AI and I - me and AI

Inspired by Damola Morenikeji's /ai manifesto I have put forward a slashpage to describe my relationship to AI.

Slashpages are a common practice of creating transparently named pages that relate to a certain topic. This is my AI page.

I don't write using AI.

Everything you can find here, except for some quotes when explicitly stated, has been written by a human. Everything went from my life, through my brain, to this site. Maybe that has its shortcomings, and sometimes I write in a less diverse, poorer, and messy blob of prose. But that is my decision, this is my site, and this is me, with my shortcomings. We should let the people behind the web be human.

I have used AI.

I've tried the GPT model on a few occasions, notably to solve very technical problems when I couldn't patch together the available documentation. Sometimes this has helped, for more basic tasks that are clearly outlined on different forums that I couldn't otherwise easily find.

Most of the time, problem-solving technical issues with AI gave a messy answer lacking in precision and not at all adapted to my specific case, even when prompted clearly. I blame LLM's lack of capability for formal reasoning. The searches have mostly led me to sites, where humans have helped me way better.

I have never used AI to "produce an artwork" outside of the scope of contextual filling of backgrounds in images with Stable Diffusion.

No, I do not trust AI.

As shared above, LLMs don't do formal reasoning and that makes them incapable of the abstract thoughts that humans have, and the abstract formulations in which we tend to get it out. For this reason, answers to AI prompts should never be taken at face value, but instead always be questioned and verified.

I don't trust AI, because it tends to be wrong way too often. Because it's not human, it doesn't really perceive its mistakes. Would you trust someone that seems way too sure of themselves? I wouldn't.

LLMs "create" crappy things

AI art sucks. This isn't a matter of how well AI can replicate "realism", it's about the type of pictures it generates. AI is not intelligent and not artificial, it's a probability model. It will create the thing that is most probable to exist according to certain circumstances. Apart from merely interesting artifacts when it fails realism, there is nothing of value coming out of it, because it's nothing more than a general average of what exists already. It's a Basquiat clone, it's a Yayoi Kusama wannabe, and it likes the racial and gender stereotypes that the creators put into it way too much.

Much the same applies for the text it generates. Word after word, one sentence after another, all comes out because it's the most probable thing to appear in context to what's already there. Do yourself a favor and turn in a blank paper instead of the digested and polished say-nothing sludge an LLM can do for you.

If you can't be bothered to write it, I won't be bothered to read it.

A huge waste of resources

AI powered scheduling. AI powered image editing. AI powered search and summary. AI powered coffee machine. AI powered reading. AI powered web crafting.

So many dumb uses, and we still haven't found a way to make AI-powered domestic tasks and shitty chores. Considering how much energy is wasted by AI, shouldn't we be using it to give us more time for fun stuff? Fix my god-damned search engine first, Big Tech. Make the robot do chores.

AI-powered bullshit.

All at the cost of a truckload of coal for a simple search query.

Stop making AI do your art, writing, music, web search, or problem-solving. Because it can't do it well, and it won't be bothered to save energy.

If you like what you just read,