shown that viable preparations can be isolated routinely up to at least 24 h postmortem, a time scale within which a sufficient number of autopsies is carried out to allow extensive neurochemical studies. When the human subject has died suddenly (see below) [not in this article], such preparations exhibit the same range of characteristics as preparations made from fresh animal tissue, or from fresh human tissue obtained at biopsy or neurosurgery. Thus incubated synaptosomes and brain slices from postmortem human brain respire, accumulate tissue potassium, maintain membrane potentials, release neurotransmitters in a calcium-dependent fashion, and possess active, sodium - dependent uptake systems (see Table 1 for references [not in this article]). Electron microscopic examination of synaptosome preparations from postmortem human brain showed them to be only slightly less pure than preparations from fresh tissue, although some degree of damage is evident (Hardy et al., 1982 [not in this article]).
As an aside, the vascular perfusion of chemical fixatives to improve stability of tissue structures prior to perfusion with cryoprotectants and subsequent storage in liquid nitrogen would seem to offer significant advantages. The main issue that would require resolution prior to such use is the risk that fixation might obstruct circulation, thus impeding subsequent perfusion with cryoprotectants. Other than this risk, the use of chemical fixatives (such as aldehydes and in particular glutaraldehyde) would reliably improve structural preservation and would be effective at halting almost all deterioration within minutes of perfusion. The utility of chemical preservation has been discussed by Drexler and by Olson, among others.
Ischemic changes do not appear to result in any damage that would prevent repair (e.g., changes that would result in significant loss of information about structure) for at least a few hours. Temporary functional recovery has been demonstrated in optimal situations after as long as 60 minutes of total ischemia[93, 94, 95]. Hossmann, for example, reported results on 143 cats subjected to one hour of normothermic global brain ischemia. "Body temperature was maintained at 36 degrees to 37 degrees C with a heating pad. ... Completeness of ischemia was tested by injecting 133Xe into the innominate artery immediately before vascular occlusion and monitoring the absence of decay of radioactivity from the head during ischemia, using external scintillation detectors. ... In 50% of the animals, even major spontaneous EEG activity returned after ischemia.... One cat survived for 1 yr after one hour of normothermic cerebrocirculatory arrest with no electrophysiologic deficit and with only minor neurologic and morphologic disturbances." Functional recovery is a more stringent criterion than the more relaxed information theoretic criterion, which merely requires adequate structural preservation to allow inference about the pre-existing structure. Reliable identification of the various cellular structures is possible hours (and sometimes even days) later. Detailed descriptions of ischemia and its time course[72, page 209 et sequitur] also clearly show that cooling substantially slows the rate of deterioration. Thus, even moderate cooling "postmortem" slows deterioration significantly.
Kalimo et. al. said "It is noteworthy that after 120 min of complete blood deprivation we saw no evidence of membrane lysosomal breakdown, an observation which has also been reported in studies of in vitro lethal cell injury[omitted references], and in regional cerebral ischemia[omitted references]."
Hawkins et. al. said "...lysosomes did not rupture for approximately 4 hours and in fact did not release the fluorescent dye until after reaching the postmortem necrotic phase of injury. ... The original suicide bag mechanism of cell damage thus is apparently not operative in the systems studied. Lysosomes appear to be relatively stable organelles...."
ADDENDUM: A 1997 paper critical of earlier work which attempted to recover DNA from ancient sources said "Whereas ancient DNA sequences from specimens younger than 100 000 years old have now been replicated independently (Hagelberg et al. 1994; Hoss et al. 1994; Taylor 1996), we have singularly failed to recover authentic ancient DNA from amber fossils."
For present purposes the distinction between 100,000 and 100,000,000 years is not critical: both are substantially longer than the time that a person might reasonably expect to stay in cryonic suspension.
As the ischemic interval lengthens, the level of damage increases. It is not clear exactly when information loss begins or when information theoretic death occurs. Present evidence supports but does not prove the hypothesis that information theoretic death does not occur for at least a few hours following the onset of ischemia. Quite possibly many hours of ischemia can be tolerated. Freezing of tissue within that time frame followed by long term storage in liquid nitrogen should provide adequate preservation of structure to allow repair [note 19].
It appears likely that short term memory, which can be disrupted by trauma or a number of other processes, will not be preserved by cryonic suspension. Consolidation of short term memory into long term memory is a process that takes several hours. We will focus attention exclusively on long term memory, for this is far more stable. While the retention of short term memory cannot be excluded (particularly if chemical preservation is used to provide rapid initial fixation), its greater fragility renders this significantly less likely.
To see the Mona Lisa or Niagara Falls changes us, as does seeing a favorite television show or reading a good book. These changes are both figurative and literal, and it is the literal (or neuroscientific) changes that we are interested in: what are the physical alterations that underlie memory?
Briefly, the available evidence supports the idea that memory and personality are stored in identifiable physical changes in the nerve cells, and that alterations in the synapses between nerve cells play a critical role.
Shepherd in "Neurobiology"[38, page 547] said: "The concept that brain functions are mediated by cell assemblies and neuronal circuits has become widely accepted, as will be obvious to the reader of this book, and most neurobiologists believe that plastic changes at synapses are the underlying mechanisms of learning and memory."
Kupfermann in "Principles of Neural Science"[13, page 1005] said: "Because of the enduring nature of memory, it seems reasonable to postulate that in some way the changes must be reflected in long-term alterations of the connections between neurons."
Eric R. Kandel in "Principles of Neural Science" [13, page 1016] said: "Morphological changes seem to be a signature of the long-term process. These changes do not occur with short-term memory (Figure 65-6 [not reproduced here]). Moreover, the structural changes that occur with the long- term process are not restricted to the [sic] growth. Long- term habituation leads to the opposite change---a regression and pruning of synaptic connections. With long-term habituation, where the functional connections between the sensory neurons and motor neurons are inactivated (Figure 65- 2[not reproduced]), the number of terminals per neuron is correspondingly reduced by one-third (Figure 65-6[not reproduced]) and the proportion of terminals with active zones is reduced from 40% to 10%."
Squire in "Memory and Brain"[109, page 10] said: "The most prevalent view has been that the specificity of stored information is determined by the location of synaptic changes in the nervous system and by the pattern of altered neuronal interactions that these changes produce. This idea is largely accepted at the present time, and will be explored further in this and succeeding chapters in the light of current evidence."
Lynch, in "Synapses, Circuits, and the Beginnings of Memory"[34, page 3] said: "The question of which components of the neuron are responsible for storage is vital to attempts to develop generalized hypotheses about how the brain encodes and makes use of memory. Since individual neurons receive and generate thousands of connections and hence participate in what must be a vast array of potential circuits, most theorists have postulated a central role for synaptic modifications in memory storage."
Turner and Greenough said "Two non-mutually exclusive possible mechanisms of brain information storage have remained the leading theories since their introduction by Ramon y Cajal [omitted reference] and Tanzi [omitted reference]. The first hypothesis is that new synapse formation, or selected synapse retention, yields altered brain circuitry which encodes new information. The second is that altered synaptic efficacy brings about similar change.".
Greenough and Bailey in "The anatomy of a memory: convergence of results across a diversity of tests" say: "More recently it has become clear that the arrangement of synaptic connections in the mature nervous system can undergo striking changes even during normal functioning. As the diversity of species and plastic processes subjected to morphological scrutiny has increased, convergence upon a set of structurally detectable phenomena has begun to emerge. Although several aspects of synaptic structure appear to change with experience, the most consistent potential substrate for memory storage during behavioral modification is an alteration in the number and/or pattern of synaptic connections."
It seems likely, therefore, that human long term memory is encoded by detectable physical changes in cell structure and in particular in synaptic structure.
