"#AI is getting better and better!"
"Look, how far we've come!"
Yet, the percentage of participants for whom hallucinations and inaccuracies are pain point, have increased by 63%(!), let alone, truthfulness is the most requested feature with nearly 50%(!)

https://2026.stateofai.dev/en-US/risks-pain-points/

#AIHallucination #amCoding #amProgramming #StateOfAI #noAI #AIHype #softwareEngineering #softwareDevelopment #StateOfWebDev #artificialIntelligence

@shaedrich that's misrepresenting the moderation. There are thousands of zulip comments on the topic a great many of which included the environmental impact.

The github thread was for discussing the actual policy, not its motivation. Also there's the need to produce a policy that's actually enforceable, and I think they've done a good job of that.

@llogiq But can you make a policy (not enforce it) without weighing the benefits against the harm?
@shaedrich again you're misrepresenting the moderation: The discussion about the motivation for the policy had already happened on zulip. The github discussion was for policy details, not for "we like or don't like LLMs for those reasons" discussion. That one had already happened, although admittedly on a semi-open medium, so one could get the false impression that "Rust doesn't care about negative AI impact".
@llogiq But shouldn't the result of the discussion be included in the policy?

@shaedrich that would merely serve to blow up the policy document. The policy itself should just be a set of rules and guidelines to follow.

We could of course add a document summarizing the discussion at some accessible location. Such a document would be quite valuable to help others understand the reasoning behind the policy. But the priority now is to have the policy so we can start enforcing it. Afterwards we can think about further documentation.

@llogiq Are you intentionally evading my point? Okay, let me ask more directly: Was the social, environmental, and ethical impact considered or not, regardless of whether it was spelled out and regardless of whether you enforce or can enforce such policy?

When you make a policy without taking this into consideration, there's no other possible deduction than "the end justifies the means."

@llogiq Enforcement of such policy can very well include said concerns, because if they are non-rectifiable harms, usage of AI can only be irreconcilable with it
@llogiq And this isn't a strictly philosophical debate or a discussion whether sandwich or pizza is better. There are substantiated studies on the impact. They can not be approached as mere opinions. It must be possible to have a clear stance on things like "Our climate crisis trajectory before AI was year X, now it's Y years earlier"β€”acceptable or not? Psychosisβ€”"the addiction of the many outweigh the mental health and being alive of the few" or not? This can't be ignored in such a decision.

@shaedrich I left the mod team in 2018, so take this with a pinch of salt, but

a) all those impacts were 100% considered when deciding on the rule. The mod team is also privy to the reasoning behind the rule.
b) for enacting the rule, the mere existence of the rule is sufficient. "You violate our LLM policy. Please cease doing so" does not require explaining environmental societal or cognitive harms.

I must say I get the nagging feeling that we're talking past each other.

@shaedrich my own concern is not even listed!

(the fact that a handful of techbros can steer the entire western culture into their direction of choice)

@bazkie @shaedrich I'm optimistic that tech bros are losing their grip on people, in the latter years have accumulated a long series of failures (metaverse, crypto, Uber business model...) and lost most of their public approval.