A Few Things To Start The Year
Vintage photos, generative AI, and slogging through learning music
The Teardown
Saturday :: Jan 7th, 2023
Happy New Year to everyone. In this newsletter, I wrote about a few articles that highlight lots of interesting topics that I’d like to write about throughout this year.
Vintage Photos Are Back, Again
The New York Times ran a piece describing Gen-Z’s latest hot gadget: the analog point-and-shoot camera. The beginning of the piece nicely captures the appeal:
Last spring, Anthony Tabarez celebrated prom like many of today’s high schoolers: dancing the night away and capturing it through photos and videos. The snapshots show Mr. Tabarez, 18, and his friends grinning, jumping around and waving their arms from a crowded dance floor.
But instead of using his smartphone, Mr. Tabarez documented prom night with an Olympus FE-230, a 7.1-megapixel, silver digital camera made in 2007 and previously owned by his mother. During his senior year of high school, cameras like it started appearing in classrooms and at social gatherings. On prom night, Mr. Tabarez passed around his camera, which snapped fuchsia-tinted photos that looked straight from the early aughts.
“We’re so used to our phones,” said Mr. Tabarez, a freshman at California State University, Northridge. “When you have something else to shoot on, it’s more exciting.”
Of course, Gen-Z’s aren’t shying away from their phones entirely either. The article describes hashtags and viral posts on TikTok and Instagram highlighting these blurry and somewhat poorly-exposed photos. After all, does an analog photo really exist if you don’t rack up social capital by posting the photo - perhaps slightly misaligned - on social media?
The article made me laugh as we’re hardly in the first iteration of this resurgent grainy-photo movement. Instagram’s early allure was exactly aligned with these sorts of photos. The app didn’t require another device. It simply made your existing higher-quality digital photos look as if they were taken on a 1970’s-era camera.
That said, the analog vs. digital topic reminds me of a conversation from my holiday break. A family member asked for a solution to a common problem: what do to with thousands of photos and videos captured with phones and other digital cameras over time. Should you keep them in the cloud? Catalog them one by one as you take them? Print digital copies of only those that mean something? There’s no right answer of course. But perhaps the trend toward analog or non-smartphone cameras will cause the next generation of adults to capture only those photos that really matter.
Ending 2022 And Starting 2023 With An AI Boom
Many of you have heard of ChatGPT by now. The wrapper around OpenAI’s GPT-3 language model arrived roughly five weeks ago and have since stirred up opinions across the internet and in analog life. Schools are banning it because it enables easy cheating. Software developers are rejoicing over the notion that they can use it to script boilerplate code without typing more than a sentence.
Generative AI, the wrapper term covering ChatGPT as well as other tools, is a fascinating world. You ask a tool to generate something based on a prompt phrase (or more), and the tool produces output based on the prompt. For example, I recently asked ChatGPT to write the code for a to-do-list app in HTMX, a markup language for building interactive websites.
My prompt was this:
Write me a to-do list application with a frontend in HTMX
ChatGPT then “wrote” the following code:
<!DOCTYPE html>
<html>
<head>
<title>To-Do List</title>
</head>
<body>
<h1>To-Do List</h1>
<!-- Add new to-do item form -->
<form htmx-form="post" action="/add-todo">
<label>
<span>Task:</span>
<input type="text" name="task" required>
</label>
<button type="submit">Add</button>
</form>
<!-- To-do items -->
<ul htmx-container="todos">
<!-- To-do item template -->
<template htmx-template="todos">
<li htmx-replace="li" htmx-class-done="done">
<input type="checkbox" htmx-click="POST /mark-done" htmx-value="done" htmx-target="done">
<span htmx-replace="task">Task</span>
<button htmx-click="POST /delete" htmx-confirm="Are you sure you want to delete this task?">Delete</button>
</li>
</template>
</ul>
</body>
</html>
It also provided some tips as to how to structure my back-end (i.e. server) infrastructure to both accept and push data to the to-do list:
The server should handle the following routes:POST /add-todo: Accepts a new to-do item in the request body and adds it to the to-do list.POST /mark-done: Marks a to-do item as completed.POST /delete: Deletes a to-do item.For what it’s worth, I haven’t verified whether this code works. But the model’s ability to regurgitate what looks like working code is certainly impressive. You might say that ChatGPT exudes confidence, and that’s part of the problem. It’s confident in its answer because it provided something, but there’s no guarantee that the answer is correct. In other words, with code, an example of a “right” answer is a piece of code that works as intended when executed. When you ask the model for the name of the current president of the United States, it should answer “Joe Biden.” But there are lots of questions and other requests that will produce answers that seem right but require further inspection.
