Vision statement

A poignant paragraph from MN’s early 2020 notes:

We believe we can ::create a studio:: that causes the creation of more transformative tools for thought. And perhaps even that ::instigates the creation of many new tools for thought, a novel practice of human toolmaking::.
Our sense is that the mnemonic medium wants to be something

  much more powerful. So we’re going to develop and scale it out
  slowly. In some sense much more research project than
  conventional SV product development. ::Long-run goal: a::
  ::transformative tool for thought that is used by hundreds of::
  ::millions or billions of people::. We may not ship that, but we can
  cause it. ::We need to make it something people _seriously want to::
  ::write in_::, i.e., something with an incredible value proposition
  for writers (which will mean an incredible VP for readers).

God, Michael’s so good at this kind of writing. But remember: he spent 17hrs on this document. In terms of actual focused time, I’ve spent much much less than that (though of course I’m benefiting from him having already written it).

I really like his “here’s what we’re doing” approach, rather than the abstract philosophizing in my attempt.

  • Can we make the mnemonic medium a 10x medium, for certain purposes?
    This needs to be true not just for readers but most especially for
    writers. What would it mean to succeed? We aren’t clear on this at
    the moment.

Michael’s really focused on building conviction. He wants to psych himself up, believe that this thing is absolutely both worth doing and is possible.


For about 90% of human history, written language did not exist. But then, just a few thousand years ago, that changed. Some small set of people invented systems of written language, and those systems radically expanded what humanity has been able to think and do. Writing isn’t a tool like the wheel. It’s a tool for thought, a technology which actually changes the thoughts people can think, individually and in groups.

I believe it’s possible to develop new tools of that kind—transformative tools for thought that radically improve how billions of people think and act. My research aims to create such tools and, ideally, to distill general principles for a new practice of toolmaking so that many others might routinely create such tools.

That’s admittedly quite a grandiose aspiration! But it is the aspiration. Let’s get more concrete.

In 2019, my colleague Michael Nielsen and I created a prototype Mnemonic medium—a medium which makes it much easier for people to remember what they read (instead of rapidly forgetting all but the gist). We build on the work of cognitive scientists, who have identified a formulaic set of steps you can take to reliably remember something. The medium integrates those steps into prose and automates them, so that people naturally reinforce their memory as they read.

Initial prototypes of the mnemonic medium show tremendous promise, but the medium is still in its infancy. I’m working to push it much further—in a sense, to discover what it wants to be. How effortless and reliable can memory become? What kinds of understanding can such systems support, in what kinds of reader contexts? How might we make it into a truly transformative tool, both for writers and for readers?

If we take such tools to their conceptual limit, memory could become almost effortless, a solved problem of everyday life. The exciting part here is not memorization itself, but rather the impact that memory has on everything you think and do. If you could effortlessly remember all the key ideas in a textbook after reading it once, you should be able to understand and apply complex subjects more easily, deeply, rapidly. The mnemonic medium illuminates a broad space of future systems which treat memory as a primitive: for example, future memory systems may accelerate creative work by helping people internalize key insights and influences.

To explore these topics, I’ve built a platform for memory system experiments (Orbit). It’s my laboratory, and I’ll use it (along with Quantum Country) throughout 2021 to better understand the mnemonic medium and what it might become. At the end of the year, I’ll publish a detailed essay describing what I’ve learned of the medium so far, including my best understanding of its limitations and challenges.

Key questions to explore and re-integrate into this vision document

  • What kinds of things should I do to explore…
    • how effortless and reliable memory can become?
    • what kinds of understanding such systems support?
    • how to make it a truly transformative tool for writers and readers?
  • What’s the ideal structure to do these things?
  • How do I need to develop myself to achieve these steps? What are the key things I need to learn? What kind of researcher do I want to become?
  • How do I get myself into a serious context of use?
  • Why now? What’s the opportunity?
  • Why me?
  • What’s missing? What should I be worried about?