"Using horseradish peroxidase (HRP) to label the presynaptic terminals (varicosities) of sensory neurons and serial reconstruction to analyze synaptic contacts, we compared the fine structure of identified sensory neuron synapses in control and behaviorally modified animals. Our results indicate that learning can modulate long-term synaptic effectiveness by altering the number, size, and vesical complement of synaptic active zones."
Examination by transmission electron microscopy in vacuum of sections 100 nanometers (several hundred atomic diameters) thick recovers little or no chemical information. Lateral resolution is at best a few nanometers (tens of atomic diameters), and depth information (within the 100 nanometer section) is entirely lost. Specimen preparation included removal and desheathing of the abdominal ganglion which was then bathed in seawater for 30 minutes before impalement and intrasomatic pressure injection of HRP. Two hours later the ganglia were fixed, histochemically processed, and embedded. Following this treatment, Bailey and Chen concluded that "...clear structural changes accompany behavioral modification, and those changes can be detected at the level of identified synapses that are critically involved in learning."
The following observations about this work seem in order. First, several different types of changes were present. This provides redundant evidence of synaptic alteration. Inability to detect one type of change, or obliteration of one specific type of change, would not be sufficient to prevent recovery of the "state" of the synapse. Second, examination by electron microscopy is much cruder than the techniques considered here which literally propose to analyze every molecule in the structure. Further alterations in synaptic chemistry will be detectable when the synapse is examined in more detail at the molecular level. Third, there is no reason to believe that freezing would obliterate the structure beyond recognition.
It seems likely that knowledge of the morphology and connectivity of nerve cells along with some specific knowledge of the biochemical state of the cells and synapses would be sufficient to determine memory and personality. Perhaps, however, some fundamentally different mechanism is present in humans? Even if this were to prove true, any such system would be sharply constrained by the available evidence. It would have to persist over the lifetime of a human being, and thus would have to be quite stable. It would have to tolerate the natural conditions encountered by humans and the experimental conditions to which primates have been subjected without loss of memory and personality (presuming that the primate brain is similar to the human brain). And finally, it would almost certainly involve changes in tens of thousands of molecules to store each bit of information. Functional studies of human long term memory suggest it has a capacity of only 109 bits (somewhat over 100 megabytes) (though this did not consider motor memory, e.g., the information storage required when learning to ride a bicycle). Such a low memory capacity suggests that, independent of the specific mechanism, a great many molecules are required to remember each bit. It even suggests that many synapses are used to store each bit (recall there are perhaps 1015 synapses -- which implies some 106 synapses per bit of information stored in long term memory).
Given that future technology will allow the molecule-by- molecule analysis of the structures that store memory, and given that such structures are large on the molecular scale (involving at least tens of thousands of molecules each) then it appears unlikely that such structures will survive the lifetime of the individual only to be obliterated beyond recognition by freezing. Freezing is unlikely to cause information theoretic death.
It is unreasonable to think that the current proposal will in fact form the basis for future repair methods for two reasons:
First, better technologies and approaches are likely to be developed. Necessarily, we must restrict ourselves to methods and techniques that can be analyzed and understood using the currently understood laws of physics and chemistry. Future scientific advances, not anticipated at this time, are likely to result in cheaper, simpler or more reliable methods. Given the history of science and technology to date, the probability of future unanticipated advances is good.
Second, this proposal was selected because of its conceptual simplicity and its obvious power to restore virtually any structure where restoration is in principle feasible. These are unlikely to be design objectives of future systems. Conceptual simplicity is advantageous when the resources available for the design process are limited. Future design capabilities can reasonably be expected to outstrip current capabilities, and the efforts of a large group can reasonably be expected to allow analysis of much more complex proposals than considered here.
Further, future systems will be designed to restore specific individuals suffering from specific types of damage, and can therefore use specific methods that are less general but which are more efficient or less costly for the particular type of damage involved. It is easier for a general-purpose proposal to rely on relatively simple and powerful methods, even if those methods are less efficient.
Why, then, discuss a powerful, general purpose method that is inefficient, fails to take advantage of the specific types of damage involved, and which will almost certainly be superseded by future technology?
The purpose of this paper is not to lay the groundwork for future systems, but to answer a question: will cryonics work? The value of cryonics is clearly and decisively based on technical capabilities that will not be developed for several decades (or longer). If some relatively simple proposal appears likely to work, then the value of cryonics is established. Whether or not that simple proposal is actually used is irrelevant. The fact that it could be used in the improbable case that all other technical progress and all other approaches fail is sufficient to let us decide today whether or not cryonic suspension is of value.
The philosophical issues involved in this type of long range technical forecasting and the methodologies appropriate to this area are addressed by work in "exploratory engineering."[1, 85] The purpose of exploratory engineering is to provide lower bounds on future technical capabilities based on currently understood scientific principles. A successful example is Konstantin Tsiolkovsky's forecast around the turn of the century that multi-staged rockets could go to the moon. His forecast was based on well understood principles of Newtonian mechanics. While it did not predict when such flights would take place, nor who would develop the technology, nor the details of the Saturn V booster, it did predict that the technical capability was feasible and would eventually be developed. In a similar spirit, we will discuss the technical capabilities that should be feasible and what those capabilities should make possible.
Conceptually, the approach that we will follow is simple:
An obvious inefficiency of this approach is that it will take apart and then put back together again structures and whole regions that are in fact functional or only slightly damaged. Simply leaving a functional region intact, or using relatively simple special case repair methods for minor damage would be faster and less costly. Despite these obvious drawbacks, the general purpose approach demonstrates the principles involved. As long as the inefficiencies are not so extreme that they make the approach infeasible or uneconomical in the long run, then this simpler approach is easier to evaluate.
We proceed in the same way for the lipids (lipids are most often used to make cell membranes) -- a "typical" lipid might have a molecular weight of 500 amu, which is 100 times less than the molecular weight of a protein. This implies the brain has about 175/500 x 6.02 x 1023 or about 2 x 1023 lipid molecules.
Finally, water has a molecular weight of 18, so there will be about 1400 x 0.8/18 x 6.02 x 1023 or about 4 x 1025 water molecules in the brain. In many cases a substantial percentage of water will have been replaced with cryoprotectant during the process of suspension; glycerol at a concentration of 4 molar or more, for example. Both water and glycerol will be treated in bulk, and so the change from water molecules to glycerol (or other cryoprotectants) should not have a significant impact on the calculations that follow.
These numbers are fundamental. Repair of the brain down to the molecular level will require that we cope with them in some fashion.
The time required for a ribosome to manufacture a protein molecule of 400 amino acids is about 10 seconds[14, page 393], or about 25 milliseconds to add each amino acid. DNA polymerase III can add an additional base to a replicating DNA strand in about 7 milliseconds[14, page 289]. In both cases, synthesis takes place in solution and involves significant delays while the needed components diffuse to the reactive sites. The speed of assembler-directed reactions is likely to prove faster than current biological systems. The arm of an assembler should be capable of making a complete motion and causing a single chemical transformation in about a microsecond . However, we will conservatively base our computations on the speed of synthesis already demonstrated by biological systems, and in particular on the slower speed of protein synthesis.
We must do more than synthesize the required molecules -- we must analyze the existing molecules, possibly repair them, and also move them from their original location to the desired final location. Existing antibodies can identify specific molecular species by selectively binding to them, so identifying individual molecules is feasible in principle. Even assuming that the actual technology employed is different it seems unlikely that such analysis will require substantially longer than the synthesis time involved, so it seems reasonable to multiply the synthesis time by a factor of a few to provide an estimate of time spent per molecule. This should, in principle, allow time for the complete disassembly and reassembly of the selected molecule using methods no faster than those employed in biological systems. While the precise size of this multiplicative factor can reasonably be debated, a factor of 10 should be sufficient. The total time required to simply move a molecule from its original location to its correct final location in the repaired structure should be smaller than the time required to disassemble and reassemble it, so we will assume that the total time required for analysis, repair and movement is 100 seconds per protein molecule.
