Manager of Advanced Applications

Predicting the future cannot be done by anyone with any certainty. Most of us would be hard pressed to predict with any accuracy what we will be eating for dinner a week from now. Predicting what twists and turns technology will take is far more difficult than ascertaining our future menus, but it is a task we all must undertake if we are to plan for tomorrow. And planning for tomorrow is something we all must do. The future has a way of arriving a little before we are ready to give up the present. Having some idea of what is coming won�t take all the surprise and mystery out of the future, but at least we will be a bit better prepared for it.

To predict the future, a reasonable thing to do is to look at trends and make reasoned guesses at which ones will continue and where they will lead. In essence we accept that the past contains the seeds of the future if only we can read the DNA of those seeds correctly. That is never an easy task.

For example, looking at the history of commercial aviation in the United States from its inception until 1958, there is steady increase in the speed of aircraft. Someone in 1958 looking at that trend and realizing that the "new" Boeing 707 jet liner could fly at nearly 600 miles per hour might have concluded that 50 years later we all would be flying at 3 to 10 times faster. But of course we are not. Except for the Concord that few of us fly, commercial aircraft fly just a few percent faster than they did in 1958.

This paper will look to the past for trends in hardware, software, networking, and education and attempt to extrapolate where they are going and what their broad implications might be. But readers should be wary. There are lots of different ways that trends can be interpreted and it is easy to pick trends that support one's thesis and ignore ones that don't. This paper may prove to be a road map for the future, but should it not turn out to be one, it will at least provide some new visions of what could have been and what still might be.

The recent past has been characterized by explosive growth of the World Wide Web, often just called the web. In August 1981 there were about 200 computers hosting web servers. By January 1998 there were over 30 million web servers. There are so many URLs (web addresses) that examining them all at the rate of one per second (an impossibly fast clip) would take over 8 years. Today, over 150 million people use the web. The user count is expected to more than double each year. Elementary math quickly shows that this trend cannot continue for very long because the number of web users would soon exceed the number of people on Earth. However, I expect this trend to continue and in fact to accelerate far into the future.

Today almost all web users are people. Tomorrow, intelligent devices operating on behalf of people and institutions will dominate the web. This will come about because of the accelerating decrease in the cost, weight, size, and power requirements of computer hardware, the increase in the speed of networks, and the growing wireless infrastructure now being put in place.

In the past the web was used to display documents and images. Today it is being used for education, research, software distribution, audio and video conferencing, and electronic commerce. Commercial uses of the web are the fastest growing web sector. The .com domains make up about half of all web domains with the .edu domains far behind. With billions of dollars in web commerce already being done, the web seems to threaten the way business is done. How long will there still be bookstores when on-line bookstore web sites are doing over 2 million dollars per day of business? Will the layoffs at brokerage houses accelerate as more stocks are traded on-line? Will any travel agents survive the ease of booking travel on the web?

Yet the web is in its infancy and has yet to show us what it will do when it matures. Today you must go to some desktop or laptop computer to use the web. Tomorrow, common intelligent devices that will number in the billions will routinely use the web with little intervention from us.

We will not make the leap to the web of the future in one giant step. There are already evolutionary signs of what is to come, if we are able to read them correctly. But we should be careful. Some of the most wonderful things are actually dinosaurs that in fact will not make it to the future.

A phone by BellSouth is a regular cordless phone, but when it gets too far away from its base station it becomes a wireless phone. In the future, all intelligent devices will be wireless, but will choose the most economical way to be so, just like this phone.

The Seiko Message Watch has its own phone number, its own IP address, and can monitor news, sports, weather, stocks, and whatever you'd like. It is a tiny, wireless, low powered, specialized, intelligent device that takes care of itself - all characteristics of the devices we will all have in the future.

The Cross Pad uses a pad on which you place a piece of paper as its input device. You write on the piece of paper and the pad converts your writing to digital ink or attempts handwriting recognition. Like devices of the future, this is a specialized intelligent device that uses something you use every day (a piece of paper) as its input device. Unfortunately you need the pad to interpret what you write. A device in the design stage by Carnegie Mellon University eliminates the pad. In fact it lets you draw 3D things in thin air.

