I keep seeing more and more discussions on the forums of some open-source projects (which I won’t name ! I’ll leave that as an exercise for the reader) where people react very strongly (and that’s a euphemism !) to the fact that some code was written, entirely or partly with the help of AI.
And it leads me to a fairly fundamental question: why does the use of AI change our perception of the result so much ?
Let’s take 3 hypothetical situations that already existed long before generative AI became widely available:
- I know nothing about software development, but I have an idea for an open-source application. I contact some developers, explain precisely what I want, they build it and, after several discussions and adjustments, the application eventually works and maybe even becomes successful and widely used. Does the fact that I didn’t write the code myself take anything away from my idea or from the result ?
- I am not particularly good at writing administrative or official letters. I need to send an important formal letter, so I contact a lawyer, explain the situation and ask them to write the letter for me. Does the letter somehow lose its value because I didn’t write it myself ?
- I need to translate a letter into a language I do not speak. So I contact a translator, give them my text and ask them to translate it for me. Is the translation somehow less valid because I didn’t do it myself ?
The same situations, with a different tool#
Now, let’s transpose these situations into the current AI era.
- For my application, I can ask an AI to produce a first draft of an app. It will most probably not be perfect, but I can test it, modify it myself, ask for corrections and improve it through successive iterations.
- For my letter, I can provide the context, the facts and what I want to say, and then ask an AI to help me formulate it properly, in a given style.
- For the translation, I can also use an AI, give it the text, the context and the tone I want, and then review and correct the result afterwards.
In all 3 cases, the “tool” changes, but the principle seems, at least partly, quite similar to me: I rely on a skill that I do not necessarily have myself in order to turn something into a concrete result. Is that, by itself, enough to make the result less valuable ?
A human professional can explain their choices, identify ambiguities, protect confidential information, and be held accountable for their work. An AI can produce explanations, but it cannot reliably guarantee that they are correct or be held responsible for the result. Just like humans, AI makes mistakes. The person using it remains responsible for understanding, verifying, and maintaining what is produced.
Why does AI trigger different reactions?#
Why can a result sometimes be perceived more negatively when AI was involved than when a human professional provided assistance ? Would that perception change if the AI tool were open source and ran locally ? Those properties could affect privacy, transparency, and control, but would they also change how people evaluate the result ? Would the perception be different if the use of AI were not disclosed ? If so, would disclosure be necessary to avoid misleading people, or would it simply encourage assumptions and accusations ?
In short, I often have the impression that attention shifts from the observable qualities of the result towards the way it was produced. The “how” often become more important than the “what”. And this difference in perception is exactly what I am trying to understand.
I understand that both sides of the coin cannot always be separated. The way a result is produced can affect its quality, security, reproducibility, licensing, confidentiality and maintainability. The question is therefore not whether the production process matters, but how much weight it should have in a particular situation.
Is the reaction primarily about trust ? Responsibility ? Intellectual property ? Quality ? Transparency ? The speed and scale at which AI can produce things ? The absence of a single identifiable person who can be held accountable ? Or simply the fact that we assign a different grade to a result depending on whether it was produced by a human or with the help of a machine ?
Are these really the reasons behind the reactions I see, or do people sometimes use them to justify a feeling of unease they already have ?
There are also other important questions that I am deliberately setting aside here, including the effect of AI on people’s jobs, ethical, economical and ecological concerns, the non-deterministic nature of AI-generated results (even though it was announced 2 years ago), the sense of pride or loss of pride associated with making something oneself, the value of craft, the sheer scale of the systems involved, and questions of solidarity. These considerations matter to me, and I do not mean to minimise them. Automation and productivity gains have accompanied many technological changes over time. Some of those changes had negative social consequences or made certain jobs disappear, while others changed existing professions, created new skills, and led to jobs that did not previously exist. However, these broader discussions go beyond the scope of this post. I am setting them aside only to focus on the narrower question of why the use of AI can trigger such strong reactions, even when the result itself is also part of the discussion.
This is therefore a genuine open question for me. I keep coming back to it when I see how much time and energy can be spent debating the use of AI, sometimes without much attention being paid to the actual qualities or consequences of the result. In some discussions, asking whether AI was used seems to become a form of gatekeeping: before considering the result itself, people first try to determine whether it deserves to be evaluated at all.
I am not trying to blame or defend anyone or AI, or even minimise the problems it may create. I am also not claiming that asking a human for help and using an AI are equivalent.
Quite the opposite: the differences between these situations are precisely what I am trying to understand.
I am not asking where the fundamental difference objectively lies, as though the answer were obvious. Nor am I suggesting that these situations are strictly equivalent. I am wondering whether the reasons commonly given genuinely explain the strength of the reactions they produce.
Where and when does AI become problematic?#
This leads to a second question: regardless of how people perceive it, when does the use of AI actually become problematic ?
One possible dividing line is not simply whether AI was used, but how it was used, what is at stake, and whether the person presenting the result can understand, verify, maintain, and take responsibility for it.
Other factors may matter too: whether confidential information was shared, whether the use of AI was disclosed when disclosure was expected or whether the person using the tool has the expertise needed to identify potential issues.
Personal opinions, including mine, do not settle the question. What matters is whether the arguments hold up, and whether they apply consistently across different situations.
Please feel free to leave a comment below, I am genuinely curious to hear where others draw the line.
Quoting a sentence I saw recently:
You don’t have to be happy about a situation to make the most of it, and making the most of it is often the best act of political warfare.
