Hello everyone! Does anyone know what the /e/ developers’ policy is on AI? Is it officially used to speed up development, is it banned, or do developers do what they feel is best?
From a curious user.
Thank you for all the work on /e/os!
1 Like
Anyone? My guess from the lack of response is that the policy is “no policy, developers do what they feel is best”. Which I suppose is a reasonable policy at this stage.
Gitlab now propose an LLM software.
Does anyone know what the /e/ developers’ policy is on AI? Is it officially used to speed up development, is it banned, or do developers do what they feel is best?
To directly answer your question, I don’t know what the officially stipulated policy is.
However, whatever the answer is, I submit that it’s a bit difficult to define in a consistent, agreed-upon way. We’d probably agree that “adding entire libraries generated by a description prompt into a commercial LLM and then adding it into a build tree with no human assessment” is the case you’re probably asking for…but let’s go down the line…
- If a prompt generated library is added after a second set of QA by a different LLM, is that a problem?
- If a prompt generated library is added after human QA, is that a problem?
- If a library with 51% AI-generated code is added, with 49% human generated code?
- If a library with 49% AI-generated code is added with 51% human generated code?
- If a library does not involve an LLM prompt by the coder, but came from a human-reviewed StackOverflow prompt, where the StackOverflow content was LLM generated?
- If a library was LLM generated, but was sourced from LineageOS?
- If a library was LLM generated, but LineageOS pulled it from the AOSP tree?
- If a library was 100% human coded, but 100% LLM debugged?
- If a library was 100% human coded and no LLM wrote a line of code directly, but was used as a debugging tool and a human copy/pasted fixes recommended by the LLM?
- If the developer used tab-complete?
Personally, I’m in the middle of writing a book, and my stance is “I’ll bounce ideas of an LLM, and I’ll let an LLM help revise and tell me if something I wrote is inaccurate, but then I’ll research the inaccuracy separately…AI will be used as an assistant, but not one word of the book will be copy/pasted from a prompt response, and no diagrams or artwork will be generated from an LLM”. Now, some will still believe that this stance still involves the use of AI beyond their scope of acceptance…but if I had to go to the library and read all the books to find everything, it’d take me years to get the book out - I want to be done by the end of this year.
So…it’s a good question to ask…but I think that “they don’t use AI” is overly reductive, and depends on what actually counts…and I’m not sure there’s a consistent definition.
1 Like
I’d love to know the answer to this too. But I worry that after the response the developers got when they “teamed up” with OpenAI to introduce accessibility features (Speech to text) they are probably silent on this question to avoid another outpouring from our community.
A privacy-focussed OS integrating with the products of the notoriously privacy unfriendly (or even privacy-hostile) AI industry is something I suspect many of us here worry about.
I trust them enough to have bought the e/OS edition Fairphone 6, and I hope I can continue to use it AI-free for years to come.
But if they turn towards this increasingly unpopular AI technology, I’ll have to move on again. Maybe dumbphones will be the only option in the future!
1 Like