Crystal Balls: Coming in 30 Years to a Theatre as Near as You Think -- Superhuman Cyborgs and Androids - Artificial Intelligence

Crystal Balls 3: Coming in 30 Years to a Theatre as Near as You Think --
Superhuman Cyborgs and Androids

Dateline: March 1, 1998

--Me. Read on . . .

THIS article is the last of a three-parter. The previous two articles looked at Forecasting the Future, showing that there are indeed methods for forecasting the future, and at Why Forecasting is Important, arguing that failure to look to the future in a time of rapidly accelerating technological change is a recipe for disaster.

In many ways, the present article represents a summation of most of what I have written in my Mining Company features to date. It may or may not be the end of the story (see postscript). If new readers need clarification of points made herein, they can find it by using the "search this site" function or going to the list of previous features. I chose medicine as the focus for this particular article because it seems to me the most striking example of what the future has in store.

Eight Key Trends

Eight key trends are visible in the ongoing development of hardware, software, and the ways in which we interact with them. Each of the eight trends individually promises to have a revolutionary impact on medicine, through adminstrative technologies that will vastly improve communications and record-keeping, neuroscientific technologies that will lead to repair and augmentation of the human brain and to android brains, bionics that will lead to repair and augmentation of the human body and robot/android bodies, and through telemedicine, which will let all this action happen at a distance.

First, software is getting smarter. It is acquiring (artificial) intelligence (AI). Key to much of the recent breakthrough development in AI has resulted from the integration of "rule-based" expert systems that operate strictly algorithmically (if this, then that) with fuzzy connectionist systems called neural networks that operate heuristically (on rules of thumb—if this, then probably that). As a result computers now think more like humans and less like computers.

Already we have programs that can beat the World Chess Grandmaster at his own game, expertly diagnose pain, identify abnormal EKGs and diagnose heart attacks better than cardiologists, and detect Pap smear abnormalities better than cytologists.

Forecast: By the year 2030, given the extension of Moore’s Law to apply to software development, software will have developed domain-specific intelligence, knowledge, and expertise of at least equivalent capability to most low and middle-level and some top-level white-collar workers in all industries and professions, including the medical and nursing professions.

Second, hardware is getting smaller, through developments in miniaturization and its extreme expression, nanotechnology. Nanotechnology is technology for building things—ultimately, anything—atom by atom and molecule by molecule. The molecular and atomic computers mentioned earlier are examples of nanoscale (billionths of a meter) devices. Nanotechnologists hope to build nanoscale devices called universal assemblers that can reconfigure their environment into something else, molecule by molecule. Instead of an entertainment center loaded with music CD player, DVD player, TV set, radio, VCR, etc., you would have a single device made up of trillions of universal assemblers that could reconfigure a TV set into a boombox, or into a perfect reproduction of the Mona Lisa, for example.

One company, Zyvex, claims to be on the road to building such a device, and Sandia National Laboratories have developed a one-millimeter-square "intelligent" micromachine incorporating computer chip controllers. Possible applications for the process include tiny drug-delivery devices. By 1995 Sandia had already succeeded in mass-producing micromachines that could perform work, turning gears each one-hundredth the weight of a dust mite at hundreds of thousands of revolutions per minute. Each gear was approximately one-hundredth the thickness of a sheet of paper, and smaller in diameter than a human hair.

Forecast: By 2030, given the extension of Moore’s Law to apply to miniaturization and manufacturing, and given the development in AI noted above, many light and heavy manufacturing processes will have been replaced by intelligent micro- or nanomanufacturing processes. Such processes may be implanted in the body to perform medical and surgical tasks.

Third, machines are becoming more mobile and more dexterous, through robotics. Robotic mobility comes in many forms, from wheels through caterpillar tracks and arthropod legs to two-legged humanoid gait. Similarly, robots have multiple modes of manipulating objects, from simple and gross two-fingered grippers to multi-fingered extremities able to manipulate objects at very small scale in multiple degrees of freedom. Unlike the unaided human hand, robot extremities can be magnetized to adhere to metals, heated, and otherwise be adapted to the materials they handle.

Forecast: By 2030, given the extension of Moore’s Law to apply to robotics, and given the trends in AI, miniaturization, and manufacturing mentioned above, mobile robots will possess the manual dexterity to perform microsurgery too delicate for human hands, not to mention more coarse-grained operations such as physical therapy, nursing, plant maintenance, and janitorial tasks.

