I do searches online for gardening topics all the time. I will search key phrases to see if I can come up with topics that people seem to be searching for or gardening questions that may be trending. I even search specific ideas to see if there is new information or research that I’m not aware of. And more and more frequently I’m noticing a disturbing trend.
As I look at the search results, often times I will see multiple pages in a row, usually sponsored results, that have the exact same key phrases as the meta description, the little blurb you see under the link that tells you what’s on the page before you click it. When I click on one of those pages I find the page to be full of ads, with very little substance in terms of gardening know how. And the verbiage and style of writing is very mechanical, disjointed, and generic. If I click another one, the page layout is exactly the same and so is the wording. These pages are all designed to look like legitimate gardening blogs or how-to sites, but they all have two things in common – they are meant to earn money from the overwhelming number of ads being displayed and they are all written by artificial intelligence.
Today on Just Grow Something we’re diving into AI and gardening advice. Can AI be helpful when asking questions in the garden? Are those answers accurate? And how can we determine whether a source is reliable or not? Let’s dig in.
References and Resources
Vote for me in the Women in Podcasting Awards!
Save on a Magic Mind Subscription with code justgrowsomething20
Just Grow Something Website (justgrowsomethingpodcast.com)
00:00:00
Welcome to Just Grow Something, where we dig into all things
00:00:03
gardening. Today, we're venturing into the
00:00:05
fascinating intersection of technology and horticulture.
00:00:08
Have you ever wondered how artificial intelligence might
00:00:11
help you grow a better garden? Whether you're a beginner
00:00:13
seeking a quick advice or an experienced gardener looking to
00:00:16
solve a tricky problem, AI tools can offer insights and solutions
00:00:20
tailored to your unique gardening needs.
00:00:23
In this episode, we'll explore how AI is changing the way we
00:00:26
garden, from diagnosing plant diseases to suggesting the best
00:00:29
plants for your specific environment.
00:00:32
We'll discuss the benefits, potential pitfalls, and how you
00:00:34
can incorporate these tools into your own gardening routine.
00:00:38
So, if you're curious about how technology can enhance your
00:00:41
green thumb, stay tuned. This is an episode you won't
00:00:44
want to miss, and if that sounds like it was written by a
00:00:49
computer, you are absolutely right.
00:00:52
I asked ChatGPT to write me an introduction to a podcast
00:00:56
episode that talks about using artificial intelligence to get
00:00:59
gardening advice, and that's what it spit out.
00:01:02
I mean, it's not bad, but it's definitely puts a very positive
00:01:07
spin on the use of AI, right? I do searches for gardening
00:01:12
topics all the time. I will search key phrases to see
00:01:15
if I can come up with topics that people seem to be searching
00:01:18
for or gardening questions that may be trending.
00:01:21
I even search specific ideas to see if there's new information
00:01:24
or research that I'm not aware of.
00:01:27
And more and more often I am noticing a disturbing trend as I
00:01:32
look at the search results. Often times I will see multiple
00:01:37
pages in a row or multiple results in a row, usually
00:01:41
sponsor results that have the exact same key phrases as the
00:01:47
meta description. That's that little blurb that
00:01:49
you see under the link that tells you what's on the page
00:01:52
before you click it. So when I click on one of those
00:01:54
pages, I find the page to be full of ads.
00:01:58
Like overwhelmingly full of ads with very little substance in
00:02:02
terms of gardening know how and the verbiage and the style of
00:02:07
writing is very mechanical. It's disjointed and generic.
00:02:11
And then if I click on another result, that page layout is
00:02:15
exactly the same and so is the wording.
00:02:18
These pages are all designed to look like legitimate gardening
00:02:23
blogs or how to sites, but they all have two things in common.
00:02:28
They're meant to earn money from the overwhelming number of ads
00:02:32
being displayed, and they are all written by artificial
00:02:36
intelligence. So today on Just Grow Something,
00:02:39
we are diving into AI and gardening advice.
00:02:43
Can AI be helpful when asking questions in the garden and are
00:02:47
those answers accurate? How can we determine whether a
00:02:51
source is reliable or not? Let's dig in.
