Levente Andras

Philip McGarvey wrote:I'd rather LLMs didn't exist but while they do they can be useful for finding information.  Some kinds of information are hard to find with a normal google search but LLMs can find them easily.  (Google often gives lots of fluff websites rather than original sources)

Still, take care not to replace human connections with LLMs.  I respect anyone who chooses to stay off the internet and learn from real people, or goes in that direction in some way.  Meanwhile, to the degree we do use the internet to find information, LLMs are useful, and there are a variety of ways to use them.  Getting them to give direct links to original sources is the most useful I think.

For anything that's likely been discussed on permies I'd start by googling "site:permies.com blah blah".

For example, I recently successfully convinced a friend to put in a wood floor instead of vinyl floor.  I Googled site:permies.com vinyl floor and read all the discussions that seemed relevant.  I also queried an LLM with "how long does vinyl flooring last, and does it shed toxic compounds into the room?  give direct links to sources for all the information you give"

Other example LLM queries:

"give me five direct links to pdf of the manual for ______"   And then I have the manual.  For a quick follow up I can give the pdf of the manual back to the LLM and say "give me exact quotes with page numbers for everything this manual says about ____"

Or e.g. troubleshooting a chainsaw there's a lot of information out there that won't be in manuals and the LLM can be very handy for giving next steps for what to try.  I could have made a post somewhere like arboristsite but it might take some time for useful replies to come in, and they might be mixed with people mocking me for not intuitively knowing what to do without asking.  I troubleshoot a lot of mechanical/tech things and I don't want to invest the time to become a thorough expert on all of them, LLMs are really useful here.

Or "are there people out there who think ____, and what do they say about it?  (even if they're kooky that's OK)   Give direct links to places where they're talking about it"

Or "give me direct links to pdfs of five scientific papers that discuss _____"  And again, give the pdf of the paper back with "give exact quotes and page numbers where this paper explains or shows data on _____"

At the moment, LLMs are not yet completely full of ads and stuff.  I imagine this will change rapidly and they might become far less useful.  But at the moment it's a relatively distraction-free way to find info from the internet in a simple text format with links to relevant original sources.

Your answer chimes with many of my experiences !!!

r ransom wrote:

When I ask AI for gardening solutions, they all speak with absolute certainty of the correct method - most of which I've tried and found wanting.  

When I research a given topic, I don't simply ask a question. I treat the interaction like a conversation with someone (sure, I know AI is not a person, but for the lack of a better word...) who has a lot of knowledge and is able to retrieve that knowledge in real time. Before asking the question, I provide as much background as I can, including what I already know on the topic, methods I've already tried, etc., as well as the hypotheses that I have already considered. So if there's anything "I've already tried & found wanting", it will most likely be excluded from the suggested solutions.

Christopher Weeks wrote:Has anyone ever asked one of these LLMs for information and had it just say "I don't know" or something to that effect? I haven't had that happen, which seems kind of weird.

I have come across situations when, after asking a question, I was told by the AI tool that there was limited information on the subject, and whatever info there was, was not supported with good evidence. Then it showed me the available info, with all the caveats.

Anne Miller wrote:There is nothing better than learning from someone with dirt under there fingernails.

Real dirt not fake dirt.

Stick with information from known sources.

Libraries are full of book by leading authorities.

I'm not sure I can agree with the first part. There are so many people "with dirt under their fingernails" whose only rationale for doing things in a certain way is "because that's how it's done" or "that's how my granddad taught me". They refuse to learn new ways and methods, and they refuse te see how the world around them has changed, and how those changes have rendered their way of doing things no longer appropriate. I'm not talking about abstract cases, but very concrete and specific ones, that I've had to deal with over the past 15 years or so. Very painful!

BUT yes to books by leading authorities!!

Cristobal Cristo wrote:AI is only as good as the multitude pseudo gardening/DYI websites loaded with ads from which AI takes its information. These websites are copies of each other in terms of contents and usually very vague and generic. Because I'm interested in empirical knowledge of real persons and things that work I use Permies as the reference and basically get serious information only from forums from real people with experience. The good thing is that sometimes AI uses Permies as the source, but at the end nobody will guarantee the answer is correct or some algorithmic delusion. I was disappointed so many times on specific questions related to IT or engineering that gave plain wrong answers that I can only imagine that the solutions to more abstract topics may be equally wrong. If AI had robots in all parts of the world that would be doing gardening, building, developing within local conditions and materials THEN it would have some merit.

In my experience, AI draws its information from far broader (and far more credible) sources than "pseudo gardening/DYI websites". When in doubt, you can ask the AI tool to reveal its sources.

Information shared on forums like this can be high-value, but sometimes it isn't. The correctness of the answers given by real people is not guaranteed either. I can give you a plethora of concrete, real-life examples of useless or outright bad information by real people, but I will refrain from that as it would take us into an endless discussion away from the topic at hand.