Fourth, machines are growing more aware of both their surroundings and of themselves, through sensor technologies. A bionic nose has correctly identified the common bacteria Staphylococcus aureus with 100 percent accuracy and E. coli with 92 percent accuracy, a LIDAR (LIght-Detecting And Ranging) invisible-beam flashlight can unobtrusively tell how much alcohol is in a person’s body in concentrations as small as 100 parts per billion, and the U.S. Defense Advanced Projects Research Agency (DARPA) is working on "Smart T-shirts"—fiber optic-laden garments with built-in bio-sensors that not only relay a soldier’s vital signs but also his location and the exact time of injury.

Forecast: By 2030, given the extension of Moore’s Law to apply to sensor technology, and given the trends in AI and robotics discussed above, very intelligent, mobile, dexterous, decision-making robots will see, hear, smell, and touch in finer resolution than we can and will reach parts of the electromagnetic and chemical spectra we cannot.

Fifth, machines are becoming more multi-modal in their ability to communicate with us, through the technologies of virtual reality. We communicate with one another not only through speech, but also through gestures, touch, "looks," body attitudes, and other non-verbal modes of communication. Multi-modality will become prevalent in human–computer communication as well as in human–human interactions over computer networks as computing power grows and sensors and software grow more sophisticated.

MIT’s Media Lab produced an eye-tracking device in the early 1990s that could tell what a subject was looking at. Zeneca Pharmaceuticals has developed a VR simulator designed to give physicians some idea of not only the physiognomy of a migraine but also what one feels like.

Advances in computing power and miniaturization are leading to inexpensive eyeglass-like devices producing high-resolution 3-D video images from Pentium-class computers. Architects and their clients can walk through and manipulate VR representations of buildings not yet built, and molecular chemists can examine and manipulate a large VR representation of a molecule, alter its atomic composition with datagloved hands, and create new compounds never seen before.

Forecast: By 2030, given the extension of Moore’s Law to apply to VR technology, VR will have replaced not just the keyboard, mouse, and monitor but entire offices and other locations, including some labs, and their equipment and instruments. To the extent that very intelligent, mobile, dexterous, ultra-sensitive, decision-making robots have not yet put us out of work entirely, our primary place of work will be the home or wherever we choose to don our VR gear.

Sixth, machines are growing more connected with one another, as well as with us, through local area networks embedded into the fabric of "smart" home, office, and factory buildings (where such continue to exist outside of VR). Like the heating/cooling, stereo, and other systems in Bill Gates’ house, appliances will be aware of each other’s state and anticipate the needs and preferences of the building’s individual occupants as they move around.

In turn, these local networks and the appliances connected to them will be connected to the Internet, enabling them to be monitored, controlled, and repaired remotely. There is already a proliferation of devices connected to the Internet, including: video "spy" cameras, a soft drink dispensing machine, a robotic arm in Australia, and several large astronomical telescopes.

Forecast: By 2030, given the extension of Moore’s Law to apply to connectivity, nearly every household, commercial, industrial, and medical device—including the very intelligent, mobile, dexterous, ultra-sensitive, decision-making robots discussed above—will be connected to the Internet, enabling them to remotely monitor, troubleshoot, and repair one another, to share information with one another and with us, to gather and analyze data on a vast scale, and to act on it.

The seventh and a chilling trend in hardware and software engines and interfaces is for machines to become more autonomous, operating without help from or recourse to humans. The software robots or "bots" that organize e-mail work quietly in the background. Physical robots such as the Mars rover Sojourner are programmed to do their own thing unless the program is interrupted and overridden by a human operator. As their programming grows more sophisticated, enabling them to handle a greater diversity of obstacles, they do better and we trust them more. As such systems grow more complex, we will have less and less choice but to trust them.

Adding to the chill is the fact that software contains or can develop "bugs" or unwanted and dangerous pathologies. The Internet itself has come close to meltdown through aberrant autonomous software, as has Wall Street, and AT&T’s national telephone network was crippled for several days in 1994 when its then new Signaling System 7 software, containing 6 million lines of code, went wrong. AI programs are increasingly used in power stations, including nuclear plants, to monitor and control critical functions.

Forecast: By 2030, given the extension of Moore’s Law to apply to autonomy in computer programs and the machines and networks they inhabit, then the very intelligent, mobile, dexterous, ultra-sensitive, decision-making, networked, and supremely informed robots discussed above will operate autonomously.

