Much has been said about the harmful effects of
YouTube’s recommendation algorithm, from Zeynep Tufecki’s
We’re building a dystopia just to make people click on ads
to James Bridle’s Something is wrong on the internet.
While these are very important (and disturbing) analyses that
need to be acknowledged, this post isn’t about that. It’s about
my personal experience with YouTube’s recommendation engine,
which has been overwhelmingly positive, largely due to the kind of
content I constrain myself to watching while on the site.
A few years ago, when running became difficult due to joint
pain, I started exercising on an elliptical,
a strange exercise machine that I originally mistook for
a stair climber.
The elliptical is strange to me because it doesn’t feel
particularly physically taxing, yet I sweat a lot while
I’m using it, so I guess it must be doing something
good. And because it’s not very intense, it leaves my
brain free to absorb content. I noticed other folks at my
gym doing all kinds of things while on these machines,
from taking conference calls to reading books.
So I decided to start watching tech talks. Mostly
GDC talks, PyCon talks and talks from various
talks folks on the 18F Slack recommended.
Because my gym had a very spotty internet connection,
I had to pre-download talks onto my phone with a public
domain command-line program called youtube-dl. Eventually
I made a simple desktop app to ease this process.
At some point I switched gyms, and the new one had a much
better internet connection, so I dropped the pre-downloading
step. Instead, I just watched the videos using YouTube’s
app, which gave the site much more data about what
I liked to watch.
After some time, I noticed that YouTube would start
recommending tech talks to me. I had never heard of
GOTO Conferences or the Code Sync family of
conferences, but YouTube introduced me to them, and I
found their talks worthwhile.
At this point I’ve discovered so many interesting channels
of educational content that it’s hard to remember which
ones YouTube’s algorithm introduced me to. I am pretty sure
the following were, but they’re worth mentioning regardless:
3Blue1Brown is an amazing channel that makes
mathematical concepts intuitive through the use of ingenious
visualizations. Its Essence of Linear Algebra series in
particular is one that I wish I’d had access to when I
took Linear Algebra in college.
Wendover Productions is a channel whose content seems
quite diverse, from detailing how forest fires are fought
to how Mount Everest is climbed to how the geographies
of various countries affect their trajectories.
Vox videos have very slick production values and
sometimes I feel like I’m watching a geopolitical thriller,
which feels bad ass. When it comes to political
content, I feel especially wary of potential bias,
but I also appreciate how Vox makes some conflicts
that have always eluded me easier to reason about. Their
The Middle East’s cold war, explained and Johnny
Harris’ Vox Borders series have been particularly
AI and Games is a channel that I think may have
been recommended because of the Procjam and
Game Maker’s Toolkit videos I’ve watched.
City Beautiful is an excellent channel about
the design of cities and urban planning. I think its
An Urban Planner Plays Sim City video may
have originally been suggested to me because of the other
game development-related videos I’ve watched, but I’ve
always been interested in urban planning too, so I appreciated
So, while I fully acknowledge that YouTube’s recommendation
algorithm can be harmful, it’s also been enormously useful.
That said, it has suggested some questionable things to me in
the past, and I’ve had to be quite vigilant about what I watch on
it, for fear that my recommendations will go awry. I suppose
this is its own problem, but despite it all, I’m grateful
that the algorithm has introduced me to a variety of content
creators that I otherwise wouldn’t know about.