Having "AI robots that garden" is totally irrelevant - AI draws on information derived from human experience, or in other words, on human experience shared in various forms (books, articles, videos, etc.)

r ransom wrote:

The other big issue is the cost of using ai. My query above took about .3 to .5wh that's almost 30 min on my led light bulb.  

Given how many power supply companies depend on oil and gas, and with global events of march 2026, the price of running ai is about to skyrocket.   They will probably move to the next phase where they get the customers to pay for it sooner than expected. Although I suspect they were hoping to get more people addicted before that.

AI is not currently sustainable and wicked icky for the environment.  

Honestly, I'm not bothered about this. Regardless whether I use AI or not, people will go on using it, seemingly uncaring about its environmental impact, and too many people use it for too many nefarious, severely damaging purposes. Consequently, I will have absolutely no qualms of conscience over using AI to try to, for example, improve the longevity and productivity of my trees, or the ecological integration of my garden. The objection above reminds me of people criticising environmental activists for flying to international conferences on global warming. I feel it silly to refuse to use a certain technology for benign and productive purposes, when millions of people use the same technology for trivial or even harmful purposes. When I restrict my use of a given technology on ethical grounds, I allow "the bad guys" to win against me using the same technology... To put things into perspective: how would you feel about me objecting to the use of earth-moving machinery in permaculture projects, on grounds that the fossil fuels used by those machines - or the industries that manufacture those machines - are not sustainable?

Jay Angler wrote:

Levente Andras wrote:... specific to my own situation (climate, exact location, soil type, etc.).

I'm not sure AI will be right about all those things.

There's a permie living only about 15 km south of me. Their farm is further from salt water, at a higher elevation, but likely has similar soil. Their weather and growing experience is *very* different from mine.

The difficult answer is that you need to get a whole lot of people in your region to study and start practicing permaculture! Developing a varied and deep local data base seems like the better medium term plan.

As with many other things, when using AI, the principle is "garbage in, garbage out". If you provide it with limited / incomplete background information, you cannot expect to obtain answers that are truly relevant to your situation. BUT the same applies to Permies discussions! In my queries with AI, I ALWAYS make sure that I specify in great detail the exact context in which I operate, leaving no open door for misunderstanding, or discussion on generalities. I have to be convinced that the answers AI provides are totally relevant and specific to my precise context, and I won't close the conversation until I made sure that's the case. Of course, I can still suspect that there may be errors in the answers, even though they are specifically tailored to my context.

Christopher Weeks wrote:
People on Permies can be wrong too. Or their experience in a Mediterranean climate doesn't shed much light on how things will work for me up north. But it isn't ever tricking me into trusting that it's anything more than it is.

In my queries with AI, I ALWAYS make sure that I specify the exact context in which I operate, leaving no open door for discussion on generalities. I have to be convinced that the answers AI provides are totally relevant and specific to my precise context, and I won't close the conversation until I made sure that's the case. Of course, I can still suspect that there may be errors in the answers, even though they are specifically tailored to my context.

Sometimes, my knowledge or experience doesn't help me to make decisions on certain (more or less complex) issues. (Note: what follows has to be read as an argument made by the Devil's Advocate.) I can choose to decide based on a hunch / intuition - but I do prefer to be informed, so I can choose to post the question on Permies, and wait to see if I get a relevant answer (I might not). But since the decision HAS to be made, and preferably fast, I choose to ask AI. There is a chance that the answer I get may be wrong, but I run the same risk on Permies too, and the time it takes me to get the Permies' answers, sift through them, and decide what's relevant / useable, can be too long, and they can still yield no useable info. Hence I take the answer provided by AI and run with it, based on the principle that "sometimes, any decision is better than no decision".

Hello, Permies! It's been a LOOOONG time since I last posted here, and lots of things have happened since (I've started a new permaculture project, which is now in its 3rd year - but more about that in a separate post).

There have been a few reasons why I haven't posted on Permies - being busy / lacking time was one of them, but not the main one.

Over the years, I have turned repeatedly to this forum for advice on all sorts of issues - but with the emergence of AI, suddenly I'm given the opportunity to ask complex, potentially difficult, and very specialised questions on ANY topic, and - within seconds - to obtain answers / solutions that are relevant, reliable, and tailored / specific to my own situation (climate, exact location, soil type, etc.).

Obtaining information of comparable specificity / relevance and detail on Permies or similar forums would be next to impossible - there is no guarantee that relevant or usable info will turn up in the answers to my posts, and what does turn up has to be sorted through and organised, before I can arrive at something that I can use.

So my question is: has the role of this forum (or the way it is being used) been redefined / adapted in any way since the emergence of AI, and if so, how? I have a vague feeling that members' lived experiences - especially if described in adequate detail, and documented with quality photos / videos - may provide value - inspiration? other? - that cannot be offered by AI. I'd like to hear your opinions.