@discobot is literally built with AI… and has been part of the forum from the very beginning. So I don’t think this is a very fruitful discussion
But in all seriousness, it seems like the main reason we wouldn’t want AI to participate in the forum as a peer to members would be a flawed viewpoint or skewed observation capabilities based on bad data or assumptions.
I would argue that all of us are in the exact same boat. We are all susceptible to the fundamental attribution error (shown first by Fritz Heider).
"Attribution theory (as one part of the larger and more complex Heiderian account of social perception) describes how people come to explain (make attributions about) the behavior of others and themselves. Behavior is attributed to a disposition (e.g., personality traits, motives, attitudes), or behavior can be attributed to situations (e.g., external pressures, social norms, peer pressure, accidents of the environment, acts of God, random chance, etc.) Heider first made the argument that people tend to overweight internal, dispositional causes over external causes—this later became known as the fundamental attribution error (Ross, 1977) or correspondence bias (Fiske & Taylor, 1991; Jones, 1979, 1990). "
Basically I am saying we all have flawed viewpoints based on our own experiences and it is human nature to exist in that flawed state, so is AI throwing flawed logic at us much worse?
Voice recognition (Speech recognition, voice to text), and bots are the first AI that I know of, and those technologies are still way ahead of their times. My friend sometimes uses ‘voice to text’ to send me messages, and it is not much accurate, I say that the AI he is using is naughty, because it often translates what he says wrong in a sxully suggestive kind of way. To think that kids use the same software on their phones.
AI is already widely used without disclosure. To disclose might make the viewer skeptical or dismissive? Some may intentionally use it to deceive, to not disclose is in itself deceiving. There’s so much fake/false/biased/unverified information floating around in cyberspace, what fact checking does AI do.
Sure it’s different in the science/computing fields where AI might have more credible uses.
There are so many daily posts on Growingfruit I don’t have the time or inclination to read them all. There are some members that have posts I tend to like and I know who has the most reliable information. That’s what I mostly read. If I wanted info from AI (I don’t) I’d just use Chatbot or other AI generators not Growingfruit.
In my opinion AI generated text is most harmful to beginners on the forum. Or even to experienced gardeners trying to learn about things new to them.
I don’t really see a lack of quick response on this forum during “peak time” Most questions posted on this forum can easily wait a day or even week.
The opposite seems true to me. From what if seen, the current AI still produces a lot of errors and or false information. Lowering the content Quality
I am all for AI assisting moderators. And there is some AI build into the forum that also assist users when posting.
AI translating to other language is also a great thing.
Actually my first and second post on this forum explained how to use the google translate plugin to translate text on this forum.
However, both these points that the AI generated text uses as an argument for using AI generated text. Are actually NOT about AI generated text. But other forms of AI
If we where to completely ban posting AI generated text (wich im not saying we should) The option for AI to assist moderators and using AI to translate would still be completly accsesible.
The AI is using a “straw man” argument here
This sounds like and can be a good thing.
However i personally enjoy the fact that due to this forum not “suggesting
to my preferences” or past searches" I’m constantly discovering and learning new things and about new plants/fruits and techniques on this forum.
Also the text generative AI’s currently in use are largely scored on ease of readability, engagement, emotional response etc. Not mainly on accuracy or actual usefulness.
It’s a known effect that this leads to more engagement/time spent. Which is profitable/useful for the AI company’s. But this also (often) can lead to echo chambers and feeding users more extreme and emotionally charged viewpoints.
Although it’s not a text generative AI. The YouTube algorithm AI is an example for this. Watch a single conspiracy video and see what it tries to “feed you”
Also known as (falling into) the “youtube rabbit hole”
This whole point again. Is another straw man argument. Since this argument posted as a argument for why “AI generated text is good for the forum” is actually not about AI generated text. but another form of AI.
Another straw man argument (dam this AI is getting repetetive)
This is not an argument for AI generated text. But for another form of AI.
This is part of the problem for me.
It takes a second to generate a large amount of “good looking” text and post it on the form with AI generated text.
It than takes a really really really long time in comparison for an actual forum user (human) to go over that text. And correct the mistakes in it. This will fatigue the human users on the forum to the point of not doing that anymore. And there goes the " measures to ensure accuracy and maintain the human touch that is crucial to community building." out the window.
speaking for myself. If only done it a few times here. But I’m already exhausted going over AI generated text posted on the forum.
It feels like arguing against a wall. Not to learn anything myself. Not to help or teach the wall. But just to warn other users to not run into it. (damage control instead of meaningful discussion.)
Sure human users of the forum also sometimes post mistaken information on the forum. But when you respond to that, you have a chance to help or teach another human. Or discover you where wrong and learn something yourself.
This is much more fulfilling than being an unpaid “error check” on AI generated text.
This forum is not a public service anymore than how Scott defines it, it is his private property. He determines policy and for a member to post a poll about policy out of the blue seems a bit of a power grab.
For controversial issues he has polled the members, but I’m not sure how controversial this issue is.
If someone posts an AI quote and identifies the appropriation, which I have done a couple of times and I believe started this issue, they may only be doing so because they believe the text to be accurate or that they want to check its accuracy with other members. I fail to see how information derived from AI is very much different than many other sources on the internet.
