The Ghost in the MP3

I’m a sucker for procrastination by way of jumping down the rabbit-holes of new curiosities (‘How’s that novel coming along, Matt?’, whispers my better self…) — & lately I’ve been sort of obsessed with the concept of AGI (“Artificial General Intelligence“), which is slightly more broad than the specific kinds of AI that populates our apps & internet searches.

The entire concept of translating human communication from linguistic concepts & human interpersonal quirks into maths & graphs & code *really* ticks some nerd boxes in my brain, & if we bumped into each other at a party right now I’d no doubt ramble on trying to explain why I find NLP (Natural Language Processing) & “word vectors“, for example, to be fascinating ways to consider the relationships between words, language, & in turn, our thoughts.

While foraging for context & more information (always moar!) I started skipping through the interwebs & landed on the DATA SKEPTIC podcast, & was taken with an episode called “The Ghost in the MP3“, which is our real topic today… how the sounds we hear can be converted to numbers & code, then translated back again depending on the situation.

That ep (from all the way back in April 2015) features an interview with Ryan Maguire, who created a project in which he rescued all of the data “lost” when the mp3 compression algorithm compu-magically reduces the size of a music file to make it more iPod-friendly. (Or Zune-friendly, if that’s how you roll.)

You can’t compete with vinyl, man…

[Short explanation for the non-technical: Bits of wavelength & frequency are removed when converting audio to the more ‘portable” mp3 format. It’s a lesser version of the sound vs the “lossless” files that have higher fidelity to the original recording, but hopefully in ways that your brain doesn’t notice. . ]

Since you’re likely listening to these on shoddy earbud headphones anyway…does it matter? That’s one of the topics Ryan address in the project, where he created a sort of isolated ghost file from only that excised audio data, so you hear exactly what you’ve been missing. There’s some nice background on what range of sounds we humans can hear, & how your senses (or lack thereof) affect your perceptions.

Follow that link about & give it a listen. And don’t be scared away by the math & programming jargon that pops up, you can get the idea without understanding any of it. ..

~ Matt

Linked: Ryan Maguire @