The Toshiba 750 CDT laptop has everything you could wish for. It is fast, it has tons of memory. It has a built-in scanner. It has a camera. Its battery lasts for 7 hours. But it is a mature technology, and mature technologies, like dinosaurs get replaced. Of course anyone would be happy to have this laptop (or whatever laptop is best at the time) because although it is definitely a dinosaur, it is the very best dinosaur there is, and the critters that are evolutionarily superior to dinosaurs have yet to appear. But they soon will and we need to prepare for their arrival.

In 1965, Gordon Moore suggested in what's known as Moore's Law, that the density of transistors on a chip will double every 18-24 months. This also implies that that the cost per transistor should halve every 18-24 months. For over three decades Moore's Law has proved remarkably accurate, but during that entire period many have predicted that the trend could not continue. They said that, in effect, Moore's law would soon be repealed as it became impossible to continue to double the density of transistors on a chip.

In fact it appears that Moore's Law has been repealed, but not because the density of transistors has failed to keep up with it, but because things are moving faster than Moore had imagined due to unpredicted technological advances. Intel, for example, has found a way to store two bits per transistor and thinks they may be able to store four bits on a transistor. The effect of this is to immediately double or quadruple the effective density of transistors and cut the cost of a memory chip by a factor of two or four.

IBM has found a way to use copper, instead of aluminum wiring on a chip, making chips 40% faster, smaller, and cheaper. One advance after another has continued to accelerate Moore's law, pushing his 18-month time frame towards 12 or even 9 months.

The effect of this is to make every hardware element smaller, faster, and cheaper. Disk storage, that today costs about a dime a megabyte, will cost less than a penny a megabyte in 2004. RAM which costs about $4 a megabyte today, will cost less than 40 cents a megabyte in 2004. One megahertz of computational speed that costs about 80 cents today is likely to cost less than a dime in 2004.

The implication of this is that any device that can tolerate an increase in cost of a dime could become an intelligent device. A toy, a pen, a pair of eye glasses, a shirt, a door knob, a paperback book, or any common object you could image will in the near future be an intelligent device that has memory and computational abilities. Of course none of these devices will be like the general purpose PCs of today. They will continue to be toys, pens, shirts, and whatever they were meant to be, but they will be "smart" versions of what they were and they will be able to communicate with each other.

Today most software is bought in shrink-wrapped packages from stores that stock zillions of titles. EggHead Software realized that having physical stores stock software made no sense. They now just have a virtual store on the web. But they need to go a step further. No one wants or needs the shrink-wrapped packages and CDROMs. It is silly to send software via UPS or the US Mail when all one really wants is the bits that are on the CDROMs. In the future all software will be obtained via networks. No physical store will sell software.

As software delivery shifts to the network, the paradigm for paying for software will begin to look like that for television. Today there is commercial TV paid for by advertising, public TV paid for by government and donations, and cable TV and pay per view TV paid for by subscribers. A similar model will emerge for software. Much software will be free, paid for by advertising. This is how most of the search engines on the web are paid for, and there is a host of free e-mail services whose costs are absorbed by marketers who want to get their messages to captive e-mail users.

The government offers some free software today. Much more will be coming in the future. Why should we buy software to do our taxes? How long will it be before IRS offers it to us on the web?

Cable TV offers another model for software delivery that will represent a sea change for us, but one that we will soon take. Companies will offer software by subscription. Pay them a few dollars a month and get the use of a basic package of software. Of course the software you subscribe to will always be kept up to date and new offerings will be frequently added to keep your interest (and your monthly fees) flowing. For users who want more than the basic software package, there will be premium packages that offer products in vertical areas such as finance, architecture, engineering, manufacturing, music, and so forth.

Finally, there will be some software packages that will be pay-per-use. This will allow you to try a package (though free trials will be the rule) or to pay for a software package for just the actual time you need it.

With fast, cheap, intelligent processors and inexpensive software many new applications will become possible. Real-time speech recognition and translation, conversions among speech, text, image, and audio formats will be possible. Vision systems that will allow gas pumps to fill your tank without human intervention (Shell Oil is installing these now) or vans that can follow the edges of roads (CMU has an experimental van that does this) will become commonplace. Software will also become worldly. It will know, for example, that "Don't drink and drive" does not refer to drinking milk.

In 1987 a group of 5 Princeton students and I entered an Apple Computer sponsored contest to design the computer of the year 2000. After just a bit of design work we realized that for Apple to succeed it had to get out of the personal computer business and invent a new industry. By 2000, we reasoned, the personal computer industry would be flooded with mature technology devices that competed on paper-thin profit margins and nonessential gadgets and cosmetic changes. We saw the personal computer industry fading and a new industry, which we called the Telemation industry, emerging. The Telemation industry would build specialized, intelligent, communicating devices that we called Telematons.