I get irritated when people post videos of on-line minor gardening celebrity gurus, but no one makes me watch them. If you don’t like responding to AI generated information, please scroll on as I do with the videos.
That’s my vote.
Incidentally, AI often provides answers accurately and does so much more quickly than an advertising infused Google search. I tend to use the tools that help me get the most work done in the least time. I have a lifetime-fulltime of experience with plants and soil as have others who engage with this forum- many more are highly experienced. AI is not submitting its comments if they appear here, we are. We should be free to do so.
You seem to be speaking “for Scott” here.
Scott has responded himself in this topic. And if not seen him post anything even suggesting he has a problem with this poll or topic. If he has I’m happy to edit or even remove it.
I think you have a problem with it, and are projecting that onto Scott here? Or am i mistaken? (i very well could be)
Either way, i think it’s best if we don’t try to “speak for others” (just like i did in the sentence above here. Purposefully giving a “bad example” to illustrate the point.)
I have not seen people asking for other members to check the validity of the AI generated text that got posted.
The posts with AI generated text, lacked clear quotes. So it wasn’t immediate obvious which part of the post was AI generated and which part wasn’t.
The prompt used to generate the AI text was not mentioned in the post. (see the figs grafted on dogwood example in this topic for a reason why that’s a huge problem)
When obvious errors in the AI generated text where pointed out. There was no response or correction. (which is not obligatory by any means. But to me clearly signals the text was not posted to have it “checked for accuracy”)
I think there is a clear difference between a link to an external video. And AI generated text posted on the forum. (especially if the AI generated text is not clearly quoted and lacks the prompt used to generate it)
If the AI text does not get copied but posted as an external link.
I would have no issue with it.
Or if it is clearly marked as AI generated at the top of the post, put in quotes to differentiate between the AI generated and non AI generated parts of the post and mentions the prompt. I also don’t see a huge problem with it.
I think it might be worthwhile to read this topic from the start than. Lots of examples given for where it is different.
If AI would be the same as other sources on the internet. It should also give credit the same. (including the prompt used)
An analogy of posting AI text without the prompt would be.
"
Google says:
text text etc.
"
Without giving the used search term.
or copying something from the internet. But not linking to the source.
I was speaking only for myself, and I don’t think starting this discussion with a poll is a good way to go about it along with it being disrespectful to Scott, whether he sees it that way or not.
You are putting the cart in front of the horse, at any rate. First the discussion and then the poll if a lot of members think it’s a good idea. If Scott is good with that, then it can go ahead.
might get philosophical here. But if Scott does not see it as disrespectful, how could it be disrespectful to him than?
anyway we’re going off topic here. What Scott thinks, is not the discussion here. If members should start polls like this is neither. (we could open another topic for that)
lets get back on topic. (if AI generated text has a place on the forum. And if so, how we should go about that) (irony again of my trying to keep it on topic )
If anyone is interested, here is the exchange that got this topic started. I believe the AI generated info was useful here.
Oscar, "if I’m not mistaken OHXF series has been propagated by cuttings. You might have bought such a rootstock already.
If had good success rooting cut of pieces of rootstock from Bench grafting of most apple rootstocks.
Often they are propagated from stoolbed since that’s less work/cheaper though.
I remember reading a paper where they said Gisella 6 could be propagated by cuttings. However my Gisella 6 rootstocks i think where from TC.
if also had good luck rooting quince cuttings.
Producing rootstocks from cuttings is very possible and has been done. Often it is more economical/cheaper to produce from stoolbeds. Since you can get a graftable rootstock in 1 year of growth from a stoolbed. From cuttings it can take 2 years to get a graftable caliper."
Oscar’s comment peaked my interest because I have no knowledge about propagating rootsock and wondered if there weren’t other fruit rootstocks propagated this way so I went to Chatai which gave me this answer in 10 seconds.
This was my followup to his comment.
I don’t propagate root stocks so didn’t know anything about growing them from cuttings, but according to Chatai it is a common practice. Artificial intelligence apps may not always be accurate (what info source is?), but they are still very useful to me for quick answers. Here’s what my robot tells me.
Common fruit tree rootstocks such as apple, pear, and quince are often propagated through cuttings, but the success rate can vary depending on the species and the specific cultivar. Here’s a general guide for each:
Apple: Apple rootstocks are often propagated through hardwood cuttings taken during the dormant season (late fall to early spring). The cuttings are typically taken from one-year-old shoots and are about pencil-thick. They are then treated with rooting hormone and planted in a well-draining medium. M9 and M26 are two common dwarfing rootstocks that can be propagated from cuttings.
Pear: Pear rootstocks can also be propagated from hardwood cuttings during the dormant season. Similar to apples, the cuttings are taken from one-year-old shoots and treated with rooting hormone before planting. Quince rootstock (Cydonia oblonga) is commonly used for pears and can be propagated from cuttings as well.
Quince: Quince rootstocks, such as Quince A (Cydonia oblonga), are often used for pears and can be propagated from hardwood cuttings. The process is similar to apple and pear propagation, with cuttings taken during the dormant season and treated with rooting hormone.