00:02:55
Hey, I'm Karen, I started gardening in a small corner of
00:02:58
my suburban backyard and now 18 years later I've got a degree in
00:03:01
horticulture and operate a 40 acre market farm.
00:03:04
I believe there is power in food and that everyone should know
00:03:07
how to grow at least a little bit of their own.
00:03:10
On this podcast, I share evidence based techniques to
00:03:12
help you plant, grow, harvest and store all your family's
00:03:16
favorites. Consider me your friend in the
00:03:18
garden. So grab your garden journal and
00:03:20
a cup of coffee and get ready to just grow something.
00:03:32
So these days it is increasingly likely that a gardener searching
00:03:37
online for answers is going to encounter content that has been
00:03:39
generated by artificial intelligence or AI.
00:03:42
So tools like ChatGPT and Googles Bert and various content
00:03:47
generation platforms are commonly used by websites to
00:03:52
produce articles and blog posts and even forum responses like
00:03:57
automated responses on a wide range of topics.
00:04:01
And this includes gardening. So there's a lot of reasons why
00:04:09
you are going to encounter more of this AI generated content
00:04:13
than before. I mean, there is widespread use
00:04:16
of AI and content generation or content creation, right?
00:04:19
Many websites, especially those focused on producing really
00:04:24
large volumes of content, tend to use AI to generate articles
00:04:29
very quickly. So this trend is especially
00:04:32
common in areas where there is high demand for information, and
00:04:36
gardening is one of these. The other reason you might see
00:04:41
it more frequently is because AI generated content tends to be
00:04:45
optimized for search engines, meaning it is designed to rank
00:04:50
really highly in those search results.
00:04:53
So even if it's not a sponsored result, meaning they paid to
00:04:58
have it at the top of the results, it is otimized so well
00:05:02
that it's going to increase the chances that when you search for
00:05:05
gardening advice, an AI generated article might be among
00:05:09
the top results. Now, some websites use AI to
00:05:15
assist human writers. So they're creating content that
00:05:19
is sort of a blend of AI generated text or responses and
00:05:24
then a human who is editing it. So this is going to make it a
00:05:27
little bit harder to figure out whether or not a piece was
00:05:29
actually written by a person or if it was written by AI and then
00:05:33
just edited by a person. And then there are a lot of
00:05:37
content aggregation websites out there, websites that like pull
00:05:41
content in from all kinds of different sources.
00:05:44
And these of course might include AI generated articles,
00:05:48
especially if they're focused on very quickly providing this
00:05:52
information. They're not necessarily sorting
00:05:54
it. And this is an automated process
00:05:56
that's just going to pull that information in from wherever I
00:05:58
can grab it. And then if you are searching a
00:06:02
question that is fairly broad or basic in nature, like how to
00:06:08
plant tomatoes, it's more than likely going to return more AI
00:06:12
generated content because these types of articles are easier for
00:06:19
AI to produce. And we'll talk about how that
00:06:22
works here in a little bit. More specialized or very complex
00:06:27
questions are still going to require more human input and
00:06:32
expertise. And so you have a lower chance
00:06:35
of getting AI generated content returned to you.
00:06:38
So if you are doing a search for something online and you do not
00:06:42
want to see AI generated content, be very, very specific
00:06:48
in your question and dive really deep into it and see if you can
00:06:52
get very specific and so you can get better information back.
00:06:55
Now, it's not guaranteed that every search is going to lead
00:06:58
you to AI generated content, but the chances are really, really
00:07:01
high, especially for the most common gardening questions.
00:07:05
And so the fair question is, do I use AI for any of these
00:07:10
episodes or for the content in my website?
00:07:12
And the answer is yes, but it may not be in the way you might
00:07:16
think. Like I mentioned in the
00:07:17
beginning, I do online searches for topics to talk about, and
00:07:21
I've often used AI to list out the most common gardening
00:07:26
questions being asked during a particular time period or
00:07:29
regarding a specific topic. I also use AI to organize my
00:07:34
material for me, so I'll take the content that I plan to talk
00:07:38
about and have it create an outline for me to follow or
00:07:41
organize the material into sections for an article to write
00:07:46
that make the most sense. So in this way, I use AI as a
00:07:51
sort of virtual assistant, a free virtual assistant I might
00:07:55
add, which is great for the budget and it saves me time
00:07:58
putting together episodes and articles.
