Our stance on generative AI

As an association working for the digital emancipation of everyone, in particular through free and open source software, we take a position on so-called “generative” artificial intelligence, which includes, among other things, language and code generation models, image, video and sound generation models, or the various assistants or “agents” that depend on them.

The exponential growth of proprietary digital tools and services controlled by large multi-million dollar companies is amplified by the arrival of AI, and represents a toxic alienation of users.

General considerations

From mine to energy, AI is an energy and material abyss and as such represents a dramatic regression in the fight to limit the serious consequences of human-induced climate change.
AI is a form of neocolonialism and exploitation of humans, from minors risking their lives, especially in war zones, to the hundreds of millions of click workers exploited in training and essential to the alignment of these AIs, in a form of modern slavery. AI also amplifies social inequalities; the educated and socially advantaged are always the ones who benefit more from it. AI is also involved and financed by the armies that use it to fight their wars. AI exploits and kills.
AI is a noxious appropriation of the common goods that art, culture and knowledge represent. It infringes copyright by siphoning off content without permission. Trained indiscriminately, AIs reproduce systemic racism, sexism, violence… and maintain, in fact, societal issues, while threatening the work and jobs of the very people who create the content they ingest and who could be able to control its biases (coders, journalists, authors, graphic designers, encyclopedists, teachers, etc.).
AI represents a significant economic risk, financed with billions of dollars without any profitability. It confiscates resources to the detriment of more urgent problems (global warming, social equity, etc.), without keeping its promises (productivity, decision-making support, reliability, access to knowledge, etc.). Promises that cannot be kept for a long time and without drastically increasing all the problems already listed.

With these important points in mind, our association would like to support its point of view specifically on the use of generative AI in the field of computer code and in relation to the emancipatory philosophy of free and open source software.

AI, open source and free licenses

Since the birth of the internet, developers have made available the source code of their tools, a large part of which has then protected the freedom of users, thanks to the free licenses on which our association is based. These licenses require that any reuse of this open code be also distributed under a free license and that the code be kept open, guaranteeing the future preservation of everyone’s freedoms. Generative AIs are largely trained on open-source licensed code, in direct violation of the principles of these licenses, while they are alienating. We oppose this dishonest use, which is in total contradiction with the emancipatory dimension of open code that we defend. Open, qualitative and public code is not intended to be digested and averaged by automatic systems that will often render it mediocre and unsuitable, widely used in private, private and proprietary code.
By making code accessible to all, generative AIs could have an emancipatory dimension, which could satisfy an association like ours. But as things stand, generative AIs largely depend on large private and for-profit companies, which are the only ones capable of bearing the staggering costs (economic and material), or ignoring them (when they are environmental or social). These companies can be software and service publishers as well as suppliers of servers and computer hardware. In this context, the use of generative AIs to develop one’s own tools amounts to swapping application publishers for AI providers, whose service can only be enshittified in their search for profits.
The use of generative AIs in code suffers from the same biases as in the field of text or media: the code generated by AIs is of poor quality and requires expert know-how to be controlled. Left unchecked, the risks range from unforeseen bugs to the most serious security breaches. These code errors, the consequences of which can be both trivial or dramatic (permanent loss of data, leaks of private data, failure of sometimes critical systems, etc.) are notoriously difficult to detect and fix, because the errors generated by AIs are very different from those of humans, who are not used to spotting them; they have a random dimension. The use of AI, which is not capable of reworking and correcting only certain parts of the code, goes against the usual iterative methods that improve and ensure the traceability and quality of the code.
Sharing the code generated in this way represents a generally unacceptable risk, without extensive verification work, which cannot be done by the less experienced people who take advantage of generative AI. It is largely for this reason that our association refuses to distribute tools whose code is not perfectly understood and mastered by their creators.

RxLaboratorio is therefore committed to remaining vigilant on the subject of artificial intelligence, and will always continue to place the emancipation and freedom of coders and users above all else, and is firmly opposed to any form of exploitation of humans. The association works for popular education and the increase of general knowledge that generative AIs oppose. The association also places the fight against climate change and damage to biodiversity as a cardinal value, in which generative AIs certainly have no place.

Notes and other resources

General considerations

AI, open source and free licenses

Other general resources about AI

English

French

La fourmi dans la coquille, Le trou noir philosophique, Podcast, Articles et livre
Cédric Villain, IA=CACA, 12 points pour refuser l’IAg
ATECOPOL, face à l’IA générative, l’objection de conscience

Please contact us if you have any other interesting resource about AI which would fit in this list!