It's about time, long back when first time i had heard about machine learning or ai, one example i recall was the one google ai model which would generate art with weird psychedelic effects. It was not the first image as such to be generated but that's one of the earliest example i recall.

Back then with barely an introduction to linear regression i had hoped to be part of some ai development aspect. Anyway a decent amount of time has passed since then, my career choices etc. have changed too. And now i am barely literate in anything AI.

With a respectful pause to probably the most important paper of this century, Attention is all you need - arxiv / wiki for same and the goated team at Google who built upon the great works earlier GOATs let's continue.

cool diagram.png
View cool diagram.png
[image of a transformer as proposed in the paper mentioned earlier]

Here we are, i want to learn, get into the maths and begin somehow but obviously didn't know where to start. I could dust out some books from my college days which mentions lots of algorithms used nowadays a bit of maths. I will have to brush up maths at khan academy and probably go through the course of mister Andrew NG. Before i do all of that, thankfully smart people have made things more easier for plebs my like me.

So one day, openai, one of the biggest AI labs in entire world released the open-model, not open source there's difference, which i could run. Obviously i was clueless but also bored and just searched a bit about how to do so, that's how i learned about ollama another great tool. I could download the models, run them locally without much headache.

I downloaded open ai 20bn parameters model, unsure whether it would work or not but it worked on my machine. A bit slow and that's how i got started with the local models. Entire day i spent installing more models from chinese dudes at deepseek and alibaba to google models. Testing things, posting about them on bharatchan to see what anons think.

Eventually some anon mentioned civitai, that's how i got into the local diffusion models. Installed tool like comfyui and models based off stable diffusion 1.5 from https://civitai.com/ and got started.

I understand that there's huge debate on this, morality of using the work of artists, and the way ai labs have scrapped data they are probably not the most ethical people out there. At one side it's really cool that one can have a glimpse at entire humanity's collective memory in a way on the other hand the ethical dilemma.

For me currently it's a learning journey, i want to explore, i want to learn the maths behind it and implement somethings. Despite my past experience with wrappers i need to understand it deeper. In a way feel it.

Somehow indirectly those local models have rekindled my interest in ai. I am glad for great researcher and scientist, to humanity and every one of the individual who have left their mark in the consciousness of these various AIs.

Hope to learn better and understand better.

JPEGView_t3nrkslA7e.png
View JPEGView_t3nrkslA7e.png

Image Refs in the Post