Carried to its logical conclusion, we would discard and replace all the molecules in the structure. Having once determined the type, location and orientation of a molecule in the original (frozen) structure, we would simply throw that molecule out without further examination and replace it. This requires only that we be able to identify the location and type of individual molecules. It would not be necessary to determine if the molecule was damaged, nor would it be necessary to correct any damage found. By definition, the replacement molecule would be taken from a stock-pile of structurally correct molecules that had been previously synthesized, in bulk, by the simplest and most economical method available.
Discarding and replacing even a few atoms might disturb some people. This can be avoided by analyzing and repairing any damaged molecules. However, for those who view the simpler removal and replacement of damaged molecules as acceptable, the repair process can be significantly simplified. For purposes of this paper, however, we will continue to use the longer time estimate based on the premise that full repair of every molecule is required. This appears to be conservative. (Those who feel that replacing their atoms will change their identity should think carefully before eating their next meal!)
We have assumed that the time required to analyze and synthesize an individual molecule will dominate the time required to determine its present location, the time required to determine the appropriate location it should occupy in the repaired structure, and the time required to put it in this position. These assumptions are plausible but will be considered further when the methods of gaining access to and of moving molecules during the repair process are considered.
This analysis accounts for the bulk of the molecules -- it seems unlikely that other molecular species will add significant additional repair time.
Based on these assumptions, we find that we require 100 seconds x 1.2 x 1021 protein molecules + 1 second times 2 x 1023 lipids, or 3.2 x 1023 repair-machine-seconds. This number is not as fundamental as the number of molecules in the brain. It is based on the (probably conservative) assumption that repair of 50,000 amu requires 100 seconds. Faster repair would imply repair could be done with fewer repair machines, or in less time.
If the total repair time is 108 seconds, and we require 3.2 x 1023 repair-machine-seconds, then we require 3.2 x 1015 repair machines for complete repair of the brain. This corresponds to 3.2 x 1015 / (6.02 x 1023) or 5.3 x 10-9 moles, or 5.3 nanomoles of repair machines. If each repair device weighs 109 to 1010 amu, then the total weight of all the repair devices is 5.3 to 53 grams: a a few ounces at most.
Thus, the weight of repair devices required to repair each and every molecule in the brain, assuming the repair devices operate no faster than current biological methods, is about 0.4% to 4% of the total mass of the brain.
By way of comparison, there are about 1014 cells[44, page 3] in the human body and each cell has about 107 ribosomes[14, page 652] giving 1021 ribosomes. Thus, there are about six orders of magnitude more ribosomes in the human body than the number of repair machines we estimate are required to repair the human brain.
It seems unlikely that either more or larger repair devices are inherently required. However, it is comforting to know that errors in these estimates of even several orders of magnitude can be easily tolerated. A requirement for 530 kilograms of repair devices (10,000 to 100,000 times more than we calculate is needed) would have little practical impact on feasibility. Although repair scenarios that involve deployment of the repair devices within the volume of the brain could not be used if we required 530 kilograms of repair devices, a number of other repair scenarios would still work -- one such approach is discussed in this paper. Given that nanotechnology is feasible, manufacturing costs for repair devices will be small. The cost of even 530 kilograms of repair devices should eventually be significantly less than a few hundred dollars. The feasibility of repair down to the molecular level is insensitive to even large errors in the projections given here.
The first advantage of on-board repair is an easier evolutionary path from partial repair systems deployed in living human beings to the total repair systems required for repair of the more extensive damage found in the person who has been cryonically suspended. That is, a simple repair device for finding and removing fatty deposits blocking the circulatory system could be developed and deployed in living humans, and need not deal with all the problems involved in total repair. A more complex device, developed as an incremental improvement, might then repair more complex damage (perhaps identifying and killing cancer cells) again within a living human. Once developed, there will be continued pressure for evolutionary improvements in on-board repair capabilities which should ultimately lead to repair of virtually arbitrary damage. This evolutionary path should eventually produce a device capable of repairing frozen tissue.
It is interesting to note that "At the end of this month [August 1990], MITI's Agency of Industrial Science and Technology (AIST) will submit a budget request for ´30 million ($200,000) to launch a `microrobot' project next year, with the aim of developing tiny robots for the internal medical treatment and repair of human beings. ... MITI is planning to pour ´25,000 million ($170 million) into the microrobot project over the next ten years...". Iwao Fujimasa said their objective is a robot less than .04 inches in size that will be able to travel through veins and inside organs[17, 20]. While substantially larger than the proposals considered here, the direction of future evolutionary improvements should be clear.
A second advantage of on-board repair is emotional. In on- board repair, the original structure (you) is left intact at the macroscopic and even light microscopic level. The disassembly and reassembly of the component molecules is done at a level smaller than can be seen, and might therefore prove less troubling than other forms of repair in which the disassembly and reassembly processes are more visible. Ultimately, though, correct restoration of the structure is the overriding concern.
A third advantage of on-board repair is the ability to leave functional structures intact. That is, in on-board repair we can focus on those structures that are damaged, while leaving working structures alone. If minor damage has occurred, then an on-board repair system need make only minor repairs.
The major drawback of on-board repair is the increased complexity of the system. As discussed earlier, this is only a drawback when the design tools and the resources available for the design are limited. We can reasonably presume that future design tools and future resources will greatly exceed present efforts. Developments in computer aided design of complex systems will put the design of remarkably complex systems within easy grasp.
In on-board repair, we might first logically partition the volume of the brain into a matrix of cubes, and then deploy each repair device in its own cube. Repair devices would first get as close as possible to their assigned cube by moving through the circulatory system (we presume it would be cleared out as a first step) and would then disassemble the tissue between them and their destination. Once in position, each repair device would analyze the tissue in its assigned volume and perform any repairs required.
The primary advantage of off-board repair is conceptual simplicity. It employees simple brute force to insure that a solution is feasible and to avoid complex design issues. As discussed earlier, these are virtues in thinking about the problem today but are unlikely to carry much weight in the future when an actual system is being designed.
The other advantages of this approach are fairly obvious. Lingering concerns about volume and heat dissipation can be eliminated. If a ton of repair devices should prove necessary, then a ton can be provided. Concerns about design complexity can be greatly reduced. Off-board repair scenarios do not require that the repair devices be mobile -- simplifying communications and power distribution, and eliminating the need for locomotor capabilities and navigational abilities. The only previous paper on off-board repair scenarios was by Merkle.
Off-board repair scenarios can be naturally divided into three phases. In the first phase, we must analyze the structure to determine its state. The primary purpose of this phase is simply to gather information about the structure, although in the process the disassembly of the structure into its component molecules will also take place. Various methods of gaining access to and analyzing the overall structure are feasible -- in this paper we shall primarily consider one approach.
We shall presume that the analysis phase takes place while the tissue is still frozen. While the exact temperature is left open, it seems preferable to perform analysis prior to warming. The thawing process itself causes damage and, once thawed, continued deterioration will proceed unchecked by the mechanisms present in healthy tissue. This cannot be tolerated during a repair time of several years. Either faster analysis or some means of blocking deterioration would have to be used if analysis were to take place after warming. We will not explore these possibilities here (although this appears worthwhile). The temperature at which other phases takes place is left open.
The second phase of off-board repair is determination of the healthy state. In this phase, the structural information derived from the analysis phase is used to determine what the healthy state of the tissue had been prior to suspension and any preceding illness. This phase involves only computation based on the information provided by the analysis phase.