Our design of the computer of the year 2000 was not a computer at all. It was the first of many Telematons that we envisioned would be built. We called our first Telematon the Apple PIE (Personal Information Environment). Although the PIE was not a computer, it had great computational power. It used its computational power to manage an information environment appropriate for its users. It took all of the information sources - radio, TV, CDs, the web, books, telephones, newspapers, other PIEs etc. - and managed the information. It drew no line between entertainment and education. Its focus was on integrating and managing information.

Although our design won second prize, Apple never built anything like it. Instead they built Performas, PowerMacs, and iMacs that have garnered perhaps 5% of the PC market. While Apple battles for a tiny share of a mature commodity market, other companies are introducing Telematon precursors.

A long, long time ago, memory, CPU cycles, hardware, and software were expensive. Because of that, general-purpose devices such as convention personal computers were built. But that was a long time ago. Every day intelligent devices become cheaper to build and general-purpose devices make less sense. Telematons are special-purpose devices that are now possible. They often have specialized input and output devices and might not even be recognizable as computers.

An example of a Telematon precursor is a system from FATS, Inc. which is used to train law enforcement agents, hunters, military units, and fire fighters. In its fire-fighting mode, the FATS (see www.fats.com) system presents a large screen to its user that covers an entire wall. On the screen is projected one of many fire fighting scenarios. A fire fighter faced with one of these simulations, grabs a hose, axe, or whatever tools and equipment are appropriate and goes to work on solving the problem.

The hose, which has the correct weight and feel of a real hose because it is one, is actually an input device, reporting back to the system its location, aiming point and so forth. The screen, rather than being static, responds to the fire fighters actions. Voice detection and recognition systems also respond to his or her commands.

The fire fighter is placed in a realistic, interactive, real-time environment in which to hone fire-fighting skills. No computer ever appears to be in sight. No keyboard, screen, or mouse is ever part of what the fire fighter sees. The heavy hose, not a mouse, is the pointing device. The firefighters' shouted commands and movements, not a keyboard, is the input device. If you peek into the office of the future there will be dozens of computers being used, but a visitor from today might not be able to detect a single one of them.

The Apple PIE group imagined dozens of other Telematons. Here's a few to get you imagining lots more.

Twenty years ago it seemed crazy to imagine anyone wearing a radio, cassette player, or CD player. Today it is hard to imagine a jogger not wearing one. Tomorrow it will be hard to imagine anyone, whatever they are doing, not wearing at least one Togamaton.

A Togamaton is a Telematon that is with you all the time. You might wear it as a watch or jewelry, or carry it with you in your wallet, purse, or pocket. Togamatons might also be part of your clothes.

For input, you might talk to them, scribble on them, or keyboard on them. Output devices would be their own displays, voice and video output, or they might use your eyeglasses as a display device. Some airline maintenance crews already use wearable computers that display their output on a lens of a pair of special eyeglasses.

What does your family use as a message center? Your PC, your phone, your TV? Of course not. They leave notes on your refrigerator. Why not have a Refridgermaton built into the door of every refrigerator? It would be a bi-directional e-mail, voice mail, web page, speakerphone, message center located right where it would be most useful. It would also include a UPC scanner that could keep track of what you put into your fridge, and with web access, your Refridgermaton could tell you what you are short of, produce shopping lists (and order over the web), and create recipes with what food you have on hand. You'd also be able to access your Refridgermaton remotely (should I pick up milk on the way home) or it might call one of your Togamatons to tell you to pick up the order it placed at the supermarket. A Refridgermaton could page your kids, remind you about birthdays, keep to-do lists for family members, and, of course, your fridge would still keep food cold.

"Cars may visit [the] Internet while on the interstate" - New York Times, September 19,1997.

Your car already has many computers and will have many more in the near future. A few cars already have GPS navigators, cell-phones, and other connections to wireless networks. An Automaton would put your car on the web, communicate with its service personnel, and optimize your driving experience. With an Automaton, if you locked your keys in your car (would there still be keys?) your Automaton in response to your voice would open your car for you.

With the profusion of Telematons, or whatever the cheap, fast, wireless, intelligent devices of the future are called, education will be fundamentally changed.

