Random thoughts on « AI »

letter to my Guix friends



---

Note: The context of this personal note is GCD 008 Standing up for human crafting . My personal opinion does not appear to me relevant for the on-going discussion. The following is fed by a challenging discussion with a long-time free software activist, and after such discussion about wide ethical considerations aside our own usage or opinions, I want to record my own thoughts at this moment in time.



---

The initial question I have hard time to resolve is: how fellow hackers who are spending uncountable hours and who advocate since years about free software are they willing to adopt GenAI in their workflow for contributing to free software? Let be clear, I’m not considering some « fictive » GenAI but the one it’s really used now, namely Claude, Codex and all the variants by Big Tech companies.

( Please don’t read Big Tech as evil; Big Tech is what they are: big companies about technology. )

The question isn’t about making individual concessions with the real world, nor individual choices. For instance, I’m using non-free firmware, Gmail for my emails, and my current laptop is running Ubuntu; I recently took a long-flight for holidays, I own a van, etc.. There is no witch hunt, nor pointing at individuals. In short, the question isn’t about some individual « moral superiority » or the reverse – poor me.

For what it is worth, I think we – free software as a whole; maybe humanity? – are at a crossroad. The world is going in one direction and, from my point of view, the question reads: Can we show another direction? It’s not because it’d not happen that we should not try; who knows the impact on the future from the tiny effects of now. Moreover, it does not imply that a stance right now about GenAI is written in stone. Today, the crossroad commits us.

When the world is moving, I’m convinced we need to still hold the principles of free software, advocated since the very beginning; section “Why I Must Write GNU”.

To me, these principles are rooted in:

1. Being benevolent.
2. Being autonomous.

Being benevolent is a worldview that implies we must care each other as human and life. It’s probably a variant of some golden rule or some ethics of reciprocity. Being autonomous is a worldview that reverses the power dynamics and encodes how the freedom articulates with the tools and technology. The four essential freedoms¹ appear to me an implementation of these two principles.

All the challenge is to draw the required actions considering all the constraints from the real world. Any collective project isn’t living in the abstract. Thus it operates using ethics: what the project ought to do or which behavior is right or wrong. And that’s where we might differentiate “strategies”.

Assume the project agrees the color of the bike-shed will be grey but then which grey? Ethics is a complex spectrum that I know nothing² about. Let oversimplify: At one end, there is consequetialism – the consequences of the conduct are the ultimate basis for judgement about the rightness or wrongness of that conduct – and at the other end, there is virtue ethics – the primary subject for judgment about the rightness or wrongness of the conduct is the virtue of that conduct itself. Somewhere in the middle, there is deontology: the duties are the central key for judging under principles but constrained.

Let take a distorted example: “worse is better” is about consequentialism – it make concessions with the worse while the consequences are better – and “the right thing” is about virtue ethics – it’s done only whether it’s right (virtue). Yes, it’s a distorted example! Another even more very distorted example: a biased lens about the debate “open source” v. “free software”; consequetialism v. virtue ethics.

This imperfect interpretative framework helps (me) to understand why some accept to trade with GenAI even from Big Tech because they are consequentialist – the consequences of using GenAI helps to reach the goal (a better “free software” product) – and why some strongly oppose to all GenAI – there is no virtue when using GenAI.

In my humble opinion, deontology helps because it focuses on the duty under principles. The question to ask is thus: Are we benevolent and autonomous when using GenAI?

Obviously, there is no answer. Still, it appears to me much more fruitful for thinking why do free software activists feel the need to resort to GenAI? When using GenAI, how does it help in being more benevolent? When using GenAI, how does it help in being more autonomous? When using GenAI now, how does it help the future generation to be benevolent and autonomous? Again, there is no answer; only an individual exercise of thought.

For the ones who tend to answer “no it never helps“, be aware of dogmatism – the tendency to lay down conclusion as undeniably true, without consideration of evidence or the opinions of others. Let tell me a short story that illustrates, I hope. Do you know that I find Amish community inspirational about their approach of technology. For one short paper: Amish Technology: Reinforcing values and building community by Jameson M. Wetmore. The title already provides an idea, I guess. Or this other short article: The Amish stance toward any invention isn’t that they reject it outright. It’s that they start by assuming they don’t need it, then adopt it only if it’s in line with their values. Now, give a look to this article: with all respect for Amish people, it seems dogmatism; although I’m sure the situation is more complicated than my poor understanding. Where do I go? Deontology considers the principles in the light of the specific situations of the changing world.

Deontology focuses on the duties as a never-ended self-reflective quest about holding the principles; duties constantly adapted to the current world.

Where I’m still struggling is about using GenAI produced by monopolistic Big Tech companies. When using GenAI respecting Open Source Initiative definition, it makes sense to me because someone chooses for the bike-shed a grey that is shaded³ by “worse is better”. However, from my point of view, GenAI from Big Tech defeats the both principles “being benevolent” and “being autonomous”, then I’m not able to reconcile. Still thinking… And, if people make concessions with GenAI produced by BigTech, then they should also accept to make concessions with non-free. Because the consequences of using non-free will be “good” (more users, more contributors, etc.). Inconsistency? Still thinking…

Now, my opinion, for what it is worth: I think that, under the current situation of GenAI, free software should pledge⁴ to be strict about the contributions using GenAI. Not because some virtue ethics but because deontology. The current situation of GenAI does not help in being more autonomous and it’s debatable about being more benevolent since the consequences on the ethics of reciprocity remains very unclear and the loss of autonomy is clear.

Last, I leave it here: the collective practise, is it the set of individual practises?

Anyway. Who knows how the situation about GenAI will evolve in the coming 2-5 years and then it’s up to us to deontologically evaluate again the current stance we’d make now.

Random thoughts. An opinion. No definitive answer!