The third phase is repair. In this phase, we must restore the structure in accordance with the blue-print provided by the second phase, the determination of the healthy state.
The first state is the starting state, prior to any repair efforts. The tissue is frozen (unrepaired).
In the second state, immediately following the analysis phase, the tissue has been disassembled into its individual molecules. A detailed structural data base has been built which provides a description of the location, orientation, and type of each molecule, as discussed earlier. For those who are concerned that their identity or "self" is dependent in some fundamental way on the specific atoms which compose their molecules, the original molecules can be retained in a molecular "filing cabinet." While keeping physical track of the original molecules is more difficult technically, it is feasible and does not alter off-board repair in any fundamental fashion.
In the third state, the tissue is restored and fully functional.
By characterizing the intermediate state which must be achieved during the repair process, we reduce the problem from "Start with frozen tissue and generate healthy tissue" to "Start with frozen tissue and generate a structural data base and a molecular filing cabinet. Take the structural data base and the molecular filing cabinet and generate healthy tissue." It is characteristic of off-board repair that we disassemble the molecular structure into its component pieces prior to attempting repair.
As an example, suppose we wish to repair a car. Rather than try and diagnose exactly what's wrong, we decide to take the car apart into its component pieces. Once the pieces are spread out in front of us, we can easily clean each piece, and then reassemble the car. Of course, we'll have to keep track of where all the pieces go so we can reassemble the structure, but in exchange for this bookkeeping task we gain a conceptually simple method of insuring that we actually can get access to everything and repair it. While this is a rather extreme method of repairing a broken carburetor, it certainly is a good argument that we should be able to repair even rather badly damaged cars. So, too, with off-board repair. While it might be an extreme method of fixing any particular form of damage, it provides a good argument that damage can be repaired under a wide range of circumstances.
One particular approach to off-board repair is divide-and- conquer. This method is one of the technically simplest approaches. We discuss this method in the following section.
If we apply divide-and-conquer to the analysis of a physical object -- such as the brain -- then we must be able to physically divide the object of analysis into two pieces and recursively apply the same method to the two pieces. This means that we must be able to divide a piece of frozen tissue, whether it be the entire brain or some smaller part, into roughly equal halves. Given that tissue at liquid nitrogen temperatures is already prone to fracturing, it should require only modest effort to deliberately induce a fracture that would divide such a piece into two roughly equal parts. Fractures made at low temperatures (when the material is below the glass transition temperature) are extremely clean, and result in little or no loss of structural information. Indeed, freeze fracture techniques are used for the study of synaptic structures. Hayat [40, page 398] says "Membranes split during freeze-fracturing along their central hydrophobic plane, exposing intramembranous surfaces. ... The fracture plane often follows the contours of membranes and leaves bumps or depressions where it passes around vesicles and other cell organelles. ... The fracturing process provides more accurate insight into the molecular architecture of membranes than any other ultrastructural method." It seems unlikely that the fracture itself will result in any significant loss of structural information.
The freshly exposed faces can now be analyzed by various surface analysis techniques. Work with STMs supports the idea that very high resolution is feasible. For example, optical absorption microscopy "...generates an absorption spectrum of the surface with a resolution of 1 nanometer [a few atomic diameters]." Kumar Wickramasinghe of IBM's T. J. Watson Research Center said: "We should be able to record the spectrum of a single molecule" on a surface. Williams and Wickramasinghe said  "The ability to measure variations in chemical potential also allows the possibility of selectively identifying subunits of biological macromolecules either through a direct measurement of their chemical- potential gradients or by decorating them with different metals. This suggest a potentially simple method for sequencing DNA." While current devices are large, the fundamental physical principles on which they rely do not require large size. Many of the devices depend primarily on the interaction between a single atom at the tip of the STM probe and the atoms on the surface of the specimen under analysis. Clearly, substantial reductions in size in such devices are feasible.
High resolution optical techniques can also be employed. Near field microscopy, employing light with a wavelength of hundreds of nanometers, has achieved a resolution of 12 nanometers (much smaller than a wavelength of light). To quote the abstract of a recent review article on the subject: "The near-field optical interaction between a sharp probe and a sample of interest can be exploited to image, spectroscopically probe, or modify surfaces at a resolution (down to ~12 nm) inaccessible by traditional far-field techniques. Many of the attractive features of conventional optics are retained, including noninvasiveness, reliability, and low cost. In addition, most optical contrast mechanisms can be extended to the near-field regime, resulting in a technique of considerable versatility. This versatility is demonstrated by several examples, such as the imaging of nanometric-scale features in mammalian tissue sections and the creation of ultrasmall, magneto-optic domains having implications for high-density data storage. Although the technique may find uses in many diverse fields, two of the most exciting possibilities are localized optical spectroscopy of semiconductors and the fluorescence imaging of living cells.". Another article said: "Our signals are currently of such magnitude that almost any application originally conceived for far-field optics can now be extended to the near-field regime, including: dynamical studies at video rates and beyond; low noise, high resolution spectroscopy (also aided by the negligible auto-fluorescence of the probe); minute differential absorption measurements; magnetooptics; and superresolution lithography." [note 20] .
One might view these cubes as the pieces of a three- dimensional jig-saw puzzle, the only difference being that we have cheated and carefully recorded the position of each piece. Just as the picture on a jig-saw puzzle is clearly visible despite the fractures between the pieces, so too the three-dimensional "picture" of the brain is clearly visible despite its division into pieces [note 21].
If it requires about three years to analyze each piece, then the time required both to divide the brain into pieces and to move each piece to an immobile repair device can reasonably be neglected. It seems unlikely that moving the pieces will take a significant fraction of three years.
A simple method of reducing storage requirements by several orders of magnitude would be to analyze and repair only a small amount of tissue at a time. This would eliminate the need to store the entire 1025 bit description at one time. A smaller memory could hold the description of the tissue actually under repair, and this smaller memory could then be cleared and re-used during repair of the next section of tissue.
Energy costs appear to be the limiting factor in rod logic (rather than the number of gates, or the speed of operation of the gates). Today, electric power costs about 10 cents per kilowatt hour. Future costs of power will almost certainly be much lower. Molecular manufacturing should eventually sharply reduce the cost of solar cells and increase their efficiency close to the theoretical limits. With a manufacturing cost of under 10 cents per kilogram[ 85] the cost of a one square meter solar cell will be less than a penny. As a consequence the cost of solar power will be dominated by other costs, such as the cost of the land on which the solar cell is placed. While solar cells can be placed on the roofs of existing structures or in otherwise unused areas, we will simply use existing real estate prices to estimate costs. Low cost land in the desert south western United States can be purchased for less than $1,000 per acre. (This price corresponds to about 25 cents per square meter, significantly larger than the projected future manufacturing cost of a one square meter solar cell). Land elsewhere in the world (arid regions of the Australian outback, for example) is much cheaper. For simplicity and conservatism, though, we'll simply adopt the $1,000 per acre price for the following calculations. Renting an acre of land for a year at an annual price of 10% of the purchase price will cost $100. Incident sunlight at the earth's surface provides a maximum of 1,353 watts per square meter, or 5.5 x 106 watts per acre. Making allowances for inefficiencies in the solar cells, atmospheric losses, and losses caused by the angle of incidence of the incoming light reduces the actual average power production by perhaps a factor of 15 to about 3.5 x 105 watts. Over a year, this produces 1.1 x 1013 joules or 3.1 x 106 kilowatt hours. The land cost $100, so the cost per joule is 0.9 nanocents and the cost per kilowatt hour is 3.3 millicents. Solar power, once we can make the solar cells cheaply enough, will be over several thousand times cheaper than electric power is today. We'll be able to buy over 1015 joules for under $10,000.