Complexity itself is the eighth key trend, and potentially the most chilling of all. Modern studies of complexity are confirming and to some extent helping us to understand the observation of philosophers and scientists over the centuries that when a system reaches a certain level of complexity it either perishes in entropy or leaps through what communication scholar Marshall McLuhan called a "break boundary," mathematician John von Neumann called a "complexity barrier," and philosopher Dan Dennett calls a "saltation," wherein the system turns into something wholly new, advanced, and unexpected.

Forecast: By 2030, given the extension of Moore’s Law to apply to complexity in the very intelligent, mobile, dexterous, ultra-sensitive, decision-making, networked, supremely informed, and autonomous robots discussed above, then (1) we may be unable to control them and (2) out of them will emerge something wholly new, advanced, and unexpected.

Administrative technologies

Automatic speech recognition (ASR) and machine translation (MT) are two communication modalities that in particular affect the administrative aspects of the physician’s (and the nurse’s) work, eliminating paperwork and enabling cross-cultural comunications in caregiving as well as in medical research.

Database technologies already figure hugely in medicine, for tasks ranging from billing to epidemiological research. The largely AI-driven trends in database technologies are to make individual database structures more accessible to one another, so that data can be more easily shared, and to "mine" them for hidden patterns of information. Data mining tools are one of the hottest commercial items for large enterprises at the present time.

Intelligent agent software called InfoSleuth from the Microelectronics and Computer Technology Corporation is being deployed jointly by several U.S. and the European agencies to improve access to and sharing of environmental data within scattered and dissimilar government databases. A goal is to utilize emerging information search, access, and retrieval technologies to vastly improve the way organizations share environmental information through standard Internet browsers. The potential of InfoSleuth to link disparate medical and health databases globally is clear.

Joe Flower presents a good overview of immediate future trends with regard to database linking in Healthcare Forum Journal. Allied to these developments in database technology is the smart card, a credit-card-like device with an embedded computer chip and memory. Already in use in Europe for personal financial transactions, smart cards will eventually carry personal medical information that can be read by (and written to) a computer at the hospital or doctor’s office. Dupont Chemical Company and Data-Disk Technology are developing a rugged medical tag device that could carry individual medical records including text, X-rays, electrocardiograms, and other test results.

By distributing data among billions of smart cards, pressure on storage capacity in the office computers will be alleviated, though data for epidemiological and other purposes may be retained long-term on servers. Developments in miniaturization discussed above plus developments in bionic and neurological engineering to be discussed below could lead to the permanent implantation of smart-cards in the human body.

Forecast: By 2030, given the extension of Moore’s Law to technologies useful in the administration of the medical practice, together with the developments in AI, miniaturization, connectivity, and VR, physician and patient alike will be equipped with small, universal translators enabling us to converse face to face, alone or in conference with others, in any language in a VR environment, and a finally paperless virtual practice utilizing data from smart cards and from other sources on the network will result. Billing and insurance claims will be handled automatically and autonomously by the network.

Neuroscience

Increasing computer power through new architectures and better engines and interfaces have led to more powerful tools and metaphors for medical researchers and clinicians alike not only to understand and treat the brain but also to augment and improve it.

The Department of Defense is sponsoring research into "hybrid information appliances" which it hopes will lead to the manufacture of bio-computers and sensors to be grafted into soldiers’ bodies and brains so they don’t have carry bulky, energy-guzzling electronic devices into battle. Rather than carrying 25 pounds of batteries, a soldier can feed his electronics on K-rations.

The technique involves building hybrid biological–electronic circuits using immature neurons that have not yet developed dendrites and axons, by placing the neuron on a silicon substrate and depositing thin lines of certain proteins in a radial pattern outwards from the cell like the spokes of a wheel. The neuron will start to grow dendrites (input wires) along the spokes. A dab of the protein laminin applied to one of the developing dendrites causes it to develop instead into an axon (output wire), and the entire neuron can then be activated by electrical impulses applied to the dendrites through the silicon substrate. The primary research goal is to figure out just what neurons do, and how they do it. Following that, the less militaristic researchers hope to see the technology developed into humane applications such as the regeneration of damaged nerves.

Such research is years away from success, but even without going to these lengths, connecting nerves to electronic components at a coarse grain is producing results today. The Huntington Medical Research Institutes in California, which have already developed approved treatments for epilepsy, Parkinson's Disease tremors, and for restoring arm movement in some quadriplegics using electrical implants to stimulate nerves directly, is working on implants to enable paraplegics to have sex and conceive children, and a longer-term goal, through what is known as neural prosthesis technology, is to restore sight to the blind and enable handicappers to control urination and artificial limbs with just their minds.

