Cookies   I display ads to cover the expenses. See the privacy policy for more information. You can keep or reject the ads.

Video thumbnail
So far we've built a neural network, learned about imaginary numbers, and trained a decision tree to recognize fingers and images.
Along the way many of you had really nice things to say.
Some of you had weird things to say and quite a few of you asked some really good questions and made some really great points.
You translated Welch Labs videos into 17 languages, used the videos in your classrooms, and dug into the supporting code.
You've shown me what I really hoped could be true about the world.
That in an age where an explosion of human knowledge has led to reductionism, or some of the most beautiful important and useful ideas
humanity has wrestled from nature are often boiled down to meaningless abstractions, you've shown that
there's still value and depth. The real understanding is worth investing in.
I wouldn't be making videos if I hadn't first been inspired by some incredible math and science YouTubers.
And I think the incredible response these channels and others, and to a much smaller but very real extent Welch Labs have received, showed
just how much you care about deeply understanding mathematics and science.
I can't express how optimistic this makes me about living on this planet.
Thank you.
Now, what's next for Welch Labs?
After careful consideration, I've decided that... I have no idea. Well, that's not entirely true.
I actually have a whole bunch of ideas.
But if there's anything I've learned after doing this for a couple years, it's that creating a YouTube series, like any good journey, is not exactly predictable.
You see, over the years, I've developed a highly refined creation process.
It all starts with research. This usually goes really smoothly and definitely never leads to feelings of crippling inadequacy.
After research it's time for writing. This generally consists of waking up early in the morning and making coffee
and staring at a mostly empty Microsoft word document for two hours.
After a few days of this I realize that I have no idea how the very simplest and most
fundamental part of what I'm trying to explain actually works.
So then it's back to the books.
Except this time instead of pretending to understand what I'm reading, I have to actually learn something.
This generally turns out to be much less pleasant than the pretend learning I was doing earlier.
After doing this for a while, it's back to writing where I quickly learn that I still have no idea what's going on.
This research writing cycle invariably goes on for a while until I realize that
my entire initial approach and conception of what I was trying to explain was fatally flawed.
If I'm lucky, along with this realization of just how bad my original approach was, comes the kernel of a new idea.
When I set out to write the imaginary number series, I actually didn't.
It was supposed to be a series on the Fourier transform.
But after I realized that I had no idea what the i was really doing in Fourier's equation,
I stumbled into what turned out to be a really fun series on imaginary numbers.
Anyway, the point here is that after some more research writing death spiraling for a while, sometimes
I'll be left with a half-decent script. I'll then edit this for a while and once I'm sick of that
it's time to make a video.
Video production starts with storyboarding and recording narration.
Narrating is super easy and I basically always get everything right on the first...
Storyboarding isn't too bad. The real issue is thinking of visuals that I actually have the skills to create.
After storyboarding it's on to shooting.
For simpler shots, I'll shoot in a light box with a few clamp lights and a simple overhead camera setup.
I completely stole this idea from MinutePhysics creator Henry Reich. Hopefully that's cool.
For larger shots, I shoot on this motion rig thing I built.
Something I learned early on is that, if you don't want your videos to look like a low budget hostage
ransom video, you really need good lighting.
The rig is lit from three sides with high CRI LEDs and overhead with a ton of CFLs.
For static shots, I use an overhead camera mount, and for fancier shots, I use this five axis motion rig I built from camera sliders.
The motors are controlled from an IPython notebook.
After shooting, it's on to color correction and editing.
Maybe I'm just really slow, but when I sit down to edit,
I'm pretty sure the fabric of space-time conspires against me to ensure that at least six hours will pass
before I have anything reasonable to show for it.
I'll then desperately review my completed video in search for errors
typically only to discover that I've missed something so painfully obvious that someone is able to point it out
within one minute of me uploading... somehow.
So as you can see, the Welch Labs video creation process is a fine-tuned well-oiled machine.
Next time, we'll dig deeper into this amazing creation process.
Just kidding! A multi-part channel update would be lame.
All right back to the question I've been avoiding. What's next for Welch Labs?
Well, a couple things. First, a new series.
Since spiraling through a few research writing cycles seems to be a necessary condition for making videos,
there won't be a new video for a little while. Let's say we'll start the next season in July.
And since it takes some time to converge onto a good topic, we won't be picking an exact topic just yet.
However, I do have a direction for you.
I really enjoyed creating the imaginary numbers series, and a topic I really, really wanted to get to was applications of imaginary numbers.
These applications are often connected to two other fascinating topics: Euler's formula and waves.
So the direction I'm heading for research is the intersection of these topics.
Secondly, today I'm launching a Welch Labs Patreon.
By supporting Welch Labs, you can help me make more and higher quality videos.
Your support is much appreciated.
Alright, thanks for watching and I'll see you in July.