While the energy dissipation per logic operation estimated by Drexler[ 85] is about 10-23 joules, we'll content ourselves with the higher estimate of 10-22 joules per logic operation. Our 1015 joules will then power 1037 gate operations: 1012 gate operations for each bit in the structural data base or 5 x 1013 gate operations for each of the 2 x 1023 lipid molecules present in the brain.
It should be emphasized that in off-board repair warming of the tissue is not an issue because the overwhelming bulk of the calculations and hence almost all of the energy dissipation takes place outside the tissue. Much of the computation takes place when the original structure has been entirely disassembled into its component molecules.
To give a feeling for the computational power this represents, it is useful to compare it to estimates of the raw computational power of the human brain. The human brain has been variously estimated as being able to do 1013, 1015 or 1016 operations a second (where "operation" has been variously defined but represents some relatively simple and basic action) [note 23]. The 1037 total logic operations will support 1029 logic operations per second for three years, which is the raw computational power of something like 1013 human beings (even when we use the high end of the range for the computational power of the human brain). This is 10 trillion human beings, or some 2,000 times more people than currently exist on the earth today. By present standards, this is a large amount of computational power. Viewed another way, if we were to divide the human brain into tiny cubes that were about 5 microns on a side (less than the volume of a typical cell), each such cube could receive the full and undivided attention of a dedicated human analyst for a full three years.
The next paragraph analyzes memory costs, and can be skipped without loss of continuity.
This analysis neglects the memory required to store the complete state of these computations. Because this estimate of computational abilities and requirements depends on the capabilities of the human brain, we might require an amount of memory roughly similar to the amount of memory required by the human brain as it computes. This might require about 1016 bits (10 bits per synapse) to store the "state" of the computation. (We assume that an exact representation of each synapse will not be necessary in providing capabilities that are similar to those of the human brain. At worst, the behavior of small groups of cells could be analyzed and implemented by the most efficient method, e.g., a "center surround" operation in the retina could be implemented as efficiently as possible, and would not require detailed modeling of each neuron and synapse. In point of fact, it is likely that algorithms that are significantly different from the algorithms employed in the human brain will prove to be the most efficient for this rather specialized type of analysis, and so our use of estimates derived from low-level parts-counts from the human brain are likely to be conservative). For 1013 programs each equivalent in analytical skills to a single human being, this would require 1029 bits. At 100 cubic nanometers per bit, this gives 10,000 cubic meters. Using the cost estimates provided by Drexler this would be an uncomfortable $1,000,000. We can, however, easily reduce this cost by partitioning the computation to reduce memory requirements. Instead of having 1013 programs each able to "think" at about the same speed as a human being, we could have 1010 programs each able to "think" at a speed 1,000 times faster than a human being. Instead of having 10 trillion dedicated human analysts working for 3 years each, we would have 10 billion dedicated human analysts working for 3,000 virtual years each. The project would still be completed in 3 calendar years, for each computer "analyst" would be a computer program running 1,000 times faster than an equally skilled human analyst. Instead of analyzing the entire brain at once, we would instead logically divide the brain into 1,000 pieces each of about 1.4 cubic centimeters in size, and analyze each such piece fully before moving on to the next piece.
This reduces our memory requirements by a factor of 1,000 and the cost of that memory to a manageable $1,000.
It should be emphasized that the comparisons with human capabilities are used only to illustrate the immense capabilities of 1037 logic operations. It should not be assumed that the software that will actually be used will have any resemblance to the behavior of the human brain.
Energy loss in rod logic, in Likharev's parametric quantron, in properly designed NMOS and CMOS circuits, and in many other proposals for computational devices is related to speed of operation. By slowing down the operating speed from 100 picoseconds to 100 nanoseconds or even 100 microseconds we should achieve corresponding reductions in energy dissipation per gate operation. This will allow substantial increases in computational power for a fixed amount of energy (1015 joules). We can both decrease the energy dissipated per gate operation (by operating at a slower speed) and increase the total number of gate operations (by using more gates). Because the gates are very small to start with, increasing their number by a factor of as much as 1010 (to approximately 1027 gates) would still result in a total volume of 100 cubic meters (recall that each gate plus overhead is about 100 cubic nanometers). This is a cube less than 5 meters on a side. Given that manufacturing costs will eventually reflect primarily material and energy costs, such a volume of slowly operating gates should be economical and would deliver substantially more computational power per joule.
We will not pursue this approach here for two main reasons. First, published analyses use the higher 100 picosecond speed of operation and 10-22 joules of energy dissipation[ 85]. Second, operating at 10-22 joules at room temperature implies that most logic operations must be reversible and that less than one logic operation in 30 can be irreversible. Irreversible logic operations (which erase information) must inherently dissipate at least kT x ln(2) for fundamental thermodynamic reasons. The average thermal energy of a single atom or molecule at a temperature T (measured in degrees K) is approximately kT where k is Boltzmann's constant. At room temperature, kT is about 4 x 10-21 joules. Thus, each irreversible operation will dissipate almost 3 x 10-21 joules. The number of such operations must be limited if we are to achieve an average energy dissipation of 10-22 joules per logic operation.
While it should be feasible to perform computations in which virtually all logic operations are reversible (and hence need not dissipate any fixed amount of energy per logic operation)[9, 25, 32, 53, 112, 120], current computer architectures might require some modification before they could be adapted to this style of operation. By contrast, it should be feasible to use current computer architectures while at the same time performing a major percentage (e.g., 99% or more) of their logic operations in a reversible fashion.
Various electronic proposals show that almost all of the existing combinational logic in present computers can be replaced with reversible logic with no change in the instruction set that is executed[112, 113]. Further, while some instructions in current computers are irreversible and hence must dissipate at least kT x ln(2) joules for each bit of information erased, other instructions are reversible and need not dissipate any fixed amount of energy if implemented correctly. Optimizing compilers could then avoid using the irreversible machine instructions and favor the use of the reversible instructions. Thus, without modifying the instruction set of the computer, we can make most logic operations in the computer reversible.
Further work on reversible computation can only lower the minimum energy expenditure per basic operation and increase the percentage of reversible logic operations. Much greater reductions in energy dissipation might be feasible. While it is at present unclear how far the trend towards lower energy dissipation per logic operation can go, it is clear that we have not yet reached a limit and that no particular limit is yet visible.
We can also expect further decreases in energy costs. By placing solar cells in space the total incident sunlight per square meter can be greatly increased (particularly if the solar cell is located closer to the sun) while at the same time the total mass of the solar cell can be greatly decreased. Most of the mass in earth-bound structures is required not for functional reasons but simply to insure structural integrity against the forces of gravity and the weather. In space both these problems are virtually eliminated. As a consequence a very thin solar cell of relatively modest mass can have a huge surface area and provide immense power at much lower costs than estimated here.
If we allow for the decreasing future cost of energy and the probability that future designs will have lower energy dissipation than 10-22 joules per logic operation, it seems likely that we will have a great deal more computational power than required. Even ignoring these more than likely developments, we will have adequate computational power for repair of the brain down to the molecular level.
While the revised data base describes the healthy state of the tissue that we desire to achieve, it does not specify the method(s) to be used in restoring the healthy structure. There is in general no necessary implication that restoration will or will not be done at some specific temperature, or will or will not be done in any particular fashion. Any one of a wide variety of methods could be employed to actually restore the specified structure. Further, the actual restored structure might differ in minor details from the structure described by the revised data base.