Other approaches to building, repairing, and augmenting the human neural network are being undertaken by Professor Ted Berger and colleagues at the University of Southern California and Dr. Hugo de Garis at Japan’s Advanced Telecommunications Research Institute. The USC project aims to restore functionality to brains damaged by stroke, head trauma, Alzheimer's, epilepsy, and other maladies, by creating a parallel-processing network out of hybrid electronic/photonic chips that could function as a brain implant. Dr. de Garis is seeking to construct an entire artificial brain with the equivalent of billions of neurons.

Much closer to real medicine today are the brain scanning technologies of magnetoencephalography, 3-D electroencephalography, functional magnetic resonance imaging (fMRI), positron-emission tomography (PET), magnetic resonance spectroscopy (MRS), and the PET reporter gene/PET reporter probe. These machines do all manner of things, from mapping the blood flow and chemical changes in the brain to showing a video of specific genes in real time as they are expressed inside the brain.

Geoff Aguirre at the University of Pennsylvania has used fMRI on subjects immersed in a virtual reality world, enabling him to test a much greater range of thought-processing mind/brain activity than is otherwise possible for a patient pinned motionless inside an 11-ton magnet. Magnetoencephalography enabled neurobiologist Karl Friston to map a subject’s brain activity as the subject decided to make small hand movements. The scan showed what areas of the brain were involved, and when, and that it took the subject more than a twentieth of a second between deciding to move his hand and actually moving it.

Forecast: By 2030, given the extension of Moore’s Law to apply to neuroscience, there will be cures for most if not all types of neural disorders, augmentation of the human brain’s processing powers, artificial brains, and, given the developments in AI-driven interconnected databases and smart cards, a deep understanding of mind, brain, and consciousness. Given the developments in AI, miniaturization, robotics, and VR mentioned earlier, neurological diagnosis, treatment, and augmentation could be carried out by very intelligent, mobile, dexterous, ultra-sensitive, autonomous decision-making robots with humanoid brains operating simultaneously in the patient’s home and a VR lab/theatre environment.

Bionics

The same technology trends that have led to surges in knowledge and treatment of the brain have also impacted knowledge and treatment of the body, and our ability to replicate much of it. The development of a bionic nose has already been mentioned, and artificial hearts, lungs, kidneys, prosthetic hands, arms, legs, and replacement hips have become almost commonplace in modern surgery. Artificial skin is in widespread use.

Some other exemplary developments include several promising approaches to the creation of artificial bone by biomedical engineers at Arizona State University’s Center for Solid State Science and at Rice University. MIT and others are working on artificial muscle, and as of 1996 at least six companies in the U.S. were testing partial blood substitutes in human surgeries. Finally, a $950 Generic Visual Perception Processor developed in France simulates and in some ways outperforms the human eye.

Forecast: By 2030, given the extension of Moore’s Law to apply to bionics, there will be cures for many if not most or even all types of physiological and anatomical disorder and augmentation of the human body’s power. In addition, given the developments in AI-driven interconnected databases and smart cards, we will have a much deeper understanding of the body. Given the developments in AI, miniaturization, robotics, and VR mentioned earlier, physiological and anatomical diagnosis, treatment, and augmentation could be carried out by very intelligent, mobile, dexterous, ultra-sensitive, autonomous decision-making robots with augmented humanoid brains and bodies operating simultaneously in the patient’s home and a VR lab/theatre environment. We will not only have created a race of bionic humans (cyborgs) but also will have all the pieces to make an artificial human—an android. Both the cyborg and the android will be physically and mentally superior to present-day humans, and potentially—through continual repair and replacement of damaged or worn-out parts—immortal.

Telemedicine/Telesurgery

At the apex of our pyramid of trends, resting, drawing from, and dependent upon all of the building blocks below it, sits telemedicine, which for convenience I will assume to include telesurgery. The U.S. federal government’s Joint Working Group on Telemedicine defines it as "the use of modern telecommunications and information technologies for the provision of clinical care to individuals at a distance and the transmission of information to provide that care." It spans every echelon of health care, from first responder/emergency medical systems to tertiary medical specialty consultations and home care, and to the collection and analysis of epidemiological and environmental health data.

An example of telemedicine in large-scale current operation is the U.S. military’s Primetime III project in Bosnia. It provides 24-hour access to computerized medical records, full-motion remote video consultation between theater medical units and tertiary care facilities, battlefront delivery of laboratory and radiological results and prescriptions, digital diagnostic devices such as ultrasound and filmless teleradiology, and medical command and control technologies.

