These are actually nice guidelines and they can serve as a clarification for the difference between reponsible AI use and “AI slop”. Would you make a comment over at the GitHub PR?
I’m a bit torn about this.
On the one hand, this can enable people who do not feel comfortable writing about technical topics in English.
On the other hand, such prompts can lead to text that is unnecessarily long and verbose, which puts additional burden on the reader. In the corporate world this already leads to absurd situations where people are using AI to generate e-mails that are so wordy that the other side has to use AI to make a summary Also, the resulting language can be very sterile and corporate-like, to the extent that the reader feels that they are talking to a machine (well, they actually are to some degree). This might be ok for technical things like GitHub issues, but not so much for the forum which is supposed to be a place for human connection. It’s just something to be aware of.
Personally, I would not prohibit the use of AI prompting (if combined with proper review and editing!), but I would not encourage it either, following the general idea of valuing human imperfection over machine-generated output. Instead, I would suggest to first write by hand and then use LLMs for translation/editing, if necessary.
Anyway, you’ve brought up an important point and I think the case of (responsible) AI prompting should be covered in the PR!
I personally have a strong stance against LLMs b/c it is violating open source licenses in a big scheme - cooperations are stealing code that was offered for free (as in freedom), but they do not give anything back to the community. Calling this a sleazy and lazy excuse to deny reality is therefore an interesting take to me, but lets keep this on topic and discuss such moral aspects in a dedicated thread.
The proposed change does not want to ban AI contributions - its intention is to ban obvious AI slop and also make it necessary to be transparent if any AI tools were used and also support the user to turn off AI tools for contributing.
IMO the proposed text is really AI friendly - most FOSS projects I come across have a simple full ban on anything AI related.
This is only my personal take, but I noticed a shift in your writing style. I personally would encourage you to write your posts in your native language and then only use a translator tool to help you bring it into english instead of relying on LLMs to outline your posts And I would also encourage you to write in english and use a translator tool only occasional or an old school dictionary once you become more familiar w/ the language.
But I totally appreciate the transparency regarding the used tools and your contributions lead me to believe that AI tools should not be banned in general since they can also be used to help users and the community!
RegEx seems quite taunting in the beginning, but tools such as regex101.com help me to come up, learn and test regex.
While LLMs can come up with valid solutions, they are often verbose and obscure, which makes it harder for the reviewing part.
I think this is good addition! Maybe add a prefix of “If you decide to use AI assisted tools”?
Though personally I’d appreciate if people would only use translators and no LLMs, but this is my personal opinion and once people are transparent about it I am fine b/c then I can decide if I want to put effort into it or also “delegate it to a LLM”^^
This month, I realised that my behaviour and my mental activity have somewhat changed since I began relying on LLMs.
Even when I expect a task could be completed more quickly without AI, I sometimes allow the LLM to attempt it first. I think this is due to laziness. The recurring difficulty with doing so is that I often end up guiding it towards producing the correct result. This happens very frequently when using Gemini.
LLMs often process my requests with excessive verbosity, both in language and in the information provided. While this can occasionally be useful feedback, I quickly become tired and lose concentration because unnecessary time and energy are consumed.
LLMs frequently supply false information presented in an engaging format, and I sometimes fail to notice the inaccuracies. At times, I even feel that my own ability to comprehend is diminished.
The most serious issue is not simply that LLMs sometimes produce seemingly successful outcomes, but that I may mistakenly assume such results are always reliable. This illusion encourages me to turn to AI automatically, even though it is still far from perfect. I should therefore use it with caution.
From these experiences, I would say that the traditional way of working without LLMs is safer. Using AI can occasionally save time and sometimes produce better results than working alone, but it remains unreliable—rather like working with inadequately trained experts.
–
Thank you for your advice!
Yes, I agree. Regarding writing in English, I feel that my own skills are no longer improving since I began relying on LLMs. It is somewhat discouraging because, although the text may appear better, my personal ability is deteriorating. To enhance my skills, I should write independently first and then use the LLM as a reference.
–
Thank you for this! I will try it again when I need regular‑expression support.
I asked an AI expert how I could train a model by scanning the PDFs and books I have. He explained that undertaking such a project as an individual would be far too expensive, so I decided to abandon the idea.
I used the following PDFs to find the exact phrases I needed with a chatbot in POE:
Ernst Heinrich Weber. ⟮1834⟯. De Pulsu, Resorptione, Auditu et Tactu: Annotationes Anatomicae et Physiologicae. Leipzig: C.F. Koehler.
Jean Baptiste Joseph Fourier. Théorie analytique de la chaleur. Paris: Firmin Didot, 1822.
However, I have no experience with the chatbots provided directly through the paid versions of ChatGPT or Claude. AI services evolve too quickly. For me, POE is convenient because it allows access to various services, but it is not the same as using each service directly. For example, using ChatGPT via POE does not provide the same experience as using it on the official ChatGPT homepage. I think the service from the ChatGPT homepage may be better than ChatGPT via POE.
–
If you use results trained on GPL‑licensed code, wouldn’t that outcome also be subject to the GPL licence? I haven’t tried it myself, so I’m not sure.
