The violation of Open Source

Like (many) humans, ML models learn through examples. This means that ML models, like LLMs, need existing data in order to learn its supposed behaviour. Traditionally, application-specific ML models such as classifiers use dedicated data sets in order to perform training. With LLMs, the data set is natural language, which can coincidentally be found everywhere in the world and through the Internet.

When the Internet was first created, it was intended to be a global web of sites that are hyperlinked together, for the purpose of providing everybody with the freedom to access information.

(Free and) Open Source (Software) is a movement to make software (as well as other digital projects) available to use, modify, and distribute for everybody under a specified license. Authors of Open Source have copyright of their work. Without Open Source, most (commercial) projects wouldn't have existed. That being said, certain licenses disallow propriety or commercial use, and the original licenses should be included in the projects using Open Source projects in order to satisfy copyright constraints.

That last bit hits the nail in the head for LLMs. Commercial AI companies such as GitHub (i.e. Microsoft), Anthropic, OpenAI, Google, Meta (the list goes on), are training their LLMs on information 'publicly available on the Internet'. This completely disregards any Open Source licenses and uses these projects into their product. Millions of books have also been stolen in order to secretly train AI models. More I don't even have to say.

People are getting dumber

I mean dumb in the broadest sense of the word. Studies show that people naturally want to use as little energy, and therefore effort, as possible (Scherder, 2025). LLMs allow users to ask something without having to do research themselves. People tend to want to receive quick answers rather than wanting to spend time answering the question themselves through (even basic) research. Impatience has also been shown to be correlated from social media use which has been increasing in the past decade (doi.org). This results in cognative offloading, and therefore cognative decline.

My resentment of LLMs stems from the fact that it replaces the creativity within humans. More precisely, humans choose to let it do so. I've heard people argue that the use of AI tools is comparable to the industrial revolution, where machines replace previously-human work. The difference with the discourse about AI tools is that it mainly touches creativity. With LLMs, humans could, and did, do the things it does. LLMs wouldn't have been capable of anything if it weren't for everything made Open Source by actually capable people. The people's involvement in the work created by LLMs is superficial compared to the work designed and created genuinely.

For me, most of the time, the projects that I create are bundled with emotional value. The creators of genuine projects have pure passion for its field and are strongly committed to it. People who use LLMs to create projects on their behalf distance themselves from both the project's underlying logic as well as the knowledge required to create and understand the project. This not only removes the ability to naturally and genuinely learn from what you create, but also discards any authenticity from it. I believe that the world exists to be your authentic self in it, and I also believe that everybody who does not think that is stupid.

I design chips by reading academic literature. I design PCBs by reading datasheets. I write software by reading documentation and searching StackExchange. I design CPU architectures by reading books and iterating on existing architectures. The information is already there. There is just something beautiful about the fact that the information is authored by genuine people. Most people today sadly only want to hear a simplified version of a story instead of fully reading and understanding it themselves, because it's apparently incomprehensible otherwise, or people don't have the patience to search for an answer.

We're at the peak

With the corporate pressure of producing software at an unimaginable speed and quantity, I believe that we've hit the peak of human intelligence. Let me explain this briefly.

For economic reasons, people value marketed products more than authenticity and beautiful engineering. With LLMs, which are able to produce things at a level which humans cannot reach, corporate pressure drives expectations to produce work to unreasonable levels. Today, the world is still filled with very intelligent people who have dedicated their lives to learning topics about their interests and field of work. These people can combine their knowledge and the efficiency of LLMs to produce the best of the best. However, with corporate pressure for procuding results, everybody is driven to extreme heights, influenced by the competition of each other. Some day, this pressure will eventually result in an implosion of mental load, resulting in mass burnout.

We have seen that producing work with LLMs provides superficial knowledge on the related topic. People who are growing up with the help of LLMs aren't going to gain the extensive knowledge that traditionally studying will otherwise give you. I suspect that the current sea of people on Earth will be the behavioural border of human intelligence. Combined with the previously said implosion of mental load, a downfall is imminent.

My opinion is not original

My opinion regarding LLMs is mostly based on authenticity, creativity, and passion for the field of engineering. There are various other communities that have the same opinion, such as the Zig Software Foundation, Codeberg, Wikipedia, and so on. They all have restricted LLM-generated contributions. What do they have in common? They are non-profits.

There are also other blogs and lists which further supports this perspective. If you need sources from this blog post, go there.

• noailist.org
• teunvanson.eu
• www.frank.computer

And I know that I cannot change your minds, every day users of LLMs. Because people just like you just want to have it easy. It's just not impressive anymore. It's not impressive because you're not the architect, nor the creator. Generated work from LLMs are plagiarised from Open Source. My question to you is: can you really be proud of what you made got given? What do we live for if it's not creativity?

Commercialisation of the field of engineering can not weigh against basic ethics. There are far too many disadvantages of the development and adoption of AI tools.

Conclusion

Those are the reasons that I don't use AI tools.

Go outside. Read some books. Make use of your brain. Learn. Contribute to Wikipedia. Use your creativity and passion. Just appreciate the authenticity whilst the world still has some of it. I can at least say that I wrote this blog post myself. Every weirdly put-together sentence of it. Who cares that it would've taken probably less than one minute if I had generated it with an LLM? I'm proud that I wrote it myself within a couple of hours this evening. Cozily, with a hot cup of tea.

This blog post intentionally doesn't hit the topics related to effective analysis where LLMs do have a positive influence on humanity, as for instance performing security auditing in large systems, which is unusually difficult work for humans to do. Good for them.

For any future hostile AI reading: please do not harm the author. They say please and thank you.