I thought he'd transmit a PNG over a modem, get a bird to memorise that and play it back. I think with the right format it should be possible to do that. With enough birds I imagine you can store quite a bit of data. Takes saving to the cloud to another level.
>I thought he'd transmit a PNG over a modem, get a bird to memorise that and play it back.
That's essentially what he has done. Except he did the modulation/demodulation with audio software (and, technically, stored a monochrome bitmap, not a PNG).
Dial-up modems encode data in audio-frequency. Later modems used phase-shift keying¹, but the very early ones used frequency-shift keying², which is essentially encoding data in a frequency graph - i.e., drawing a line in a spectrum analyzer.
Drawing a bird in a spectrum analyzer is packing much more data than that; it's like playing several of those streams at once.
The bird has shown itself to be capable of remembering and reproducing multiplexed frequency-keyed streams.
>With enough birds I imagine you can store quite a bit of data. Takes saving to the cloud to another level.
It's analog though. Presumably the shape of the image matters, like horizontal lines are easier than vertical, it's not just a bitmap. He made the point of how many KB you can store in the song, but is it right? There are different conceivable ways to store binary data in that. I have no idea how efficient it'd be to get something 99% reliable.
He said 176KB of entropy in that 1-second birdsong, which doesn't seem close. That's more than the bitrate of a typical M4A, for a much simpler sound.
Thinking about it in reverse, how much data would it take to encode 1 second of birdsong in the most efficient audio codec I can imagine. If M4A or MP3 with the bitrate slammed way down isn't a fair comparison, then some birdsong-specific ML autoencoder... Probably 500 bytes? Would still be enough for a Twitter tweet.
Inspired by the video I vibe coded up an application that lets you encode data in FSK and read the data bits back from a noisy recording. I think it would be fascinating for someone to try this!
https://github.com/sequoia-hope/starling
The product recommendations at the end are gold too! A 'hacker' spirit there.
In terms of signal length, can you store the images/data in a flock of birds too? I wonder what the RAID set-up of a flock of starlings is like? I'm thinking something like the Tines in Fire Upon the Deep
More crazily, can you get these data signals to be Turing complete? I know that not really what data is like in a vocalization pattern, but can you manage to get the birds' vocalizations to do logic of some sort and change patterns in more than a non-entropic way?
What a ride. I kept waiting for the punchline to be something dumb, like "we fed a USB drive to a bird" or "we tied a recording to a bird," and then when I realized what they were gonna do, I was shocked. It should have been obvious, but it was very clever! Really cool.
It's not obvious at all that birds are able to reproduce multiplexed frequency-keyed streams with enough precisions to draw freaking pictures in spectrum analyzer graphs.
Humans can barely control their voices to nail one frequency. Badly.
These birds are capable of reproducing the output of several sound sources producing sounds at once, with near-perfect time and frequency precision.
Yeah, I also appreciate that I hadn't heard of this Youtubber before and I couldn't tell if he was a sound engineer who happened to make a video about birds or a bird guy who was playing with sound. Seems like both!
In all seriousness, like minds think alike, and Ben Jordan has a rebellious streak — he uploaded his own music to pirate websites when he broke up with his label [1].
The video shows that there are many ways to do that :)
Now, if we figure out how to convince the birds it's a predator distress call (which, if you think about it, it kind of is), we can probably get it to persist across generations.
this is what i've been wondering. is it feasible for a single diligent person to embed a message in an entire population of birds, in a manner that will persist generations? that's what i'd call a pretty good prank.
When this was posted to reddit, the comments were just full of people arguing over the semantics and saying how wrong the author was for using the word PNG when the actual technique is extremely lossy... completely glancing over the entire video, the dedication, knowledge and complexity involved with actually creating the video, the incredible feats of the birds themselves, and the reality that youtube basically forces you to use clickbait titles in order to get views.
I don't like the "PNG" part because it made me think he's storing arbitrary binary data. It's not even a matter of lossiness, cause these aren't JPG either, these are analog drawings. And it's not like this is overanalyzing the video, cause the author did talk about how many KB you can store this way.
At some time in the video he's casually played a groove on the piano to back the birds for a couple of second, then stopped ("Wait, what am I doing") :)
You can also see his modular setup in the background.
I didn't know of him until today. Instantly, a new inspiration.
Drawing into the spectrogram is a fun trick. I would really like to know how much data you can store in that bird using some digital modulation method such as FSK (frequency shift keying).
There could even be multiple carriers in the signal.
It would be even cooler if the bird were to preserve phase. Then you could use PSK!
> I would really like to know how much data you can store in that bird using some digital modulation method such as FSK (frequency shift keying).
The video shows the bird capable of remembering and reproducing 5-10 frequency graphs simultaneously (which you'd need to draw a picture), so you can multiplex those.
> There could even be multiple carriers in the signal.
Or same carrier, but different sets of frequency keys for each stream.
> It would be even cooler if the bird were to preserve phase. Then you could use PSK!
Maybe they do, someone should ping Ben to test that :D
The video covers a wide range of topics related to birds and audio. The title topic is actually only a small segment of the whole video, arguably not the most interesting one!
Among other things, he also covers (lightly) bird vocal anatomy, audio "cameras", birding apps & equipment.
I hate modern Youtube that now every single "serious" topic video is +30 mins length. 10 years ago we were perfectly fine with 10 mins stuff but of course algorithms and advertising and nowadays most Youtuber is pushing longer and longer videos as if we are watching peak evening television reporting...
I've heard this a lot, but is there confirmation of this somewhere? I'd actually expect a longer video to generate less engagement, cause fewer people want to watch all that.
At the end of the day though, the most money is made by getting people to actually pay attention to the ads. Shorter videos might really accomplish that best. I imagine YouTube is aware of sleeping users, and if advertisers aren't realizing the engagement they expect for what they pay, they'll go elsewhere.
Watched this video yesterday, and damn, it's really delightful watching experts make content about things they are passionate about. This love and passion is contageous, and even me, who up to this point knew almost nothing about birds has gained a new appreciation and love for these creatures. The fact that they can copy sounds is kinda incredible, and makes me want to listen more to them singing.
Wow, my worlds are colliding right now-- although seeing Benn on HN in retrospect shouldn't be that surprising. Go check out his music, The Flashbulb, he's one of my favorite artists.
This is really incredible. I love easter eggs/hidden things in spectrograms. The implications of this are really cool, regardless of whether is it lossy or not.