It is a good idea to examine that stance. The problem is that genie is out of the bottle and not going back in. There’s lot of good things these new tools can do. If you let them do that.
Right now I also see that people not using AI are insulted by calling their work low effort and AI. I also see people spending lot of effort with AI present and undervalued. This is as bad as ai bros telling people they lose their work to AI.
I’ve spent lot of my time improving my thinking skills with AI. I also have improved my writing and communication skills same way.
That’s why the visceral reaction to “no AI slop”. Just agree that slop is not accepted, were it human or AI slop. There’s no need to differentiate.
In other fields, the situation is very different. In physics, I know a grad student that her dissertation is about using AI to study blackholes with an order of magnitude of less calculation, and a result that is acceptable as a very close preview of the strict way of computing it (it may take hours or days, IDK)
You seem to know a bit more than me. I wonder what could be achieve if you have your own machine and you train your model, could you use it to dsp graph creation, for example???
I unfortunately know very little. I’ve been finetuning LLMs though, and know that LLMs can do plenty of things where some kind of linguistic effort is required. Aside that there are other AI algorithms that can also achieve neat things.
It could be possible to generate DSP graphs with AI. I think it depends on what you want to achieve. Somebody could know if you elaborate what problem you’re wanting to solve or what kind of input or output you expect?
You can choose a chatbot from Explore, and then start a chat with one of them.
You could also try the following service:
–
how can you trust those things?
I check the results using my own knowledge and by consulting other sources.
Regarding Latin and French:
When I try to understand texts in Latin and French using dictionaries and grammatical knowledge, it takes too much time. Using AI reduces the effort considerably. For text comprehension, I let AI identify the meaning of each word with its inflection (declension or conjugation), then compare this list with the text in context. When something does not match, I consult dictionaries and, if possible, a published official translation.
It is often not straightforward, and it still requires effort and time, but it is more efficient and faster than working alone with only a dictionary.
Yes, “human slop” exists, but there is difference in scale. With AI it takes almost zero effort to generate large amounts of code or text in a way that is not possible otherwise, so it feels worth calling it out specifically.
Then there is also the humanistic aspect. The forum is a place for human exchange and we should try to preserve the human element as much as possible, with all its imperfections. If someone could not participate without AI due to language barriers, that’s fine. But in that case they should still try to write every sentence and only use an LLM for translation/editing.
Personally, I’m really against AI prompting for communicating with people because I find it dehumanizing. If one doesn’t put in the time to use their own words, why should I take the time to engage? Also, the overly slick style of AI prose often reminds of marketing speech, which makes it even more depressing. I strongly prefer people writing in their own personal style with all their little idiosyncracies. Of course, that’s a personal moral stance and others might disagree with it.
I’ve spent lot of my time improving my thinking skills with AI. I also have improved my writing and communication skills same way.
That’s great! Don’t get me wrong: I also use AI for certain tasks and it can be very helpful. Then again, I also see the potential (and real!) dangers to society, so I probably fall more into the AI sceptic camp. All that being said, using AI prompting to communicate with other people on a forum is crossing the line for me and that’s why I agree with the proposed guidelines. I want to stress again that the proposal does not ban AI use – except for obvious “slop” – it only discourages it.
I have a little rant about this. It’s a bit off-topic perhaps, so I’ll “hide details” so you can skip it if you’re not interested. TL;DR is that a holistic examination of LLMs quickly gets darker than genies and bottles.
Oh well, I tried
It is very important here to remember that the reason why OpenAI, Anthropic, Google, Meta etc. are creating these models is not for your benefit, or for society’s benefit.
It is 100% because they believe whoever implements AGI first will be insanely rich, and they don’t care who gets hurt. (More insanely, they don’t even care how short is the duration of time in which they will be able to enjoy that wealth.)
It’s problematic to write this, because AI enthusiasm reflects an extreme degree to which the (short-sighted) will of a small number of corporations is being forced onto everyone, while both ignoring pitfalls and masquerading as pragmatism – so, for me to express doubts about that coercion to a similar degree will look like extremism! (I half predict I’m going to be told here to calm down.)
But, read up on AI psychosis: the ways that chatbots reinforce the user’s words leading to a break with reality. Because LLMs can’t differentiate between sensible and delusional ideas, they reinforce everything. As we’re aware, audio feedback can amplify to the point where the audio is unusable. An unchecked, un-tethered-to-reality feedback loop of ideas can do the same, even to a prominent AI investor who ought to know better (really, read the quotes in the article – now imagine someone you care about going on and on about a “non-governmental system… originated by a single individual with me as the original target” etc.).
If chatbots are, as some say, “optimized for dopamine,” that’s… well… kinda like drugs. We regulate drugs when they’re shown to have dangerous side effects. (Continuing the audio feedback analogy, audio engineers identified ways to regulate feedback so that we can use it safely.) But, for some reason, with LLMs, mounting evidence of the potential for damage to mental health is met with a shrug and a “oh well, it’s inevitable, you can’t stop it.”