Telesurgery, a method by which surgeons will be able to operate on patients halfway around the world using datagloves, high-speed communication links, 3-D imaging, and cameras and robots at the patient's location, has advanced to the point where animals have been operated upon from a distance of five kilometers and a human patient had a gall bladder removed by a surgeon from across the room. The main constraint on telesurgery is the speed of the communications link. The current maximum distance for a wireless link is 50 miles, and 200 for a cable connection. Internet2 is designed partly to find ways to increase the speed of the Internet so it could be used, inter-alia, for telemedical purposes.

DARPA is currently working on a five-dimensional total body scanner known as a "Medical Avatar." The scanner will produce enough data to enable a physician to examine a living, breathing, holographic VR replica of a remote patient, even reaching inside to feel the heart beating or feel a broken bone. It’s about 25 years from production, DARPA estimates.
Already, we have floating holograms of hearts you can feel beating in your datagloved hand, surgical glasses that superimpose digital images over real organs to give surgeons X-ray vision, and "robodocs" that assist in hip replacement surgery.

The message is that telemedicine is already here, albeit in limited form. Doctors use videoconferencing networks to view x-ray images or examine a skin lesion of a remote patient. VR already helps patients overcome claustrophobia, agoraphobia, fear of public speaking, fear of heights, impotence, premature ejaculation, and to control pain for burn patients and women in childbirth. Given the trends we have discussed in other sections, telemedicine is destined for spectacular achievements in coming decades.

Forecast: By 2030, given the extension of Moore’s Law to apply to telemedicine and all of the technologies upon which it rests, there will be cures for most if not all types of neurological, physiological, and anatomical disorder, and methods for the augmentation of human and machine power. These cures and augmentations will be performed by androids or cyborgs drawing their expertise from the global networked knowledge base.

Implications

Steve Heimoff, writing in Healthcare Forum Journal, thinks healthcare professionals view "replacing the paper medical record" as the "killer application" of the computer in medicine, and properly notes its more advanced role as a decision support tool for caregivers. Heimoff also makes reference to "fears that computers could undermine the doctor–patient (or nurse–patient) relationship."

But drawing upon the key assumption that Moore’s and Metcalfe’s Laws will continue to drive development in all computerized technologies, we have arrived at a much more startling scenario; one in which cyborgs and/or androids don’t just undermine but take over the doctor/nurse–patient relationship, and the administrator’s role too. For some this will sound horrifying. Some will scoff— Heimoff quotes one physician-administrator as arguing that it will take 5–10 years for many medical institutions just to get their in-house records computerized. I would argue that any medical institution lacking computerized records five years from now will be out of business.

For others, however, our startling scenario will carry a message of hope, telling us that the time and the capabilities are almost upon us to explore new frontiers in inner and outer space with ageless, re-engineered bodies and augmented brains, and with intelligent androids not just as tools or aides, but as companions and co-discoverers.

I have focused on medicine in this article because it medicine is a major adopter of technology, and because I'm preparing a lecture presentation on the topic for medical associations. But the same kind of analysis applied to any other professional field, such as the law, government, manufacturing, etc., will yield results just as startling. I hope soon to start work on a future analysis for the law.

Until next week,

NEXT WEEK: Two possibilities for next week: (1) A review of Ward Systems Group's NeuroShell Easy Predictor and NeuroShell Easy Classifier, OR (2) Another edition of Intelligent Machine Post. It depends on whether or not I can do a thorough review of NeuroShell in time. I've already checked out the downloadable demo version of Predictor and it's good--VERY good. Not only the program itself and the genuinely easy and intuitive interface and help, but also the built-in tutorial on how neural nets work. The demo is worth getting for the tutorial alone, though as Ward Systems says, it's there only for those interested in the underlying technology. You DON'T need to understand the technology to get good practical results in predicting results or classifying data.

Help Wanted: Got questions or comments on this article or on any other AI-related subject under the sun? Post it in the AIBB!

POSTSCRIPT: I've had a year's sabbatical financed by the sale of my ownership in the Internet company I founded. The sabbatical not only gave me time to maintain this site and write these articles, but also to draft a book manuscript. The book might finance me going forward, but it is not yet in publishable state and I am too much an unknown to get much of an advance out of a publisher. The money won't last forever and I need to find income, and when I find a job it may be that I will no longer have the time to keep this going. We'll see; I just wanted to prepare you, just in case.

Previous Features