Where are the Physics Microworlds? Who’s been working on Papert’s vision? You can see them all around if you figure out the right phrases. Physicist Jakob Schwichtenberg’s Twitter feed is full of notes like this:
Michael Nielsen, the co-author of the “Mike and Ike” quantum computing textbook, has a series of essays at Cognitive Medium (see “Thought as a Technology” for an example) describing his attempts to build new tools for thinking.
Bret Victor is the modern prophet of this subject. He runs Dynamicland, a research lab that’s turned an entire building into an interactive computing environment. You write programs on pieces of paper, lay them on a physical table and, boom, they begin executing. This code, laid on the table:
When /bot/ is a *dragglebot*, /arena/ is a *arena*, /arena/ contains laser dot /dot/:
Wish (bot) has head position (dot).
Wish (bot) has head active (1).
Makes this happen:
When you’re done, put the paper in a binder and put it on a shelf. It’s a magic spell, waiting to be physically discovered by the next person.
Imagine writing down a description of a physical system — a "Lagrangian”, and having the physical system appear in dancing light on a wall of your room. Reserve judgement on whether or not you can “actually learn physics” this way. I’ll discuss that in the next post. But imagine how exciting it would be to think, “are these symbols really saying anything about the real world?“, and answering the question by bringing a new world to life from the symbols.
The Humane Representation of Thought
Bret Victor has a deeply powerful talk called “The Humane Representation of Thought” that has fleshed out so much of this area for me. The thesis of the talk is that our current tools for thinking are all heavily optimized for symbolic thinking. Paper makes it easy to create knowledge on tiny rectangles of paper using a pen and small motions of our hands. The printing press made it easy to broadcast knowledge on those same small rectangles… and now, in the 21st century, even though our tools make it possible to break out of the rectangle, we still stare at flat phones and monitors. We’ve never reconsidered the original design.
Compare this with a furniture workshop. Tools physically hang on the walls all around you. The chair you’re building has a smell, a feel. The goal is to make the chair feel right. You can walk around it, sit in it, use your kinesthetic sense of how the wood is responding to gauge exactly how much wood you’re shaving off with each pass of the plane.
The shop is dangerous because you can be more effective than you’re used to, immediately. There are no prerequisites to using anything in the shop. Maybe to using them safely! But not to getting started. And you can sort of figure out what everything does by looking at it.
Back to the chair that you’re building. How do you know if you’re done? If it looks right, it is right. Think of the brain machinery that has to exist for a collection of sticks to fall on either side of the “chair” or “not chair” line.
We are heavily shorting ourselves by forcing knowledge work and learning to happen in environments where all of that machinery is left running idle and atrophied.
If you are good at this stuff, you’ll think: “But isn’t it a beautiful thing, that we can develop new senses and modules in our brain? That we can develop the ability to feel mathematics, and get a glimmer of this sense that an equation ‘looks right’?”
It absolutely is. But this isn’t how most people feel. I was having a conversation with a friend recently and asked, trying to figure out where to start: "do you know what a partial derivative is?"
I could almost feel his soul tighten up and his skin get clammy. “Fuck… is this a quiz? I should know this…“ Almost everyone has this reaction to the math and science they went through in school. Imagine if you asked a kid to grab a broom out of the broom closet, and at the words “broom closet”, their eyes got wide and they started shaking. You’d go on red alert! “Oh my god. Was this child abused?“
Or if asking a friend to read you a restaurant menu got them sweating. Just because some people can work with the system doesn’t make it humane. Something is deeply wrong with the way we teach and learn and communicate research. It’s fine if some tiny percentage of people go into math and science professionally. But why does everyone else have to leave with such trauma?
Look around the world today and see the devastation that comes from the fear and anxiety that people feel about numbers.
Here’s Bret Victor’s talk, “The Humane Representation of Thought”. Please, take an hour to watch this and imagine what the world would look like if we went down this road together.
“If these tools are so obvious, why hasn’t anyone built them yet?” We’re in a difficult place, where the tools that Bret Victor describes take serious programming chops to build. The code that runs Dynamicland doesn’t seem easy to build in a Dynamicland environment. Dynamicland isn’t “self hosting”. So we have to build the new world using old tools.
To make better tools for building something like a Microworld engine, you have to know:
how to build software tools,
at least a few of the domains where the tools will apply. (It’s hard to build a workshop without having a project that needs a workshop.)
Many of the best research scientists seem to be caught in the grip of an incentive structure that rewards the production of PDFs compatible with a 15th century printing press. This is backwards compatibility taken to its insane extreme. Michael Nielsen, in “Reinventing Discovery“, describes how this incentive structure means that any time spent building tools to explore existing ideas is time wasted for a career.
Tool building isn’t explicitly rewarded. And then people like me, who know how to build software tools, don’t have any incentive to try and teach their math-and-science hobbies. We produce tools like TensorFlow and PyTorch for others to use.
My optimistic belief is that this odd swirl of bad incentives means that this area is a powder keg of intellectual acceleration, ready to explode.
I’m sure I’m going to learn more about why this has been so slow, and how to apply leverage, but for now it’s clear that the path forward involves:
Building tools that make it easier to produce interactive Microworlds
Actually producing on the order of 50 of these Microworlds in a form that others can steal and use and try in their own work.
In the next post, I’ll discuss, finally, the concrete project I’ve taken on in this direction, and some ways you can help me if you find all of this interesting.
Come say hi in our Discord chat room, and send me an email if you’ve been exploring similar ideas and are looking for ways to work on this. I’ve been assembling a small network of people excited about these ideas, and I’m going to start connecting folks who are feeling alone in this as we reach critical mass.