Tag: machine learning

Brain Food

Visual Dispatch #6

A photograph of two bits of paper. The first has "How do you want to feel?" written across the top with a list of adjectives below. They include connected, fulfilled, stable, lauded, validated, helpful, effective, accomplished, calm, grounded, proud, challenged. Below that are phrases like "Survival without worry" and "Creating empathy in others". The page on the right lists five categories "Financial, Emotional, Internal, External, and Visible".

I was emptying out my backpack this morning and a bunch of bits of paper fell out of a hidden sleeve. There were postcards from my trip to Hawai‘i in 2017, thumbnails from the comic about Federated Learning I made for Google in 2019, and a couple scraps of paper containing (I’m fairly sure) notes for this talk on success—also from 2019.

I wish I could remember where the lines in quotation marks came from. Maybe they were responses from the participants at that workshop? Maybe they came up before I even gave the talk? Either way, it reminded me how much I’d poured into those two presentations (the other was about money, surprising no one), and how glad I am to have recorded them.

I’m stuffed up with a head cold this week and feeling very sorry for myself, so this video window into a me who was, well, with it feels really validating. So much has changed and twisted and refined in the past two years. I feel smaller in some ways—like the circles of influence I previously occupied have shrunk considerably—but also deeper. In the five spheres of success above, I feel as if I’ve traded Visible, External, Financial, and possibly even Internal success for something deeply Emotional. And I’m still trying to let the dust settle on that decision and learn how to make peace with it.

The lesson remains: these questions are never truly answered. Not really. Not if I’m doing it right.

Blog, Reading

Literary Archipelagos

Last week, in a moment of Peak Bellwood Weakness I signed up for an online class/study group called Literature at Sea: A Brief History of Existence. The facilitator shared something in today’s intro call that I can’t believe nobody sent me when it was released back in July. It’s called An Ocean of Books and it looks sort of like this:

A screenshot of the homepage for An Ocean of Books. It shows a pale greenish blue chart with a mass of tiny, tan islands spread across it. They're loosely grouped by subject: History, Science, Novels, Classics, etc.

This “poetic experiment” was made by Gaël Hugo during his time as an Artist-in-Residence at the Google Arts & Culture Lab. It pulls from the entire Google Books library and uses a bunch of (I’m waving my hands vaguely here) technology to generate a chart of Author Islands whose distance from each other is determined by their relationships on the web.

The site’s a little awkward in places, but I find the whole concept delightful. The weird aesthetic mix of pixelated game art and old nautical chart elements!1 The playful mechanism for revealing keyword searches within a bank of fog! There’s also little factoids beside various islands, like this gem about Maurice Sendak:

A screenshot from A Sea of Books showing a drawing of a boy riding a horse. The text beside it says The original title of Sendak's famous book was Where the Wild Horses Are but he couldn't draw horses. So, when his editor asked what he could draw, his reply was Things.

Anyway, I spent a lovely afternoon poking around in here, but what it really got me hungry for was a similarly attractive way to organize one’s own library for others to explore. The trouble is that I’m just not moved by reading lists—even ones curated by subject. I’m a visual thinker, and I need to make a big mess and tack a lot of red string to the wall before I can truly understand how all these ideas are contributing to the electric pinball machine.

I don’t want the map to be dictated by an algorithm; I want to play cartographer.

I rediscovered a piece of technology this week that might hold the key, but I’m saving it for now. You’ll just have to wait.

1. Fun pedantic terminology fact: if it’s to do with the ocean, it’s a chart, not a map. Yes, there will be a quiz on this later.↩
Comics

New Comic: Federated Learning for Google

Hello, blog friends!

I know things have been pretty quiet around here lately, but that’s mostly because I’ve been tied up making this new comic for Google!

A header image titled “Federated Learning: Building better products with on-device data and privacy by default. An online comic from Google AI” A smiling woman and an excited corgi flank a row of smart phones.

Federated Learning is a new field of machine learning research that just hit the big-time at Google’s developer conference this week. I landed a gig working with Scott McCloud and an internal team to translate the basics of the field into an explainer comic.

Most importantly: it’s got a Moby-Dick joke in it.

You can read the whole thing here.

There’s a lot to talk about with this gig, and I’m going to be diving into what I learned from my first major corporate client experience over on Patreon. This job is the most lucrative freelance contract I’ve ever taken, and I want to talk about how that’s felt (Complicated! Emotional! Empowering!) and what other folks can do to pursue similar gigs. I’m really proud of what we produced, and I think it does a good job of explaining something I would’ve never otherwise learned about.

I also took lessons learned from building the accessible edition of 100 Demon Dialogues and made sure we had functional alt-text throughout the comic. If you use a screen reader to browse the web, do let me know how it works for you. I’m still trying to get better at making this site and other projects I work on accessible for folks who traditionally get left out of the digital comic-reading experience.