qua 01 jul 2026 às 15:27:13 (1782930433), [email protected] enviou: > Hi, > > André Batista <[email protected]> writes: > > [...] > > > To my understanding, this GCD asserts a general distrust of GenAI usage > > for the various reasons refered to in the motivation section. > > > > As such, people more aligned with the project, people that somehow are more > > officialy tied with guix and that can be seen as project > > representatives, are being asked to avoid its usage (on the project's > > context and on the designated use cases) and to commit to that. > > And I think this is where this GCD makes reaching consensus difficult: > it tries to have us agree on a shared world view that GenAI as a whole > should be seen negatively, for environmental, social and other reasons. > If you don't agree (or simply do not think this kind of stance belongs > in a free software project), then the exercise of reaching consensus on > the commitments or policies, knowing the reasons that motivated their > existence, becomes difficult. > > I'm also firmly of the opinion that introducing > expectations/policy/rules bound to divide our community or taint our > view of some of its members is not a good thing, especially since genAI > has been with us for a couple years already and we so far haven't seen > serious practical issues with it (such as being buried under a mountain > of genIA slop review requests). > > In my view, the "let's take a stance against genAI" appears more like a > PR exercise, than solving an actual problem we've been dealing with. >
For the initial draft, I think this criticism was valid. We had to have a shared negative view on GenAI and be on agreement that action was to be taken against it in order to reach consensus. Not so much for the current version. The way I see it, the current proposal has moved away from trying to reach a negative consensus and taking decisive action against GenAI and towards an agreement that the current situation warrants some caution on our part. IMO, one does not need to have made up their mind against GenAI nor to fully agree with any of the stated motives to _accept_ this proposal. What's required is a modicum of humility in face of uncertainty. The precautionary principle, which comes from envionmental law, is applicable precisely when there is no consensus or conclusive proof that supports a definitive action on a given subject matter. According to it, when faced with uncertainty regarding the possibility of a harmful outcome from some action we do not try our luck. Instead we take preventive measures to contain the risks of such negative outcomes until a more definitive answer is found. In a similar vein, the 'Cost of Reverting' section addresses another aspect of this decision which is the 'periculum in mora' (danger in delay). Right now we can act to safeguard our project from the potential risks that GenAI poses. If later on we find out that these measures are not actualy needed we will be able to drop them or make them more lax. However, if we take no action and just wait to see if those outcomes come about, later on it may well be impossible to revert the damage done. Since these are precautionary measures, maybe we could stipulate a term for these measures to be reassessed. Two years, maybe? Then instead of having to go the normal GCD route to revise or revert the conclusions of this GCD, we would have a shortened procedure where we go straight to deliberation period where everyone is asked if they support further extending the term or if they want to reopen the discussions/conclusions of this GCD. WDYT?
