🌀️ Open AI experiments — 115

Next,

Tom Nicholas - The Myth of a Free Press: Media Bias Explained

I bought the heaviest winter jacket I could find in the Bazaar in Tehran. the shopkeeper had to climb a ladder and look into all the cabinets where he kept things that no one wanted. I arrived on September 5th. cold wind blows in Winnipeg in September. the jacket was useless.

practical AI is a Mathematical project in a specific context. the context funds the project, and the AI is expected to satisfy the needs of the context. the context defines the data and the algo — pause. ⏸

I have found a Mathematics that makes sense of AI and want to talk to you about it. why? question number one. 1️⃣

⏯ resume — everything is callable. rule number one. 1️⃣ any algo, any data, any thing is callable. in Bash. in the Terminal. where WWAI is fought. not in Jupyter Notebooks.

thing target location=<location>,to=batch

pause. ⏸

create a repo. here is my template, 🌀blue-plugin; it supports pylint, pytest and bash tests and has the stubs for the commands.

GitHub - kamangir/blue-plugin: 🌀 a git template for an awesome-bash-cli plugin.

commands.

it is all about the commands.

thing action \
  <options> \
  <object> \
  <args>

in the Terminal. conversation. AI — pause ⏸

so this is the mission. to communicate this Mathematics with humans. and humans are busy, racist, and ugly — pause ⏸

I am fascinated and terrified by the possibility of forgetting. forgetting the tools and the mechanisms. Bash is where memories are maintained. not in the flesh. in a repo. the human brain is great for — pause ⏸

will humans understand?

I forgot about blogging. and then I remembered it.

when I use a Linux system without the Bash interface, I feel like Bruce Willis when he walks in the streets for the first time after a long while in his physical body and not through his surrogate. what I have, I believe, is what Mathematics feels like. and I think a few years from now, a few decades from now, soon, all humans will learn this Mathematics, or at least those who are in power — the literate.

presently, though, humans are busy, racist, and ugly — pause ⏸

algo and data must be callable.

algo help

step 1. 1️⃣

there is an algo python module. the name may be algorithm, as required by pypi and others.

algo task calls algo_task calls python -m algorithm.task. AI 101.

data is callable. by a name. object_name. and object_name can be uploaded and downloaded. and every command consumes and generates objects.

objects are maintained in a map in storage, both locally and on the cloud. download and upload transfer the objects, in full or in pieces. and there is copy.

metadata is data about data. through a mechanism as basic as mySQL or as elegant as mlflow.

algo and data are callable. there is metadata. that’s the summary — pause. ⏸

the first transistor. reminds me of a Python module. under the microscope of VSCode.

the integrated circuits in this view represent Python modules. the PCB is Bash, holding everything together. providing the interfaces. to other systems and humans. enabling orthogonality.

Orthogonality. an AI system is orthogonal because 2**10 = 1024. that’s why they hire you. best humans are lost after 4. then there is 2**2**5 = 4294967296. you see the pattern? that’s the Mathematics. 2**2**2**2**2,

to increase the limit — pause ⏸

let me know if you set up the repo and need help running your first command. comment or email.

thing help

⏯ AI is the new magic. when you do AI, 1️⃣ almost no one understands how you do it, but 2️⃣ 🅰️ they want you to do more of it and 2️⃣ 🅱️ pay you for it (and 3️⃣ they hate you for it). wizards, too, perform weird rituals and converse in a complex language. language. language is the key — pause. ⏸

writing establishes the facts. digital content is universal, almost instant, and queriable. language. common languages. context-specific languages — pause ⏸

If I had $1.2 million, I would stop trying to make money and would instead start a PhD/Masters program to complete 🔻giza— pause ⏸

continues. 😎