The other problem, at least with some models, is their perceived bias. ChatGPT famously avoids controversial requests. I’ll use a simple example: you ask the model if the n-word is racist and it produces this:
Acknowledging the content policy via the button simply erases the question, despite the lack of an authoritative racist statement.
We can all of think of questions that we probably don’t want it to answer, but the opinionated nature of response raises eyebrows. Should AI decide what’s right and wrong? Should AI decide what can be asked and what’s inappropriate to ask? I don’t think we’ll ever truly settle on an answer, but I’m interested to follow how these questions are addressed or not by generative AI providers as the space evolves.
Music Lessons Are Hard
J.R. Patterson wrote an emotive piece for The Walrus discussing why many kids and adults abandon their musical instruments and lives. He describes how he learned violin and his frequent disinterest in the process:
I think of Dupea when I remember my childhood violin lessons. I loved music, but something about the methodological lessons rankled me. I was a pupil within the Suzuki method, which values learning by ear, memorization, and parental involvement.
…
His solution was nurture over nature; his evidence was language—an infinitely complex skill learned naturally by almost everyone. If there were a secret to mastering an instrument, it was only “practice and practice of the right things.”
…
But, for me, this relatively soft teaching style was outweighed by the formality and uniformity of the method. The rote memorization and emphasis on practice over anything else left little room for improvisation or imagination. It was the music equivalent of conjugating verbs—alluring to the pedantic grammarian but not to someone who wants a more streetwise edge to their dialect. There was no part of me in the sound I was making; even when playing well, I was only a vessel for the music between the page and the air. Each week, from the age of six to seventeen, I’d spend the hour-long drive it took to reach those lessons imagining events that might preclude me from playing that afternoon—and perhaps forever: I’d fall suddenly ill; I’d realize I had forgotten my violin at home; there would be a pileup on the highway. As we got nearer, I’d wonder if I could make jamming my fingers in the car door appear like an accident. When I inevitably arrived safely at the class, I stumbled through the arpeggios I’d neglected to master between lessons, and the disappointment on my teacher’s face was less a catalyst to buckle down than an indictment of everything I had already failed to become.
Like J.R., I too learned to play over years of lessons. I first started on saxophone in fourth grade, roughly around age nine or so. I wasn’t forced into an instrument like many kids. Instead, I was somewhat naturally interested coming from a musical family, with a parent that played piano.
Somehow, I started on saxophone and quickly dove into formal lessons. They focused on a lot of familiar and classic techniques: understanding scales, reading music, working through the dynamic range of the instrument, and etc.
Much of my practice work was those scales, both to internalize and build dexterity. And much of my practice music was some variation of classical genre. I quickly gathered skills and became “good” at saxophone, performing as the top student of my teacher Valerie.
Over time, I met other players in school who were taught by another local teacher named Mark. Mark taught some of the same fundamentals but based everything on jazz rather than classical. His students - my peers in school - alway seemed to be having way more fun. Jazz was cool, loose, improved, and groovy. Classical felt rigid in contrast.
I eventually switched instruments, moving drums instead. I took lessons but not for long, instead choosing to teach myself what I needed to know. I was also choosing to have more fun. I often just sat down to play and didn’t worry much about what was coming out. I played drums regularly for many years and still play them today, albeit less given other demands on my time.
How did I keep up my practice? Well, in general, I was just interested to play music and kept at it. That interest allowed me to accelerate my skills. But the same isn’t true for every young learner, as J.R. details later in his piece:
Lessons are a considerable commitment for parents; they were for my mother, who worked full time, paid the fees for my sister and me, and spent countless hours driving us to tutorials and concerts. But the parental involvement entrenched in Suzuki brought the judiciary aura of lessons home and created a sticking point that lasts to this day. Tears were shed over the attempts to cajole me to practise. Time sheets, bribery, nothing could persuade me to engage in what felt like a buttoned-down, dot-joining activity that turned leisure into duty. (I did feel moments of liberation whenever I went off piste and played a little Fritz Kreisler.)
Many kids feel forced or are explicitly forced to play. Fortunately, I wasn’t forced. But another important factor was the commitment of my parents, in particular my father. He was interested in my progression in a non-forceful way. He wanted me to feel confident and encouraged me in all sorts of ways, whether it was visiting stores with old saxophones or calmly asking if I wanted to play along with music he was playing across the house.
I’m in the early part of this progression with my own kids. I don’t force them into music, but I try to make it an interesting part of life. We dance to music. I play drums for them. I play guitar (poorly) for them. And I let them mash on keyboards when they those instruments are nearby.
I’ll love it if one or both of them pick up an instrument or more, and hope I can remember that inspiration rather than force will be best for sparking their skills and interest.