00:08:01
But I don't rely on AI for the information itself.
00:08:06
Number one, I want you to get information from my perspective
00:08:11
and information that I have researched myself and verified
00:08:15
the validity of. I've also seen too many times
00:08:18
where I've just used a simple search in Bing which now gives
00:08:22
you an AI generated summary at the top of the page before
00:08:26
showing you the search results and it has given completely
00:08:32
incorrect information about whatever question I have asked.
00:08:36
So no, I don't trust the machine to do my writing for me, for the
00:08:41
podcast or for my blog post. And Speaking of those blog posts
00:08:46
and the search results that had tons of advertisements all over
00:08:50
them, I do have ads on my blog post.
00:08:52
Nothing overwhelming, but they're fed from Google just to,
00:08:56
you know, make a few pennies to help keep the lights on in my
00:08:59
office. And I choose how many ads are
00:09:02
placed on the page so that it's not overwhelming.
00:09:05
The same goes for the ads that you hear in the podcast that are
00:09:09
automatically inserted. I get to select which topics I
00:09:13
don't want advertisements for, so you shouldn't ever hear any
00:09:17
ads for medications or anything because I choose not to have
00:09:22
those included. And anytime you hear me mention
00:09:27
a product of any kind, like anything that I talk about in my
00:09:31
own voice, that is a product I have actually tried and love or
00:09:35
plan to order regardless of whether it's an affiliate, a
00:09:38
sponsor, or there is no compensation at all.
00:09:41
Magic Mind has been at the top of that list the past few weeks
00:09:44
because not only is it peak farming season, I'm doing this
00:09:48
podcast, I'm training for a marathon, and then I have 3
00:09:52
ultra marathons right in a row behind that.
00:09:55
Plus we have a couple other things brewing in the
00:09:57
background, and that all takes a whole lot of energy and a whole
00:10:00
lot of focus. The ingredients in Magic Mine
00:10:03
compound their effects overtime, which is what makes it so
00:10:07
magical. The nootropics and the
00:10:10
adaptogens do start to work right away the first day you
00:10:12
take it, but after seven days they have really hit their peak
00:10:16
and it's just sustained energy. No spikes and crashes like you
00:10:20
get with energy drinks or caffeine shots.
00:10:23
Just even flow with natural science backed ingredients.
00:10:28
And you know how I feel about evidence based goodness.
00:10:32
If you want to take the Magic Mind seven Day challenge like I
00:10:34
did, head to magicmind.com/just Grow Something and use Code.
00:10:39
Just Grow Something 20 to save big on a subscription or 20% on
00:10:44
a one time purchase. Take it for seven days and you
00:10:47
will be amazed at the difference. magicmind.com/just
00:10:51
Grow Something with Code. Just Grow Something 20 at
00:10:55
checkout. And As for those AI generated
00:10:58
pages with all of their ads, not all of them are 100% AI.
00:11:05
Yes, there are plenty of pages that have been coded to
00:11:10
automatically generate based on keyword searches and optimize to
00:11:15
hit the highest rank in the searches, but there are some
00:11:18
that are sort of hybrids. So how can you identify whether
00:11:23
gardening advice has been written by AI or an actual
00:11:27
knowledgeable person? First thing is to check for
00:11:31
depth and detail. So AI generated content usually
00:11:37
just kind of provides generic advice.
00:11:40
It's going to lack any depth, right?
00:11:42
It might cover the basics really well, but it's going to struggle
00:11:46
with more nuanced or complex issues.
00:11:49
Typically, if it's a person writing it, especially an expert
00:11:52
in the field, it's going to include some detailed
00:11:55
explanation, some personal experiences, maybe some case
00:11:58
studies. If you're getting very
00:11:59
scientific about it, you're going to get insights that go
00:12:02
beyond just like the basic information that you can find
00:12:05
anywhere. It's going to address specific
00:12:07
challenges or provide more advanced tips that you may not
00:12:11
see elsewhere. The other thing is to look for
00:12:15
more of a personal touch, I guess.