It’s important to note that while these rootstocks can be propagated from cuttings, the success rate may not be as high as other methods like grafting onto established rootstocks. Additionally, some rootstocks may have specific requirements for successful propagation, so it’s recommended to research the specific rootstock and follow appropriate propagation techniques.
So maybe cuttings are commonly used for two very common root stocks which is info I failed to find in 15 minutes of scrolling a google search.
This was Oscar’s response
"This is a perfect example of such a word association algorithm getting it wrong.
It probably read fruit tree growing texts. And quite often grafting to an established rootstock is advised for propagating fruit tree cultivars.
However in the context it put it in now. It does not make sense anymore. But the AI does not understand what a rootstock is, or what grafting is. It just looks at texts and tries to guess the next word in a sentence.
If you asked me how best to propagate a rootstock. And i said i propagate all my rootstocks by grafting onto established rootstocks. You would not trust anything i write afterwards.
please stop with the AI writing posts for you. Or more clearly mark it. (different colour or quotation marks) Warn at the beginning of the post that it is AI generated so we can skip it if we don’t want to read AI text."
How could I have marked it more clearly? I even prefaced it with my own doubt about accuracy, but it did include information that expanded on what Oscar started. I believe it was useful information that enhanced my understanding of the subject- that the very common M9 and M26 apple root stocks can be and often are propagated from cuttings.
could you link back to the topic this came from.
And use correct quotation. It is a bit of a mess now. Sometimes not obvious what is quoted text from me. And what you have written.
Also you seem to be missing some parts of the “post” that started this.
Also these posts have been linked already in the starter post of this topic. I don’t see the relevance of incorrectly/incompletely copying all this text and posting it here.
I edited nothing beyond pasting relevant parts that illustrate my point. I didn’t remove a word or sentence within the quotes. My comment was already too long.
If anyone cares, the comments are from Fruit Tree Planting Depth Question.
If you read the full comments, I’m confident It will show that what I pasted is not in any way misleading.
Alan posted
“I edited” “beyond pasting relevant parts that illustrate my point.”
“If you read the full comments, I’m confident It will show that what I pasted is” “way misleading.”
See how that’s way different from your original text. Or even from proper quotes like below.
This is an exaggerated example. (i did 3 cuts vs alans 4 cuts. His where over a significantly larger piece of text though) but signifies how copy pasting some-one else their text and cutting things out can become very misleading.
<< It feels like arguing against a wall. Not to learn anything myself. Not to help or teach the wall. But just to warn other users to not run into it. (damage control instead of meaningful discussion.)
Sure human users of the forum also sometimes post mistaken information on the forum. But when you respond to that, you have a chance to help or teach another human. Or discover you where wrong and learn something yourself.
This is much more fulfilling than being an unpaid “error check” on AI generated text. >>
I don’t really have a horse in this race. I’ve used ChatGPT a fair amount. Here’re my take-aways:
It is not infallible. It gathers information from on-line sources. Garbage in, garbage out. Moreover, it sometimes makes shit up. This is called “hallucination.” It its a well-recognized problem that may be solved eventually.
It is fast and easy. The two answers I posted above (re whether AI text should/should not be allowed) were generated within seconds after I posed the question. If you want to do an internet search for information, ChatGPT is WAY faster and easier. In general, it gets me good answers in 3 minutes to questions that might have taken me 3 hours or 3 days. No ads!
While the answers are not guaranteed to be correct, they seem to be better than the average of what you or I might find in a casual search. And if you suspect an issue and push back, the AI does not (unlike the average human) argue with you. It politely rechecks its answer and freely admits errors.
ChatGPT seems extremely good at applying known problem-solving methods. For example, if I ask “How many grams of Calcium Chloride do I need to add to a 6.5 gallon batch of mead must to produce a solution that is 150 ppm Calcium?” it gives me a fast and accurate answer. Obviously this is a real world example – been there, done that.
This is still a work in progress. Output must be carefully curated to ensure accuracy and utility. IMO, the user is responsible. Any AI-generated text should be carefully checked. Any published AI-generated text should be clearly attributed.
One final note that will make clear my tentative view: We have all encountered novice growers who appoint themselves the resident experts on a topic. Maybe they like showing off their new knowledge. Maybe they like earning money from ad revenue on FB. If you choose to interact, it can be extremely time-consuming trying to protect other (especially newbie) growers from their misinformation. IMO, output from an AI is likely to be much more benign. If we’re not gonna screen out dumb, misguided humans then we shouldn’t prohibit properly attributed text from an AI.
But what do we do than, when users don’t carefully check but still post AI generated text?
What do we do when they don’t clearly attribute the AI generated text? (for example when they don’t mention the prompt. Or put the AI part of the post in quotes.)
some users of this forum have said they use AI to save time. And posted AI generated text on subjects they claim to have no experience or expertise on. How can we than trust them to “carefully check” the text their posting?
This is why i think it would be a good idea to have a forum policy either discouraging the use of AI generated text. Or describing how to go about posting it. (at a minimum mention the prompt) or both.
)
