The title is something I read on twitter on a long since deleted tweet, my memory faded about its original context - crucially I did write down and it somehow stuck - something I resist putting a datapoint on - a contradiction to an action - an artistic expression there to be studied.
It’s now the name of my journal, which is intended to be read by an informed audience interested in data, the music industry, ai and whatever it is we’re calling blockchains today.
A tool that helps studios recycle music from the cutting from floor. It has no front end, meaning you have to message it via Teams, Slack, Google Chat - making it cheap to build.
A former business partner reiterated to me that once you choose a frame work, you’re stuck with that stack forever. Can ChatBot’s help circumvent this or at least delay building one until you have customers?
AI Products
I’ve been interested in AI since 2018, principally as consequence of researching ‘blockchain’ technology. In a lot of ways I don’t consider blockchain a ‘human’ technology, more it is a tool for Artificial Intelligence to create automations.
My prediction is that won’t be until 2029 that we’ll be talking about the killer dApp that is getting 100,000 installs per day, all powered by this new wallet app from company that we’ve never heard of.
Until them, my ‘academic’ area of AI research is focused on ‘translator’ use cases for document summarisation and inference- ie how these system make predictions and ‘fill gaps’ - particularly with regards to legals in media production.
claim Infranodus gives you an overview into any discourse, revealing the blind spots and enhancing your perspective. Here, I show you how with some music metadata
Coda has an ‘AI Block’ that can classify and summaries values you place in it. Read how I used it to prepare for a job interview. Did I get the job...? No! But you can discover how I used the AI block to manage a GPT context window using a database and what insights into the company I was able to gain from it.
, an automation tool akin to Zapier and Make. I originally adopted it because you could use NPM packages in your workflows - and it marketed itself as an automation tool for developers - but its pivoted away from that slightly and grown more into a ‘backend-as-a-service’ - for example, it has an inbuilt SQL editor, which is vital incorporating it into any wider production grade system. Fundamentally though, Pipedream still takes the strain out of securely setting up OATHs between APIs, something I still find challenging.
Below I am documenting some of my scripts, more for my reference than anything as I recycle them a lot.
Script to determine an audio file’s type when stored in a Pipedream temp storage workflow.
This is useful for gatekeeping packages like FFMPEG, Music-MetaData... that our expecting audio but your input funnel might be contaminated with album art, press releases, *.ini’s...
I’ve found the mime-type package to be a worthy CPU trade off to writing my own javascript as this handles uncommon propriety audio types (eg m4a = audio/mp4) correctly.
Annoyingly I’ve struggled to identify mime-types from URL streams & file buffers in the Pipedream environment.
A lot of audio processing favour CLI’s and lower-level languages like C++ - those higher up typically require inputs in an uncompressed format, so its a handy script for working with audio in Pipedream’s environment.
that let you convert CSV files to ALE - however, this script is expecting to receive XML from a JSON array.
This is useful for programmatically preparing a collection of media assets for ingest into Avid, as you can cache everything and write the ALE only once before distributing your files over to your post-facility.