00:12:17
As you you know, saw in that first paragraph or two that I
00:12:22
read for the intro, AI generated contact usually lacks a personal
00:12:28
voice. Now it's getting better, but it
00:12:33
still tends to be just more clear and concise and
00:12:36
informative, but is very sort of impersonal or genetic.
00:12:41
It's not going to be able to share, like, a personal anecdote
00:12:43
or, you know, an observation about something.
00:12:46
Obviously, people can do that. Expert written content is going
00:12:49
to include, like, personal touches and opinions or
00:12:52
references to the writer's own gardening experiences.
00:12:56
So this makes the content a little bit more engaging and,
00:13:00
you know, it seems a little bit more trustworthy when it seems
00:13:02
like it's coming from a person. The other thing that you can do
00:13:06
is check for repetition and redundancy.
00:13:10
This is probably the thing that I see the most.
00:13:12
When you have articles that are generated intentionally to meet
00:13:17
a specific word count, it tends to repeat the same points over
00:13:23
and over again, just in different words or in different
00:13:27
ways, or it just includes really unnecessary filler content.
00:13:32
So this is a sure sign that they're just trying to meet a
00:13:35
specific word count in order to fit that page.
00:13:38
If you've got an expert writing some content, it's going to be
00:13:41
fairly concise and to the point. You're not going to have a bunch
00:13:43
of unnecessary repetition. Everything is going to clearly
00:13:46
be serving a purpose. It's not just filler words,
00:13:49
right? And then, you know, kind of
00:13:52
going back to the fact that AI content is very broad, the
00:13:58
advice itself is going to be very generalized.
00:14:02
It's going to lack specifics that address any type of a
00:14:07
unique circumstance. If if you were to go, you know,
00:14:10
onto my website and do a search for say nutrient deficiencies in
00:14:15
plants, all of those articles that come up are going to
00:14:19
include specific details that show a deep understanding of the
00:14:25
subject. It might talk about particular
00:14:27
soil types. It might talk about climate
00:14:28
considerations. If you look for pest management,
00:14:31
it's going to talk about strategies that are specific to
00:14:34
certain pests. AI generated content just isn't
00:14:38
going to have that depth of information.
00:14:41
And then of course, the tone again, is just got to be off
00:14:46
like it, you know, AI might produce content that is, you
00:14:52
know, quality, but it's very formal or it can be very
00:14:56
awkward. Again, it is getting better.
00:14:58
And honestly, I mean, there are some writers out there who write
00:15:01
like they speak. And so it also tends to be a
00:15:05
little awkward, but at least it tends to be more natural.
00:15:09
And at least it reflects the author's personality.
00:15:12
It's, you know, engaging. It's, it's tailored to who's
00:15:15
reading it, not just a, you know, a generic audience that
00:15:18
might just, you know, do a simple, a simple word search or
00:15:21
keyword search. When you're on these sites, try
00:15:26
to check the credentials right? There should be, you know, a
00:15:32
page and about us page. If, if there's nothing there,
00:15:38
there's no author byline, it's published under a generic byline
00:15:42
like staff writer or something like that.
00:15:45
If it doesn't mention the the writer's bio at the bottom,
00:15:50
there's no background or qualifications.
00:15:53
It's very possible this is being written by AI.
00:15:55
If it's a, if it's a single individual who is, you know,
00:15:58
running the entire website, like my website, you're going to see
00:16:01
an about Us page or about me page or something that's going
00:16:05
to include information about who this person is, what their
00:16:08
expertise is, what their qualifications are to be writing
00:16:11
about this subject. You know, references at least to
00:16:14
their gardening experience or something.
