LLM/AI contributions seemingly being accepted now?

[Meister1593]

I see that in https://github.com/tutao/tutanota#contribution-policy it's explicitly prohibited to use LLM/AI in any shape or form for contribution.

But i see:

• this commit being added with assitance of claude (and it's a HUGE commit)

- this issue is pretty much clearly generated and admitted - being okay?
Ok i admit that was month before adding a policy, my bad

What is going on?

I thought it was clear from this discussion already.

[hrb-hub]

Hi @Meister1593 👋

Perhaps I can help clarify things.

While GenAI-assisted external contributions are prohibited, LLMs are sometimes used internally during development. Our reasoning here is that there's a big difference between accepting LLM-assisted commits from internal teams versus accepting LLM-assisted PRs from the public.

Public code contributions worked in the past because the barrier to entry was high enough that any non-trivial PR would have been so much effort that its existence alone could act as a rough proxy for quality, this assumption does not hold any more.

Internal use is different because whoever is committing LLM-assisted code takes full responsibility for it and is therefore staking their reputation on that code being good. Git blame will also still show their name.

Hope this helps :)

[mpfau]

Thanks for your elaborate feedback, but you are comparing apples with pears – as we say in Germany.

At Tuta, we are very energy efficient, and this because we ourselves want to do our part to make things better. For instance, our infrastructure consumes less than 1g CO2 per active user per year. That includes all hardware manufacturing and used energy. We implemented a lot of small and bigger measures to get to this state and run our whole infrastructure exclusively on renewable energy since 2019.

That is why I strongly disagree with your framing that we don't do anything at all. In fact, I am very proud of what we achieved so far and always happy to have constructive discussions about further improvements.

While I agree with you on the fact that resources consumption from AI usage must be further improved, I strongly disagree with you on how to achieve this. The fact that a technology has environmental costs does not mean that the only responsible response is to reject the technology altogether. I think the more useful question is where its use provides sufficient benefit to justify those costs, and how we reduce those costs. This is, for instance, exactly what Fairphone is doing.

We are currently using AI on a very low level but already started exploring using renewable energy for it and will continue to do so. While running inference on renewable electricity does not fully solve the issue, it substantially reduces emissions.

I also disagree with your conclusions about branding individuals as unethical by reducing the entire evaluation to their usage of a certain model while ignoring everything else.

[suleyth]

where its use provides sufficient benefit to justify those costs

There is no benefit worth selling our humanity and our planet to murderers. I am very sad to hear that you think coding Tuta a little faster is.

[nepmark]

Wow, this is actually incredibly disgusting. It's not even environmental issues only, but many others like worsening many real engineers' life. This is why I quit programming, it is babysitting a robot and not logical reasoning anymore... If you're using AI, then why not going with Proton? Not any less ethical at the point. Better support, better users, better UI...

[alexoliva91]

I agree on the principle that AI is mostly in the hands of few that have a future in mind which is worse for many (most) of us. I also agree that the planet's resources are scarce and the whole AI industry is worsening the long standing exploiting of poor countries by the tech industry.
Having said this, it is clear that most of the people don't care and AI will be used more and more until this bubble finally explodes. Exploits are found daily and serious security breaches are becoming the norm thanks to AI and even if we all stopped using them, the respective companies would still push their models in this daily cyber-warfare that is clearly sponsored by certain governments.

So I agree that using AI for developing a new theme in a couple of minutes is sickening (yes, I am talking about Omarchy) but I honestly believe that at this point you need AI to keep up with all the security issues. The only point I have here is that Tuta should use an open model and not Claude or similar because, other than the environmental issues mentioned above, it is like giving away your expertise and also paying for that.

[taelspinner]

An utterly myopic and tone-deaf response. Nothing says "we aren't changing our minds for anything, leave if you don't like it" like a dismissive and condescending "thanks for your elaborate feedback, I strongly disagree, we are doing fine" response to detailed and thorough customer feedback from knowledgeable professionals voicing legitimate concerns.

I already left Tuta for unrelated reasons, but I feel bad for those who haven't, and will now be taking at least one paying customer with me. Good luck to you in pursuing your mission while betraying your stated principles and blowing off your customers. Time will bear out how well that works for you.