Hi folks,

I'm impressed with everyone's patience to continue in this conversation. Yet, I 
am also under the impression that continuing this for much longer wouldn't be 
productive. My impression is that there are some folks who do not wish to have 
a policy of this sort at all.

>From my reading of it, this policy does not ban LLM contributions, and it 
>doesn't put LLM users at a disadvantage. It merely makes explicit our 
>expectations of any contribution. These expectations are particularly relevant 
>to LLMs because it is easy to enter a passive state where you do not read the 
>code produced and generate large amounts of low quality code. I think we are 
>all against this.

I agree with others that as this thread has gone on, the policy has been 
watered down somewhat. I think the key thing is the recommendation (R1). I like 
how it clearly follows as a consequence of the policies.

> Aside from that, I share the post-apocalyptic concerns about LLMs (climate, 
> alignment), however it is completely baffling to me how banning LLMs from GHC 
> development would help with that

Let me draw this out. LLM companies require a vast amount of money to keep 
producing new models. These companies fund this through a mixture of revenue 
and investments/debt. In order to access investments, they make promises about 
future profitability. Right now, many if not all of these companies are not 
profitable. The longer that they are not profitable, the more difficult it is 
for them to access investment, and so to make better models, etc.

As such, starving these companies of revenue is an effective way of avoiding 
apocalyptic fears about LLMs. The impact of any individual is small, but if we 
believe that the harm are as large as you mention, then even that is effective. 
Collective action is better than individual action. If the entire GHC project 
didn't contribute to LLM companies that would be better (under this premise) 
over an individual contributor not doing so.

Often collective action can make folks feel powerless since there's a lot of 
coordination necessary. If one person goes out to protest an oppressive regime, 
they will surely be arrested. If 1 million go out to protest, then they will 
topple the regime. Yet, it isn't really like this for using LLMs for 
contributing to projects like GHC. Not using LLMs likely wouldn't lead to any 
negative repercussions.

It's also worth emphasizing that this needn't be all or nothing. You can take a 
harm reduction strategy and try to minimise your contributions to LLM 
companies' incomes rather than eliminating them altogether. If the harms are 
great as you say, then even that is worth doing, if more action isn't possible.

There's also non-financial effects to a project not using LLMs. LLM companies 
more so are selling not an economic reality but an economic narrative. There is 
hype on the one hand and doom about attempts to resist on the other. Narratives 
are very fragile things. I think Zig's anti LLM policy was very damaging to 
this hype/doom narrative. Other open-source projects rejecting LLMs would help 
re-assert that these things aren't necessary to do development and that there 
are step that can be taken to avoid using them. I think that's a powerful 
thing. Although, I don't think that on it's own should dictate the policy of 
the GHC project.

> establish more effective political channels

I'm not sure what these channels would be. I'm not aware of any major political 
parties that are proposing a ban of LLMs, and I would especially doubt that the 
American political system would lead to LLM companies being disbanded. If we 
want to change the policies of polticians, then I think there's two ways (1) 
vast amounts of money, or (2) a large bottom-up organized campaign. (1) is 
likely not possible for most of us. (2) seems like the better option and then 
you are stuck doing bottom-up organizing again.

In any case, you are not restricted to engaging in one strategy. You can both 
engage in political organizing and not give money to LLM companies.

In general, I find this narrative that "LLMs are here to stay" or that the 
future is already fixed to be very telling. If that was clearly the case, then 
there would be no need to say it all the time. In fact, what these narratives 
reveal is that the future is not set. LLM companies are in an extremely 
vulnerable position. They require vast amounts of money and the transformation 
of work as we know it to merely keep existing. Just doing nothing is all that's 
required to not make that come about.

If you'd like to discuss these things more with me, I'm happy to chat in 
person, but I'm unlikely to reply further on this mailling list.

Cheers,
Teo
_______________________________________________
ghc-devs mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to