The complexity of the program that determines the healthy state will vary with
the quality of the suspension and the level of damage prior to suspension. Clearly,
if cryonic suspension "almost works", then the initial data base and the revised
data base will not greatly differ. Cryonic suspension under favorable circumstances
preserves the tissue with good fidelity down to the molecular level. If, however,
there was significant presuspension injury then deducing the correct (healthy)
structural description is more complex. However, it should be feasible to deduce
the correct structural description even in the face of significant damage. Only
if the structure is obliterated beyond recognition will it be infeasible to
deduce the undamaged state of the structure.
It seems likely that many people today will want complete physical
restoration (despite the philosophical possibilities considered above) and will
continue through the repair planning and repair phases.
We know it is possible to make a human brain for this has been done by traditional methods for many thousands of years. If we were to adopt a restoration method that was as close as possible to the traditional technique for building a brain, we might use a "guided growth" strategy. That is, in simple organisms the growth of every single cell and of every single synapse is determined genetically. "All the cell divisions, deaths, and migrations that generate the embryonic, then the larval, and finally the adult forms of the roundworm Caenorhabditis Elegans have now been traced.". "The embryonic lineage is highly invariant, as are the fates of the cells to which it gives rise". The appendix says: "Parts List: Caenorhabditis elegans (Bristol) Newly Hatched Larva. This index was prepared by condensing a list of all cells in the adult animal, then adding comments and references. A complete listing is available on request..." The adult organism has 959 cells in its body, 302 of which are nerve cells.
Restoring a specific biological structure using this approach would require that we determine the total number and precise growth patterns of all the cells involved. The human brain has roughly 1012 nerve cells, plus perhaps ten times as many glial cells and other support cells. While simply encoding this complex a structure into the genome of a single embryo might prove to be overly complex, it would certainly be feasible to control critical cellular activities by the use of on board nanocomputers. That is, each cell would be controlled by an on-board computer, and that computer would in turn have been programmed with a detailed description of the growth pattern and connections of that particular cell. While the cell would function normally in most respects, critical cellular activities, such as replication, motility, and synapse growth, would be under the direct control of the on-board computer. Thus, as in C. Elegans but on a larger scale, the growth of the entire system would be "highly invariant." Once the correct final configuration had been achieved, the on-board nanocomputers would terminate their activities and be flushed from the system as waste.
This approach might be criticized on the grounds that the resulting person was a "mere duplicate," and so "self" had not been preserved. Certainly, precise atomic control of the structure would appear to be difficult to achieve using guided growth, for biological systems do not normally control the precise placement of individual molecules. While the same atoms could be used as in the original, it would seem difficult to guarantee that they would be in the same places.
Concerns of this sort lead to restoration methods that provide higher precision. In these methods, the desired structure is restored directly from molecular components by placing the molecular components in the desired locations. A problem with this approach is the stability of the structure during restoration. Molecules might drift away from their assigned locations, destroying the structure.
An approach that we might call "minimal stabilization" would involve synthesis in liquid water, with mechanical stabilization of the various lipid membranes in the system. A three-dimensional grid or scaffolding would provide a framework that would hold membrane anchors in precise locations. The membranes themselves would thus be prevented from drifting too far from their assigned locations. To prevent chemical deterioration during restoration, it would be necessary to remove all reactive compounds (e.g., oxygen).
In this scenario, once the initial membrane "framework" was in place and held in place by the scaffolding, further molecules would be brought into the structure and put in the correct locations. In many instances, such molecules could be allowed to diffuse freely within the cellular compartment into which they had been introduced. In some instances, further control would be necessary. For example, a membrane- spanning channel protein might have to be confined to a specific region of a nerve cell membrane, and prevented from diffusing freely to other regions of the membrane. One method of achieving this limited kind of control over further diffusion would be to enclose a region of the membrane by a diffusion barrier (much like the spread of oil on water can be prevented by placing a floating barrier on the water).
While it is likely that some further cases would arise where it was necessary to prevent or control diffusion, the emphasis in this method is in providing the minimal control over molecular position that is needed to restore the structure.
While this approach does not achieve atomically precise restoration of the original structure, the kinds of changes that are introduced (diffusion of a molecule within a cellular compartment, diffusion of a membrane protein within the membrane) would be very similar to the kinds of diffusion that would take place in a normal biological system. Thus, the restored result would have the same molecules with the same atoms, and the molecules would be in similar (though not exactly the same) locations they had been in prior to restoration.
To achieve even more precise control over the restored structure, we might adopt a "full stabilization" strategy. In this strategy, each major molecule would be anchored in place, either to the scaffolding or an adjacent molecule. This would require the design of a stabilizing molecule for each specific type of molecule found in the body. The stabilizing molecule would have a specific end attached to the specific molecule, and a general end attached either to the scaffolding or to another stabilizing molecule. Once restoration was complete, the stabilizing molecules would release the molecules that were being stabilized and normal function would resume. This release might be triggered by the simple diffusion of an enzyme that attacked and broke down the stabilizing molecules. This kind of approach was considered by Drexler.
Finally, we might achieve stability of the intermediate structure by using low temperatures. If the structure were restored at a sufficiently low temperature, a molecule put in a certain place would simply not move. We might call this method "low temperature restoration."
In this scenario, each new molecule would simply be stacked (at low temperature) in the right location. This can be roughly likened to stacking bricks to build a house. A hemoglobin molecule could simply be thrown into the middle of the half-restored red blood cell. Other molecules whose precise position was not critical could likewise be positioned rather inexactly. Lipids in the lipid bi-layer forming the cellular membrane would have to be placed more precisely (probably with an accuracy of several angstroms). An individual lipid molecule, having once been positioned more or less correctly on a lipid bi-layer under construction, would be held in place (at sufficiently low temperatures) by van der Waals forces. Membrane bound proteins could also be "stacked" in their proper locations. Because biological systems make extensive use of self- assembly it would not be necessary to achieve perfect accuracy in the restoration process. If a biological macromolecule is positioned with reasonable accuracy, it would automatically assume the correct position upon warming.
Large polymers, used either for structural or other purposes, pose special problems. The monomeric units are covalently bonded to each other, and so simple "stacking" is inadequate. If such polymers cannot be added to the structure as entirely pre-formed units, then they could be incrementally restored during assembly from their individual monomers using the techniques discussed earlier involving positional synthesis using highly reactive intermediates. Addition of monomeric units to the polymer could then be done at the most convenient point during the restoration operation.
The chemical operations required to make a polymer from its monomeric units at reduced temperatures are unlikely to use the same reaction pathways that are used by living systems. In particular, the activation energies of most reactions that take place at 310 K (98.6 degrees Fahrenheit) can not be met at 77 K: most conventional compounds don't react at that temperature. However, as discussed earlier, assembler based synthesis techniques using highly reactive intermediates in near-perfect vacuum with mechanical force providing activation energy will continue to work quite well, even if we assume that thermal activation energy is entirely absent (e.g., that the system is close to 0 Kelvins).
An obvious problem with low temperature restoration is the need to re-warm the structure without incurring further damage. Much "freezing" injury takes place during rewarming, and this would have to be prevented. One solution is discussed in the next two paragraphs.
Generally, the revised structural data base can be further altered to make restoration easier. While certain alterations to the structural data base must be banned (anything that might damage memory, for example) many alterations would be quite safe. One set of safe alterations would be those that correspond to real-world changes that are non-damaging. For example, moving sub-cellular organelles within a cell would be safe -- such motion occurs spontaneously in living tissue. Likewise, small changes in the precise physical location of cell structures that did not alter cellular topology would also be safe. Indeed, some operations that might at first appear dubious are almost certainly safe. For example, any alteration that produces damage that can be repaired by the tissue itself once it is restored to a functional state is in fact safe -- though we might well seek to avoid such alterations (and they do not appear necessary). While the exact range of alterations that can be safely applied to the structural data base is unclear, it is evident that the range is fairly wide.
