Where's It All Going?

One of the hardest things for people who haven't been intimately involved in the computer business for at least a decade to grasp is the nature of the exponential growth underway in our industry, and the consequences of continued exponential growth. I believe that many technology-oriented companies fail because the financial or management-oriented executives who run them don't understand how rapidly the underlying technology is developing, and how quickly their companies must change to survive in such an environment.

Every now and then it's worth reminding ourselves that we're on a wild rollercoaster and that failing to aim high enough is as sure a prescription for disaster as falling off. Here's a view of the immediate future I presented at a company meeting in early 1988.

Recently, a lot of people have been asking me ``where's it all going?''. Well, I've been spending some time thinking about that, both because if you get paid to design products, it's nice to know something about the world in which they'll be used, and also because it's embarrassing to have to answer that question ``Beats me''.

I guess you can't live in Marin County for 14 years without being prone to having ``flashes''. I was driving through San Rafael last Friday, and suddenly I flashed on this.

``Hey, this is the future.''

As a typical 1950's technoid kid, I lived for the future. You know:     Mechanix Illustrated, Analog science fiction, ``Science Fiction Theatre'', ``The Outer Limits''. Now you drive through San Rafael and what do you see? On this block there's a computer store. Around the corner, a software store. Down the street, there's a hologram store. And two blocks over there's a satellite ground station store.

There's moon rocks in the museums, a space station in orbit, computers on our wrists, and telephones in our shoes. So now what? Would you believe....

But first, what does this have to do with Autodesk? In thinking about starting Autodesk, I proceeded from two basic assumptions. One: it takes money to develop creative and innovative products and bring them to market. Two: it takes creative and innovative products to make money, or at least to have any fun making money. So when I thought about how to build a successful company, it seemed pretty obvious: hook up a creative engine to a cash machine and throw the switch. And hey, it not only works, there's money left over. Kind of like a high-tech cat and rat farm.

So what's going to be happening to this industry we're in? Things are going to get very weird, very fast. We all know what an IBM AT is, and what it feels like to use one. The current top end machines, such as the Compaq 20/20 and the Sun 3, run about ten times as fast as an AT, and it's that kind of performance that lets us go in for solid modeling, shading, and full 3D on a desktop. But it isn't going to stop there. A couple of weeks ago I took a little machine home that runs forty times faster than an AT. And let me tell you, that feels totally different from anything you've used before.

But it isn't going to stop there.

The two magazines I always read are ``Aviation Leak'' and this one, Electronics. I'd just like to quote a few items from just the last two issues of Electronics.

``Stellar Computer to introduce $75,000 to $100,000 machine in the next few weeks. 20-30 million instructions per second, 40 million floating point operations per second.'' That will run floating point code like AutoCAD about eighty times faster than an AT.

``Motorola announces a totally new RISC chip for this spring. First shipment will be 17 MIPS, but we'll show our customers how they can hit 50 this year''. That's about a hundred and fifty times faster than an AT for our kind of work.

``Ardent launches first supercomputer on a desk. Runs up to 64 million floating point operations per second. Less than $150,000.'' Call it 200 times faster than an AT.

``Apollo unveils a desktop supercomputer. A true 64 bit workstation, it hits up to 140 million floating point operations per second. Priced below $80,000.'' Oh, say 450 times faster than an AT.

And where do those guys say they're going? After Cray, who's up there at about 3,000 times faster than an AT. And Cray's heading for 30,000, with IBM, ETA, Fujitsu, Hitachi, and NEC in hot pursuit.

And it isn't going to stop there.

Now you might ask, ``What in the world do all of these $100,000 to 16 megabuck machines have to do with little ole Autodesk?'' The answer lies in the fact that if you open up one of those gilded crunch-o-matics and look inside, you discover that there's really nothing in there but sand and profit.

We aren't running out of sand, and competition will take care of the profit.

So what this means is that the products that we start to develop today will, in the middle part of their life cycle, have machines available for less than ten thousand dollars, which run from one to five hundred times faster than the IBM AT--the platform that carried AutoCAD to its initial success, and from ten to fifty times faster than anything our customers have access to today.

Changes of that magnitude mean much more than nozzles appearing on the screen in 50 microseconds. They portend another qualitative shift in the kinds of products we can deliver to our customers, and those products will change the engineering design cycle at least as much as all of the computerisation to date.

Let's try to look at an Autodesk customer in two or three years. This person will be using a computer that looks little different from those we use today. It'll probably have a larger screen, and that screen will have higher resolution. But look at what's on that screen! All controlled from a common user interface, into which additional products can be plugged like chips into sockets, this user may be dragging the diameter of a weight-reducing hole in a mechanical link. In one window, the shaded image is updated in close to real time. In another, the weight and centre of gravity is being recalculated as the hole changes size, and in a third window, running a little behind, a stress analysis is being run on the part with the load displayed in colour. The design of the part may be overseen by a constraint manager that digests the design rules that bound the problem, and not only helps the designer move among acceptable designs, but even offers advice when a design limit is hit. In minutes, the designer can produce a photographic quality image of the part. And naturally, the manufacturing drawings are automatically updated as the design changes.

The designer will have access to a vast database of design data, will be able to track the design history of a project, and will be able to access large bodies of data and to work with others with the computer facilitating their collaborative endeavours, thus aiding in the evolution of knowledge and the development of sound designs. All of this will shorten the design cycle, increase productivity, and feed back into the process of creating the tools. And, we get paid for it.

Autodesk is creating incentives that spawn vibrant, burgeoning, and highly competitive markets for add-on products, extensions to our product line, and new hardware platforms. This will continue to exponentially expand the armamentarium of tools the designer can apply to create better products and bring them to market faster. Those who attempt to define a ``total solution'', or ``control the market'' will continue to discover what Autodesk has been teaching them for the last half-decade: that an open system unleashes the talents of tens of thousands of creative people each motivated by their own self-interest, and that their energy and dedication will carry that open system to successes unimagined by those who think they can ``plan for the future''.

Variation, selection; replication, extinction; innovation, competition; it's been working for billions of years, and it isn't going to stop tomorrow.

I think that most great business successes are the result of somebody tripping over an exponential curve, driving a spike into it, and holding on for dear life. That's what we're doing now. And there's no sign that our curve is turning back down, or that we're losing our grip. Every part of that utopian dream system is the logical outgrowth of work which is already underway at Autodesk, and by the time it comes together in, say, two years, there's no doubt in my mind that the computer power it needs will be in the hands of our users.

And the beauty of it is, we don't have to disappear into a corner and make something new. We will evolve our way there, product by product, release by release, feature by feature, with the market we've created guiding us and everybody else in it toward the best solutions to the problems it faces.

Well, time's up. Enough of the distant blue-sky horizons 24 months from now. We have to get back to the future.

Every now and then I try to push a new feature of AutoCAD off the deep end to see if anything gives way. After adding three dimensional space polylines to the AutoCAD Release 10 development version, I used a map database to build a three dimensional model of the Earth's surface. This model, over 3.5 megabytes on disc, can be viewed from any point in space with an AutoLisp routine. This full-Earth view is centred on Cleveland, Ohio.