AI Transparency

I want to give some information of how and where I use or do not use technologies like generative artificial intelligence.

TLDR: For all my projects I aim for authenticity and individuality. In some rare cases I use assistants for inspiration and code completion, never for my writing. Because of ethical reasons, I try not to use generative artificial intelligence at all.

Writing

All of the blog posts on the "text." subdomain are entirely written by me. I do not use any "intelligent" tools like LLMs (e.g. ChatGPT), writing assistants (e.g. Grammarly), or translators (e.g. Deepl).

This increases the probability for grammar and spelling issues, especially since English is not my native language. However, I believe that this is the only way to keep an authentic writing style unlike the bland linguistic smoothies that assistants may come up with. Thinking of the future of writing, it's obvious how authenticity and individuality will be the most relevant aspect to make it distinct and worth reading.

I suggest a different method of perfecting my communication: If you, as an individual, find any major and relevant mistakes, please let me know and I will fix them accordingly. This gives a kind of "distributed intelligence" that, unlike LLMs, is distinctly directed by individual people.

Inspiration

With the goal of transparency, I admit to have used assistants like ChatGPT during relentless searches for inspiration and new ideas. I do believe that certain interactions had a influence on the motivation of some of my projects and articles.

The use of generative artificial intelligence as a tool of chaotic exploration without any specific goal or motivation is in my opinion one of their only actually useful applications (at least in their current form).
I believe that the best ideas grow from questions that nobody dared to ask or most people would laugh or be confused about. From this aspect, the "hallucinations" that are common in generative artificial intelligence can in rare cases actually be used as a seed for such ideas.

Code

I try to implement most things I work with or write about. Some of those implementations are large or repetitive. If I have a strong belief that it could save time, I use(d) GitHub Copilot to complete common code snippets.

This is especially the case for C and Python projects where some implementations are well-known and trivial to implement. In C, this could be a simple implementation of algorithms for a linked list. In Python, it could be some plotting boilerplate or numpy transformations.

I don't use Copilot anymore but won't promise anything.

Reasoning

While I already mentioned some reasons for not (or if, rarely) using artificial intelligence there are many more reasons why one should not use artificial intelligence carelessly.

Because of those points I actively try to further reduce my usage of AI.

---
Marvin, 2024