Are We a Model or a Document?

Our dreams for this technology are pulled in different directions by competing traditions as to what we put into computers and what we expect to happen there. Hypertext straddles this division and is thus able to go either way. But our dreams are most likely aligned with one or the other.

In the early days of Lisp there was a similar division between the east coast way from MIT and the west coast way from PARC. MIT hackers thought of a program as a living thing coaxed this way or that while it ran. The PARC model had the program safely stored in a file and only brought to life when it was briefly run as written.

Wiki and agile grew up together with agile leading the way. The brand new personal computers on desktops were largely inert until brought to life by the hazy dreams of a potential users interpreted fresh every day through the suddenly social process of programming.

Wiki enabled these computing pioneers to articulate what happened when their work shifted daily by writing in hypertext that also shifted daily. Wiki was a model for this new way as well as a document of its progress. Hypertext straddled this divide.

# Roots

I am consumed by the notion of rendering the work I do as a model similar in many ways to the model train that was the other hobby besides Lisp that consumed MIT hackers.

The model train was something from outside brought inside with exquisite detail that would be revealed when one leaned forward to look closely. It also ran. The trains moved. But the train signals, the lights and sensors, are what attracted the electrically competent.

Contrast this vision of purpose with PARC founded by a copier company and chartered to invent the "paperless office" where documents flew electronically from workstation to workstation to be revised and ultimately printed where what you saw was what you got forever after.

The Learning Research Group lurked within PARC and told a story much bigger than paper. Alan Kay asked that they consider not the book, but the dynabook, that participated in the constructivist learning of children of all ages. Alan was thinking the east coast way, connected to MIT through Minsky and Papert and from there, Piaget. The program was alive, made from a million little cells, doing what cells do, organizing to realize emergent behavior.

Aside: Smalltalk was distributed as a memory image that would be trained in new behaviors through "meta" programs contained within itself. This living thing could be frozen and then revived elsewhere. The Smalltalk I encountered was a revived copy of the very Smalltalk that the Learning Research Group had coaxed into existence years before. More model than document.

# Displays

Leon Chua started his professorial career at Purdue while I studied there in the late 60's. He introduced nonlinear dynamics to the calculus of electrical circuits. He also brought grant money and spent some of it on a computer with a graphical display, the second on campus. I hung out there.

All graphical displays in those days were made from WWII radar tubes where the beam traced out figures under computer control. More figures took longer. The computer was fast, but not fast enough to fill the screen. We dreamed then of a computer memory so vast that there could be one bit allocated to every spot on the screen and they could all be on at once. Total white. It would be awesome.

Soon the engineers at PARC were building desk sized computers with enough memory to display all white and then put black letters on it and maybe even diagrams. They made a hundred of these computers and gave one to every researcher to see what they could do. Alan Kay called it the interim dynabook. Not the dynabook of his imagination but it would have to do.

By the time Smalltalk made it to Oregon I had a bit-mapped display on a desk sized computer in my office for my own use. The future had arrived but what would we do with it? I had been doing display research with vector graphic projectors so fast we could fill the dark screen with a page of Kanji characters which one could touch to menu through the alphabet. But the little cells of Smalltalk, the objects, and their emergent behavior. That was new and unexpected.

I explain all this so that you, dear reader, might understand what if felt like to have the display of our dreams appear and have the dynamic concepts at hand to draw them. I explain this because a better dream is about to come true.

# Models

Bret Victor describes thinking with models that we can hold in our hands and use all our human senses to comprehend. He is exploring that vision by projecting onto plain paper with enough marks on it that the computer can see it handled and adjust accordingly. Abstraction. Dynamics. Creative thought.

Jeri Ellsworth is a maker that loves games. Valve asked her to build the team that would invent the future of games. Having tried many things the one that stuck for her (but not Valve) was, like Bret, projecting an interactive model into the creative space occupied by people. In her case the projection was on a reflective table mat again with enough dots that the computer can see it and respond.

Bret and Jeri are both visionaries augmenting thought with models that leap out of the computer and into the room. Bret projects from the ceiling to fill the room, Jeri from the viewer's glasses to fill the table. Although smaller, with Jeri's system the view is unique to each viewer. What I find most astounding about Jeri's system is that you can lean in close and see exquisite detail just like the MIT model railroad.

I've built a model of my employer's work that is incomplete even as it approaches a million cells. I view this as relationships, drawn as boxes and arrows, maybe up to 50 at a time. Any more and the picture turns into a jumble from which no new though is forthcoming. I catalog and parameterize these view so that one can peek into parts of the graph known to be interesting. I've made it possible to conjure new views with a little practice and share them between colleagues online.

I've built a similar system that can conjure diagrammatic views of the federated wiki hypertext. This works for viewing over the horizon to see pages you may soon visit or may pass by. The mere existence of these diagrams is known to mold the thought and writing of at least one author. We talk about this weekly and make adjustments to improve the work.

# Wiki

So this is how my distinction of model and document come around to the forward growth of our system. I said it would be a "data wiki" and explained that there one could do work as well as talk about doing work.

You should be thinking of Lisp's east coast way. The thing is alive and running in the system ready to be coaxed one way or another. In this wiki's first year, nine years ago, I made sure it could run and evolve the sustainable manufacturing models of interest to my sponsor and still meet the business interests of every member of their sustainable business association.

Over the years we have smoothed and improved this wiki's collaborative writing experience but less so improved its modeling capabilities. This I see as the challenge for next year and maybe a decade after that.

The most important move this year has been spawning independent but still connected implementations we call outposts. This creates the flexibility we need to run highly interactive, data intense and long running computations. It also creates a place to plug in Jeri's image projecting glasses to see deeper into the models we build and share.

The challenge for us is to shed the document way of thinking. I see it hobbling how we think of even our most powerful computers. We engineers speak of observability and then render what we could know as a "dashboard". Arragh. Imagine if we could see the datacenter as a model train, look at it real close, and then put the derailed locomotive back on the track.

Computers can respond to how we think. And the way we think depends on how we learn. Computer engineers have powerful computers close at hand. Management might believe engineers teach the computer how to think but it is often the other way around. Make something. See it work. Make something else. See it work. Over and over. This is constructivist learning. Kay to Piaget and even further back.