An obvious modification which would allow us to re-warm the structure safely would be to add cryoprotectants. Because we are restoring the frozen structure with atomic precision, we could use different concentrations and different types of cryoprotectants in different regions, thus matching the cryoprotectant requirements with exquisite accuracy to the tissue type. This is not feasible with present technology because cryoprotectants are introduced using simple diffusive techniques.
Extremely precise control over the heating rate would also be feasible, as well as very rapid heating. Rapid heating would allow less time for damage to take place. Rapid heating, however, might introduce problems of stress and resulting fractures. Two approaches for the elimination of this problem are (1) modify the structure so that the coefficient of thermal expansion is very small and (2) increase the strength of the structure.
One simple method of insuring that the volume occupied before and after warming was the same (i.e., of making a material with a very small thermal expansion coefficient) would be to disperse many small regions with the opposite thermal expansion tendency throughout the material. For example, if a volume tended to expand upon warming the initial structure could include "nanovacuoles," or regions of about a nanometer in diameter which were empty. Such regions would be stable at low temperatures but would collapse upon warming. By finely dispersing such nanovacuoles it would be possible to eliminate any tendency of even small regions to expand on heating. Most materials expand upon warming, a tendency which can be countered by the use of nanovacuoles.
Of course, ice has a smaller volume after it melts. The introduction of nanovacuoles would only exacerbate its tendency to shrink upon melting. In this case we could use vitrified H20 rather than the usual crystalline variety. H20 in the vitreous state is disordered (as in the liquid state) even at low temperatures, and has a lower volume than crystalline ice. This eliminates and even reverses its tendency to contract on warming. Vitrified water at low temperature is denser than liquid water at room temperature.
Increasing the strength of the material can be done in any of a variety of ways. A simple method would be to introduce long polymers in the frozen structure. Proteins are one class of strong polymers that could be incorporated into the structure with minimal tissue compatibility concerns. Any potential fracture plane would be criss-crossed by the newly added structural protein, and so fractures would be prevented. By also including an enzyme to degrade this artificially introduced structural protein, it would be automatically and spontaneously digested immediately after warming. Very large increases in strength could be achieved by this method.
By combining (1) rapid, highly controlled heating; (2) atomically precise introduction of cryoprotectants; (3) the controlled addition of small nanovacuoles and regions of vitrified H20 to reduce or eliminate thermal expansion and contraction; and (4) the addition of structural proteins to protect against any remaining thermally induced stresses; the damage that might otherwise occur during rewarming should be completely avoidable.
The proposal that all four methods be used in combination is open to the valid
criticism that simpler approaches are likely to suffice, e.g., precise distribution
of cryoprotectants coupled with relatively slow warming. The proposals
advanced in this paper should not be taken as predictions about what will happen
or what will be necessary, but as arguments that certain capabilities will be
feasible. A belt is sufficient to hold up a pair of pants. When explaining that
it is possible to hold up a pair of pants to someone who has never seen or heard
of clothing, it is forgiveable to point out that the combined use of both belt
and suspenders should be effective in preventing the pants from falling down.
Off-board repair consists of three major steps: (1) Determine the coordinates and orientation of each major molecule. (2) Determine a set of appropriate coordinates in the repaired structure for each major molecule. (3) Move them from the former location to the latter. The various technical problems involved are likely to be met by future advances in technology. Because storage times in liquid nitrogen literally extend for several centuries, the development time of these technologies is not critical.
A broad range of technical approaches to this problem are feasible. The particular form of off-board repair that uses divide-and-conquer requires only that (1) tissue can be divided by some means (such as fracturing) which does not itself cause significant loss of structural information; (2) the pieces into which the tissue is divided can be moved to appropriate destinations (for further division or for direct analysis); (3) a sufficiently small piece of tissue can be analyzed; (4) a program capable of determining the healthy state of tissue given the unhealthy state is feasible; (5) that sufficient computational resources for execution of this program in a reasonable time frame are available; and (6) that restoration of the original structure given a detailed description of that structure is feasible.
It is impossible to conclude based on present evidence that either failure criterion is likely to be met.
Further study of cryonics by the technical community is needed. At present,
there is a remarkable paucity of technical papers on the subject [note 24]. As should be evident from this paper
multidisciplinary analysis is essential in evaluating its feasibility, for specialists
in any single discipline have a background which is too narrow to encompass
the whole. Given the life-saving nature of cryonics, it would be tragic if it
were to prove feasible but was little used.
It is the authors pleasant duty to acknowledge the many people who have commented on or encouraged the work on this paper as it evolved. The reviewers were not selected because of their opinions about cryonics: some support it, some don't, and some reserve final judgment. While the quality of the result could not have been achieved without their help, the author must accept responsibility for any errors in the final version. The author would like to thank: Dave Biegelsen, Arthur C. Clarke, Mike Darwin, Thomas Donaldson, Eric Drexler, Greg Fahy, Steve Harris, Leonard Hayflick, Hugh Hixon, Peter Mazur, Mark Miller, David Pegg, Chris Peterson, Ed Regis, Paul Segall, Len Shar, Irwin Sobel, Jim Southard, Jim Stevens and Leonard Zubkoff.
2) There is no implication here that the most powerful repair method either will (or will not) be used or be necessary. The fact that we can kill a gnat with a double-barreled shotgun does not imply that a fly-swatter won't work just as well. If we aren't certain whether we face a gnat or a tiger, we'd rather be holding the shotgun than the fly- swatter. The shotgun will work in either case, but the fly-swatter can't deal with the tiger. In a similar vein, we will consider the most powerful methods that should be feasible rather than the minimal methods that might be sufficient. While this approach can reasonably be criticized on the grounds that simpler methods are likely to work, it avoids the complexities and problems that must be dealt with in trying to determine exactly what those simpler methods might be in any particular case and provides additional margin for error.
3) An atomic mass unit is the same as a Dalton. Different authors in different fields have different preferences for the name used to describe this unit, and so no single abbreviation will satisfy everyone. The use in this paper of the atomic mass unit, abbreviated as amu, was a compromise intended to be most easily understood by the widest audience.
4) A wide variety of mechanical computer designs are feasible. Perhaps the most famous proposal for a mechanical computer was made by Charles Babbage in the early to mid 1800's. Mechanical systems can be scaled down to the molecular size range and still function, although the analysis of such molecular mechanical systems requires the use of (appropriately enough) molecular mechanics: a thriving field which models molecular behavior by the use of force fields to describe the forces acting on the individual nuclei. The time evolution of the locations of the nuclei can be followed using relatively straightforward computational methods.
5) To fully specify the state of each atom would, strictly speaking, require that we specify the states of all its electrons. For the most part, however, these states are known or can be readily inferred once the type of atom is given. For example, a sodium atom in solution will normally be the ion, Na+. Likewise, the bonding structure of two carbon atoms separated by a certain distance can normally be inferred from the distance. The state of magnetization, while relevant for computers (the state of magnetization of a floppy disk is obviously of importance) is of negligible importance in biological systems. People are routinely exposed to magnetic fields of several Tesla to make diagnostic images, and appear none the worse for the experience. While coordinate information should be sufficient in almost all cases, we can always add a few bits of additional information if there is some ambiguity. This won't increase our estimate of 100 bits per atom by very much, and because 100 bits is a conveniently round number we'll continue to use it.
6) Because proteins are always produced as a linear chain, they must of necessity be able to adopt an appropriate three dimensional configuration by themselves. Usually, the correct configuration is unique. If it isn't, it is usually the case that the molecule will spontaneously cycle through appropriate configurations by itself, e.g., an ion channel will open and close at appropriate times regardless of whether it was initially started in the "open" or "closed" configuration. If any remaining cases should prove to be a problem, a few additional bits can be used to describe the specific configuration desired.