00:16:16
But if it's all AI content, that's not going to be out
00:16:19
there, right? And then if there is a really
00:16:23
large volume of articles, especially across a bunch of
00:16:29
different topics, if the website itself covers like a wide range
00:16:34
of subjects and it updates really, really frequently, it
00:16:38
very well may be using AI. If you have one person or even
00:16:42
if you have a staff of like 5 people who are contributing
00:16:46
articles to a website, it is not very likely that those are that
00:16:51
that website is going to be publishing five articles a day
00:16:56
all about the same subject or five articles a day on a wide
00:17:00
range of subjects. Unless it's a national
00:17:02
publication like the New York Times or something like that,
00:17:04
right? You're going to find that if
00:17:08
it's written by an expert or somebody who knows what they're
00:17:11
talking about, or at least a actual human being, you're going
00:17:14
to find it on some pretty niche websites.
00:17:16
You're going to find it on blogs, platforms that are
00:17:19
focused focused specifically on gardening or like in the terms
00:17:23
of those larger publications, a gardening section of a larger
00:17:27
site. But they're more likely to
00:17:29
feature content that is written by people who have real world
00:17:33
experience or whether they're just enthusiasts or they are
00:17:36
actual professionals. But it's not going to be AI
00:17:40
generated. So just kind of review the
00:17:43
content strategy of the website and you might be able to figure
00:17:48
out whether or not it's AI generated or it's actually
00:17:50
expert written content. So you might be able to better
00:17:54
judge whether the gardening advice you're reading is coming
00:17:57
from a bot or if it's actually coming from a knowledgeable
00:18:01
human. Now, is all the information you
00:18:08
get from AI going to be bogus? No AI generates answers to
00:18:15
questions through a combination of techniques that involve
00:18:20
natural language processing, machine learning, and vast
00:18:25
amounts of data. So AI models like ChatGPT are
00:18:30
trained on very extensive data sets.
00:18:33
And this is, you know, books and articles and websites and any
00:18:37
other written resources they can get their hands on.
00:18:39
I mean, not their hands, not the AI hands, the the trainer's
00:18:43
hands, the human hands. So these data sets are, you
00:18:47
know, basically a wide range of topics and it lets the AI learn
00:18:53
the structure and the content of human language.
00:18:57
So AI models, so GPT, the GPT in ChatGPT stands stands for
00:19:04
generative pre trained transformer.
00:19:08
So these AI models are pre trained on a general data set
00:19:13
that covers a wide variety of topics.
00:19:15
And so the idea is to get it to generate text that is
00:19:19
grammatically correct and that seems appropriate to the context
00:19:23
of the question. And they continue to learn and
00:19:26
they add more data to that database to be able to answer
00:19:30
more completely the next time the question is asked.
00:19:34
The problem is that these are all probability based responses.
00:19:39
AI doesn't always know the correct answer, so it generates
00:19:45
responses based on probability. If it's not sure, it's just
00:19:50
going to give you like a general or very vague answer that covers
00:19:54
multiple possibilities, and it can sometimes produce answers
00:19:59
that are completely incorrect if it's finding information in its
00:20:03
data set that is incorrect. But it has been repeated so
00:20:08
often, meaning it has been published and written over and
00:20:11
over and over again. That it's now integral to that
00:20:15
data set, it's going to repeat that information because it
00:20:17
doesn't know any better, right? So until the correct information
00:20:22
is more prevalent in the data set, then the incorrect
00:20:26
information, AI is just going to keep regurgitating the wrong
00:20:31
answer. So AI generates answers by
00:20:35
predicting the most likely sequence of words based on the
00:20:41
patterns that it has learned from the vast amounts of data
00:20:45
that it's been trained on. It doesn't understand.
00:20:49
It uses algorithms to produce responses that are supposed to
00:20:54
be relevant, coherent, and in the proper context.
00:20:58
It just may not always be accurate.
00:21:02
Until next time, my gardening friends, keep on cultivating
00:21:04
that dream garden, and we'll talk again soon.
00:21:07
Thanks for listening to another episode of the Just Grow
00:21:09
Something podcast. For more information about
00:21:12
today's topic and to find all the ways you can get in touch
00:21:14
with me or support the show, go to justgrowsomethingpodcast.com.
00:21:19
Until next time, my gardening friends, keep learning and keep
00:21:22
growing.