Gen “AI” disclosure

I feel it is necessary for me to define how I use the so-called Generative Artificial Intelligence so others know what to expect from my work.

This page is subject to change as my use of Gen AI changes.

Note

What some refer to as "AI" is not Artificial Intelligence. Machine learning models such as LLMs or Stable Diffusion are blobs of data generated from information scraped from the internet. They may be able to generate human-like output, but they are incapable of reason, and therefore mustn't be treated as "intelligent" in any way.

What the Fuck is Gen AI

Let’s first start by defining what I consider to be “Generative AI”.

Generative AI is:

Large Language Models
Image and video generation models with realistic output or stylized output resembling art
Audio-generation models capable of generating complex sounds such as voice or music

Generative AI isn’t:

Speech-to-text models
Translation tools
Image edition tools incapable of generating entire images (eg. background removal, object detection, object removal, upscaling)

This should give you a pretty good idea of what I treat as Generative AI, now let’s move on to how I use it.

How I Use Gen AI

TLDR: My use of Gen AI is mostly limited to using LLMs as coding and writing help.

How I Run My Models

My hardware is incapable of running any useful models at a reasonable speed, so I primarily use free online tools offering LLM chats such as duck.ai. I avoid services retaining user data off-device, and I refuse to financially support AI companies because of their unethical practices.

I always use LLMs in my browser or a dedicated app for running them locally, currently Alpaca. I never integrate them in my IDEs.

Coding Help

I occasionally use LLMs for explaining errors, writing large, repetitive chunks of code, or creating large data structures from data formats such as JSON.

Writing Assistance

I sometimes struggle to express myself even in my native language (Polish), not to mention English. I use a spellchecker to catch spelling errors, but that’s obviously not enough to detect syntax errors, logic errors and repetitions.

I might run some of my texts through an LLM for it to try and find those mistakes, then I correct them by myself or apply Clanker’s suggestions.

Occasionally I also ask the LLM to rewrite a sentence or two for me if I get a minor writer’s block, but that happens less and less often as I get more comfortable with English.

Translation

LLMs have greater context awareness than traditional translation tools, so I might use them for translating large texts from language I don’t know to a language I’m proficient in.