7) "For many years, it was thought that irreversible cellular damage unavoidably occurs after only a few minutes of complete cerebral ischemia. This opinion has been modified during the past decade [omitted reference]. Provided that the conditions for recovery are optimal, short-term restoration of brain functions may be achieved after periods of ischemia lasting as long as 60 minutes...". "Most clinical and experimental studies suggest that the normothermic brain is not able to withstand complete ischemia of >8 to 10 min. There is, however, firm experimental evidence of functional and biochemical recovery of a substantial part of the brain after complete cerebrocirculatory arrest of one hour [omitted references].". "It turned out in fact that appropriate treatment of post-ischemic recirculation disturbances led to recovery of energy metabolism and neuronal excitability after complete cerebro-circulatory arrest of as long as 1 hour at normal body temperature [omitted reference]".
8) Definitions that are similar or identical to the one given here are well known in the cryonics literature.
9) This issue is of great concern to computer users. A variety of tools and techniques exist for recovering information from damaged or otherwise inoperative disk drives, with the intent of recovering the memory and "personality" of the computer so that the user will not suffer a (sometimes traumatic) loss.
10) Cryonics will also fail if a person is prematurely thawed. This failure mode, however, is not an argument against cryonics, rather it is an argument for reliable refrigerators. A person injured in a car crash might die if their ambulance was struck by a train. This is not an argument that we should cremate accident victims rather than use an ambulance to transport them to a hospital!
11) There is fairly general agreement that death by the information theoretic criterion will not occur during storage of tissue at the temperature of liquid nitrogen, confer note 1. For this reason we neglect the possibility that significant information loss occurs during storage even though this might be viewed as theoretically possible.
12) Criticisms of cryonics are not supported by the extant literature. Interestingly (and somewhat to the author's surprise) there are no published technical articles on cryonics that claim it won't work. As one might suspect, there are also no articles in the neuroscience literature that address the issue of erasure of memory in the information theoretic sense, and there are no articles in the cryobiological literature that address the impact of freezing on the retention of long term memory in the information theoretic sense. There is an almost absolute conceptual failure to either understand or consider the implications of the information theoretic criterion of death. This conceptual failure is a severe impediment to research in this area. Even worse, the Society for Cryobiology has gone so far as to adopt by- laws calling for the expulsion of members who support cryonics. Members in good standing who support cryonics have been threatened with firing if they discuss their views publicly. Open discussion and review has proven to be a remarkably effective engine for driving scientific advance. The suppression of open discussion by a scientific society runs counter to one of the most central principles of scientific research and seriously impedes progress.
13 Many non-mammalian animals can be frozen to temperatures as low as -50 degrees C and survive.
14) Unpublished work by Darwin, Leaf and Hixon suggests that penetration of glycerol into the axonal regions of myelinated nerve cells is poor, and that increased damage to the axon results. This is consistent with the observation that penetration of glycerol through the many layers of the myelin sheath would presumably be slowed. However, myelinated axons are relatively large and serve a relatively well defined function: the transport of information. Even significant damage to the axon would not obliterate the fact of its existence or the path over which it carried its signal. As a consequence, this damage is unlikely to result in information theoretic death.
15) Suspension usually begins immediately upon pronouncement of legal "death." If legal death is pronounced upon cessation of heartbeat in a terminally ill patient who is being continuously monitored, then the ischemic interval can be kept under 5 minutes.
16) There are at least two likely ways in which cryonic suspensions conducted prior to legal death could be legalized. First, polls support the creation of a "right to die" for the terminally ill patient and an active movement is seeking to translate this support into some form of law. Second, terminally ill patients have sought and will presumably continue to seek the legal right to be cryonically suspended before (rather than after) a deteriorative disease (brain cancer, for example) has destroyed their brain. It is difficult to argue that such individuals should be forced to suffer an agonizing death, knowing that this agony is also destroying their brain and hence any chance for a future life.
17) There are various reasons for delay when a person is cryonically suspended, ranging from purely pragmatic issues such as delay following abrupt and unexpected accidents to legal and social forces that mandate that suspension not be started until after a legal declaration of "death." Whatever the cause, the effect is to increase the level of damage that takes place prior to suspension.
18) It should be clear that the claim of "irreversibility" is unsupported. Mitochondrial function is well understood: they provide energy for the cell. Even the complete absence of mitochondria would not cause death by the information theoretic criteria.
19) Much current work advances the (correct) claim that cellular, organ, and body function is lost under certain conditions. This loss of function is incorrectly and misleadingly labeled "death," "irreversible injury," etc. This work forms the backdrop against which tissue damage to cryonically suspended patients is measured by most biologists, cryobiologists, doctors and other health care workers. Clearly, this work predisposes such workers to dismiss cryonics because, by these criteria, much "irreversible" damage has occurred in most cryonically suspended patients. The implications of adopting the information theoretic criterion of death have simply not been considered, and we can reasonably expect a delay of several years to a few decades before they are. This would be consistent with historical data concerning the slow acceptance of new ideas. Ignaz Semmelweis demonstrated in 1848 that washing your hands in chlorinated lime after leaving the autopsy room and before entering the maternity ward reduced maternal deaths from childbed fever from as high as 25% to about 1%. Despite this, his proposal was widely ridiculed and little practiced for several more decades. Interestingly, few of even the most severe critics of cryonics claim that death by the information theoretic criterion is likely to have occurred when the question is posed to them directly.
20) It is interesting to note that current research into the three- dimensional structure of neurons often embeds neural tissue in plastic, and then produces a series of thin sections (typically 50 to 100 nanometers thick in electron microscopic reconstruction work) by using an ultramicrotome. The serial sections are then examined by a person (typically a graduate student) and the structures of interest in each section are outlined on a digitizing tablet and entered into a computer. The resulting data-base is used to build a three-dimensional image of the neuron. This work has been quite successful at determining the three-dimensional structure of small volumes (small enough for a graduate student to examine in a few weeks or months) despite the adverse effects of tissue preparation and sectioning. Sections vary in thickness. They also buckle, fold, and tear. Despite these difficulties, the human visual system can reconstruct the original shape of the object in three dimensions. Current electron microscopic reconstructions are quite capable of analyzing even the finest dendrites and thinnest axons, as well as determining the location and size of synapses[27,28], and even finer detail. It seems reasonable that the less damaging method of inducing a fracture at low temperature, and the more informative and less damaging analysis possible with nanotechnology (as opposed to destructive analysis of thin sections by a high energy electron beam) will produce more information about the structure being analyzed.
21) Under favorable circumstances, we might be able to terminate the division process sooner. That is, it might be that a relatively large piece of tissue (several tens of microns or larger) was relatively intact, and required little if any repair. Devising methods to take advantage of the minimal damage that might occur under favorable circumstances is beyond the scope of this paper.
22) For those concerned about the omission of water molecules and the like, we could just as easily store the coordinates of every molecule. This would increase the storage requirement, but would still be entirely feasible.
23) Despite the notorious difficulty in obtaining accurate information about specific aspects of brain "hardware," as discussed by Cherniak, it is still the case that rather rough bounds can be usefully derived.
24) A literature search on cryonics along with personal
inquiries has not produced a single technical paper on the subject that claims
that cryonics is infeasible or even unlikely. On the other hand, technical papers
and analyses of cryonics that speak favorably of its eventual success have been
published. It is unreasonable, given the extant literature, to conclude that
cryonics is unlikely to work. Such unsupported negative claims require further
analysis and careful critical evaluation before they can be taken seriously.