The mnemonic medium is particularly well suited to platform knowledge, so I’ll focus much of my work there. Quantum Country is the flagship example, and it’s a good place to experiment with improving the scheduler and understanding reader behavior. But it’s difficult to run more substantive experiments since Michael’s no longer involved.

Short-term steps


TODO

::what steps am I doing to explore these topics?::
::how do I need to develop myself to be able to do that?::
::what kind of a researcher do I want to be?::
::but how? what’s the structure? what are my practices? where’s my context of serious use? what is the day-to-day, and how does it connect to this?::

::why now? what’s the opportunity? maybe that’s a topic for another document. I really don’t understand it well yet.::

  • my skills, my interests
  • the key things I need to learn
  • what’s missing? what big-picture thing isn’t here? what should I be worried about?
  • what would I like to do? be proud to do? what’s possible to do? do I really believe it? is it worth doing?

Exhaust

That’s an admittedly lofty goal. What kinds of tools might aspire to it? One way to think about tools for thought is that they alter the fundamental set of objects and actions available in a person’s environment. Different primitives make different thoughts accessible or salient. For most tools, the distinction is subtle or mundane. But the most powerful tools for thought introduce primitives which substantially shift the range of human thought. For example, written texts support non-linear reading—darting backwards and forwards as needed—which make it easier to introduce and maintain complex terminology. Hindu-Arabic numerals introduced an efficient place-value representation of numbers, enabling rapid and universal arithmetic operations. What new sets of objects and actions might shift the range of users’ thoughts so dramatically?

My current research explores memory systems—environments which build on powerful ideas from cognitive psychology to make durably remembering something into a primitive action. Prior work on memory systems (e.g. Spaced repetition memory system) has focused mostly on explicit memorization of simple facts, like country capitals and key dates. But memory-focused primitives can augment more complex and more meaningful tasks. I’m interested in how they might helpfully suffuse and augment the activities of everyday life—reading, writing, conversation, art. For example, I believe we can create memory systems which make it much easier to learn abstract, conceptual knowledge (Spaced repetition memory systems can be used to develop conceptual understanding). And I believe these systems can accelerate creative work by helping people internalize key insights and influences.

These are some of the topics I hope to explore first:

  • Augmenting non-instructional and non-technical texts
    • A quantum computing textbook is an ideal topic for memory systems: the context is explicitly instructional, and there’s a great deal of formal content to be learned. What value can memory systems provide in non-instructional, non-technical texts, like persuasive essays and informal articles? What are readers perceptions of memory systems in such texts? How should the mechanics of the mnemonic medium be modified to support such texts? How lightweight can they (and should they) be? (How might the mnemonic medium enable readers in genres outside platform knowledge?)
  • Characterizing the dynamics of practice and memory
    • Quantum Country’s relatively large user population offers an opportunity to understand the details of how long-term memory develops over time. Are there reasonably consistent functions which describe what happens when a piece of knowledge is reinforced? What are the high-variance inputs? That is, how do memory traces vary with content, timing, and prior behavior? To put this another way: in what ways can we use Quantum Country to better fundamentally understand memory and learning—and then to use those understandings to improve our memory systems?
  • Understanding the downstream impacts of memory systems
    • The big-picture potential of memory systems is best understood not by studying how well it helps people remember things, but rather by how well it helps people perform meaningful activities in their lives.
    • How does fluent memory transfer to performance on open-ended tasks in the subject? That is: does memorizing Quantum Country’s material mean you can explain topics in quantum computing to someone else? Solve problems you haven’t seen before? Create circuits for a purpose? What elements of the system promote these effects, and how might they be amplified? What are the limits of these effects? What kinds of skills are relatively indifferent to memory?
    • What is the impact of memory systems on learning downstream topics? That is, if you study chapter 1 with a memory system, can you learn chapter 2 more rapidly? Accurately? Deeply? Can you learn topics you wouldn’t practically have been able to learn before? What are the key interactions here, and what are their limits?
    • In what ways does fluent memory translate to insight generation? Behavior change? Artistic expression? How should memory systems designed around these goals differ from those designed for traditional learning performance measures?