When I was your age I had to draw hands to point at things
Hi friends,
(welcome to my email dispatch! You can sign up for these or read the archive at buttondown.email/thesephist 💌)
The arrow symbol (”→”) was invented in the late 18th century. Its earliest known use is in 1737.
This is a remarkable fact for a couple of reasons:
Somebody had to invent the arrow symbol. This ubiquitous, globally recognized piece of visual vocabulary had to be INVENTED! In its first use, it still looks more like an archer’s arrow than a symbolic one.
The arrow symbol is only about 300 years old. The invention of the arrow symbol is closer to the discovery of the electron than Newton’s theory of gravitation, which also means that when Newton invented gravity and Leibniz invented the modern notation for calculus, THEY DIDN’T KNOW ABOUT ARROWS! “∫” is older than “→”.
Now, you might be asking yourself, “What did they use to point to things before the invention of the arrow symbol?”
They used an intricate drawing of a hand with the index finger pointing. Like: 👉. Check out any map before the 18th century with a pointer, or table of contents and margins of old books.
I routinely bring up this bit of history in conversations about designing information interfaces because I think it demonstrates two really powerful ideas:
If a new piece of notation is truly useful, it can spread very, very quickly. It can expand language in a very real sense.
New notations are not at all obvious, and it’s quite possible we have yet to discover the most useful ways of expressing ideas outside of our minds.
I spent much of the last two weeks thinking deeply and reading broadly about data visualizations, which aren’t quite as simple and remarkable as arrows, but nonetheless a ubiquitous and modern addition to our notational vocabulary for ideas. Data visualization is a technology, in the sense that language and writing are technologies.
This week, I want to share with you my favorite links and resources about tools for seeing information in new ways:
What I read
Josh Horowitz has done phenomenal work in the space of data visualizations and interactive explanations. Check out his website index or his gallery of concept visualizations.
Google Brain’s Big Picture Group also did lots of phenomenal work, listed here. Among them is a classic interactive, direct manipulation UI to understand how a small neural network learns, called TensorFlow Playground.
Martin Wattenberg, an ex-member of the Big Picture Group, has also collected many great examples of interactive data graphics on his website here. I see Wattenberg’s name often in many of my favorite papers studying how humans understand and visualize ML models’ outputs.
Though not online reads, there are a few books I’ve come back to often as classic references in this field:
Edward Tufte’s The Visual Display of Quantitative Information is a classic work on data graphics, especially in an era of print design.
Giorgia Lupi’s Dear Data is a beautiful collection of more creative examples of hand-drawn data visualizations.
The maps of US National Parks are beautifully crafted, and manually designed each time. Felt has done a great interview with one of the early contributors here.
Michael Nielsen’s Reinventing explanation is a great high-level review of many of these big ideas about visualizing information.
What I’m working on
I gave a demo of a design prototype for a “computational notebook for thinking with language models” at South Park Commons in New York. You can watch the talk here, or read through my blog transcript here.
I wrote about how I think about exploratory, open-ended work like research or interface prototyping after some conversations with friends. It’s called A beginner’s guide to exploration, and talks about balancing vision with execution and how I think about long-term motivation. I’m quite proud of how it turned out, but astute readers have pointed out:
Research often happens in a collaborative environment inside organizations, where there are other factors beyond your control that I don’t account for. (tweet reply)
I don’t really talk about the role of funding and attracting resources for exploratory work, though my friend Anson helped write this great piece related to it.
Patrick on Twitter pointed out another factor in cycles of transformative innovation often depends on a tick-tock cycle between better instruments (letting people see further into nature) and better science (motivated by those insights, and enabling better instruments).
On the more technical side, I’ve begun sketching out how a real implementation of the “computational notebook” idea from above may look like, both technically and visually. I’m zeroing in on a few workflows initially to motivate my prototypes:
Exploratory data analysis, specifically understanding LLM training datasets
High-stakes decision support, specifically medical diagnoses and legal decision making, maybe also early stage investing
Literature review, perhaps something similar to what Elicit does
A reminder: you can reply to these emails :-) They go straight to my inbox.
Wishing you a happy and safe week ahead,
— Linus