That may be true, in the sense that we’re now in a position where failure of LLMs to deliver on their promises would tank the stock market in a way that would make 1929 look like a yacht party, while success of AGI research is likely to result in the obsolescence of the human species. This is nuts. Seriously, nuts. I’m going to be told here that I’m overreacting, but concretely, in what sense is the current situation not insane?
Now, I’m not saying to ban outright. But, if AI-company CEOs are aware of people losing their grip on reality through use of their product (up to and including documented cases of suicide), if they are decent human beings, they would be appalled and self-regulate. If they can’t regulate, then society should regulate (same as you can’t go to the pharmacy to buy opiates over-the-counter, even though Purdue Pharma of OxyContin infamy would probably like it if you could). The fact that we are not taking this seriously is… not a good indicator.
If someone derives benefits from LLMs, again, happy for them. But “well I like them” doesn’t negate the damage or the need for regulation.
You’re reacting like you should. These companies aren’t exactly showing altruistic behavior. Unchecked these tools do cause harm such as amplify people’s thoughts out of their own control. There is also this problem of how it’s imposed as necessary when it affects everybody’s lives in dramatic ways simply by existing.
But how it’s going down onwards from here. Nobody knows yet. Usually it’s not black and white. It’s likely going to be something in the middle. Your distrust is justified, yet it doesn’t mean we still are only harmed in the process.
One person’s psychosis amplifier is another person’s thinking tool. I have no idea how to differentiate people who are prone to harm, but maybe somebody does. There are plenty of ways these tools have already helped people quietly. I agree that safety is important. It also ultimately helps to deploy more equitably as well.
I also think it’s injustifiable for companies to simply cause this kind of panic that we see. People are really worried for their future and I can’t tell it’s not justified. We lack nuance in dialogue and only talk in extremes. It really harms everybody’s lives and decision making overall. It’s something we all can work on together though.
@Henri_Tuhola For what it’s worth, I largely agree with your take, which strikes me as pretty well balanced. I focused on the negative in my post because there are serious problems, but LLMs are not the only product that has benefits and risks, and we don’t summarily ban everything just because of those risks.
It is frustrating/infuriating to me that these tech bros get to say “more AI!” while passing money amongst themselves and calling it revenue when it really isn’t, and differing viewpoints have essentially no power to object to anything. Who elected them?
I’m out of time on this topic for today – in any case, thanks for listening.
I’m also super tired to tech bros and technofeudalists. I neither signed up for world where my works are used to dehumanize people or remove them entirely.
The catch here is that is literally impossible. The confusion comes from using the word “intelligence” for those things. I’m not saying anything new, but neuroscientists (not to mention philosophers or CSs) who have studied this all their careers for decades will all agree on this.
But the illusion that it is possible can make you rich. They are, like, dollar signs $$$$!!
–
EDIT
I can’t not leave a reference to this text, that just feels so relevant now.
The virtue of formal texts is that their manipulations, in order to be legitimate, need to satisfy only a few simple rules; they are, when you come to think of it, an amazingly effective tool for ruling out all sorts of nonsense that, when we use our native tongues, are almost impossible to avoid.
Instead of regarding the obligation to use formal symbols as a burden, we should regard the convenience of using them as a privilege: thanks to them, school children can learn to do what in earlier days only genius could achieve. (This was evidently not understood by the author that wrote —in 1977— in the preface of a technical report that “even the standard symbols used for logical connectives have been avoided for the sake of clarity”. The occurrence of that sentence suggests that the author’s misunderstanding is not confined to him alone.) When all is said and told, the “naturalness” with which we use our native tongues boils down to the ease with which we can use them for making statements the nonsense of which is not obvious.
It may be illuminating to try to imagine what would have happened if, right from the start our native tongue would have been the only vehicle for the input into and the output from our information processing equipment. My considered guess is that history would, in a sense, have repeated itself, and that computer science would consist mainly of the indeed black art how to bootstrap from there to a sufficiently well-defined formal system. We would need all the intellect in the world to get the interface narrow enough to be usable, and, in view of the history of mankind, it may not be overly pessimistic to guess that to do the job well enough would require again a few thousand years.
Remark. As a result of the educational trend away from intellectual discipline, the last decades have shown in the Western world a sharp decline of people’s mastery of their own language: many people that by the standards of a previous generation should know better, are no longer able to use their native tongue effectively, even for purposes for which it is pretty adequate. (You have only to look at the indeed alarming amount of on close reading meaningless verbiage in scientific articles, technical reports, government publications etc.) This phenomenon —known as “The New Illiteracy”— should discourage those believers in natural language programming that lack the technical insight needed to predict its failure. (End of remark.)
From one gut feeling I derive much consolation: I suspect that machines to be programmed in our native tongues —be it Dutch, English, American, French, German, or Swahili— are as damned difficult to make as they would be to use.
Also, the resulting language can be very sterile and corporate-like, to the extent that the reader feels that they are talking to a machine (well, they actually are to some degree). This might be ok for technical things like GitHub issues, but not so much for the forum which is supposed to be a place for human connection. It’s just something to be aware of.
And I would also encourage you to write in english and use a translator tool only occasional or an old school dictionary once you become more familiar w/ the language.