There are three variables in instruction; the material covered, the level of mastery of the material, and the time allocated to cover the material. Any two of these can be constrained. The other will always be variable. For example, if the material covered and the time allocated is fixed, then the level of mastery will vary from student to student. This is the model used in most classrooms today, and the one that makes least sense. Why bother to teach at all if one is not interested in mastery of the material? What does it mean when someone gets a C in chemistry 101? Does it mean that the person only knows 75% of what she should have learned? If that is true, how do we expect her to survive in chemistry 201 with students who got an A in chemistry 101? Can she still be a chemist having no knowledge of 25% of basic chemistry? Do you want a doctor operating on you who got straight C's?

Making mastery of material a variable is a compromise caused by the economics of teaching large numbers of people. Having decided that most people must be educated, it was too expensive to ensure that all of them were well educated.

It was not always like this. In Plato's day the material covered and the level of mastery were fixed. The time to cover the material was variable. The only acceptable level of mastery was complete mastery. One didn't move on to the next level until the previous level was mastered. This was accomplished using a one-on-one tutor system, with the material and presentation tailored to each student. This model simply doesn't scale up in any affordable way when hundreds of thousands of students need to be taught thousands of different subjects.

It has been thought that technology might be able to allow us to afford to get back to the Plato model of education, but so far the efforts have not been very successful. TV didn't do it. Movies and slides didn't either. Even web courses and distance learning mostly only reduced the costs of delivering the same old "talk and chalk" kind of instruction we were delivering without all the new technology. Students still got C's in courses and went on to struggle in the next course in the sequence.

While universities struggled with technology in education, TV stations mastered the art of delivering the evening news. Universities presented students with scholars who often had no teaching skills and used all the multi-media features that a blackboard, awful handwriting, and chalk could deliver. TV stations had trained actors and news readers backed by world-wide news teams, graphic artists, animators, and a studio full of support people, present the evening news to us. A five minute evening news piece was almost always more memorable than any number of one hour lectures in a classroom.

A first essential step for universities is to adopt the evening news model of education, deleting the bad features and taking the good features a step further. Some evening news plusses are:

You are where the action is live. At some point it is essential to see a diagram of a nuclear reactor (if that is what one is studying), but a live tour of one before seeing the details makes learning more meaningful.

The use of multi-media. Using as many of your senses as possible improves learning. And making material compelling and entertaining improves retention.

The evening news brings in the specialists and pros. Doctors, lawyers, pilots, and a host of consultants are always on hand to provide the in-depth inside information. There are also a host of staff specialists such as weathermen, sportscasters, financial analysts, and so forth.

The evening news is scripted, rehearsed, polished, and checked for correctness and good pedagogy. News is delivered by professionals who know how to present information.

Some evening news minuses are:

• There is no interactivity.
• Many newscasters are just newsreaders and have no depth in the subject they are presenting.
• There is a tendency to oversimplify and to do sound bytes.
• The evening news is linear.
• Attempts at humor and levity often fall flat or are just distracting.

While adopting the good features and avoiding the bad, the evening news model we should adopt should throw away grading entirely and instead insist on mastery. Everyone gets either an A+ or gets no credit. And they take as long as they are willing to take to get it. It should also throw away the awarding of degrees. Students would simply get credit for whatever work they mastered. A transcript would simply consist of a list of the material a student had mastered.

Of course to implement this model would require a sea of Platos skilled in every discipline offered by every college and university. Even if we could afford such a group of Platos, no such group exists. What we need instead, and what will soon be possible, is a software aid (that would run on Telematons, of course) that I call SMILE, for Software Managed Instruction and Education.

The SMILE paradigm creates a personal software mentor model. In essence, SMILE presents a student with a personal Plato-like tutor for all areas the student is studying. SMILE lets every student go at her own pace. It soon learns what pedagogy works best for a particular student and soon learns a student's weak and strong points. It cajoles, urges, entertains, and manages a student's education. Its presentations use the modified evening news model described above.

SMILE is network connected and also has access to NOAH (Network Optimized Allocation of Help), a new network service that allows network connected peers and specialists to provide live interactive help to supplement what SMILE can do.

Can we really build a SMILE system today? All of the hardware and software pieces needed are getting close to the point where we could afford them. The more difficult problem is our difficulty in giving up the present and forging forward into the uncertain future.