Subscribe To Clarity Lab
If you like this issue of Clarity Lab, sign up below to get new issues full of tactics, trends, and news to help you grow your knowledge commerce business. It'll make your life better. Promise.
Clarity Lab Issue 99 – Bro, do you even AI?
read time: 15 minutes.
research time saved: 4.75 weeks.
Hey there friend,
It’s not-AI-generated-Forest writing to you from a cozy cafe in Boulder, CO, while drinking a mocha and waiting for a rainstorm to come over the mountains and wash through town. (All of those are things AI can’t do, by the way.)
Speaking of AI, I’ve spent the last few months researching, testing, and playing with artificial intelligence. The AI rabbit hole is fascinating, unsettling, and concerning all at once, but I’m finally at a point where I’m ready to share my thoughts on it. I want to give you some perspectives on what’s going on with AI that could help you if you haven’t already gone down the AI rabbit hole.
So today, I’m doing something different than my usual short intro and a handful of short paragraphs about things I’ve curated online. Today, I’m going long. What’s below is a deep dive into a few facets of the AI explosion, how AI works, how it’s already impacting some online businesses like yours and mine, and some valuable perspectives to have when holding all of this scary-fascinating-terrifying-mind blowing AI stuff in your “organic intelligence” powered mind.
With how fast things are moving in the AI space, it feels like I’m a few years late to the “here’s my take on how AI will impact the world” hot take party, but when I go down rabbit holes as complex as this one is, it can take me some time to absorb it all and formulate some practical perspectives.
(If you ask ChatGPT to write on the topics I’m writing on today, it would have already completed a mediocre version of this article in less than 30 seconds. That’s not for me, though. I’d rather spend a few months researching all of this and then spend 6 hours writing and editing this piece for you. Spoiler alert: old-school human research and writing will still win the day on the internet, at least for a few more years.)
What is AI doing, how does it work, and why does this matter?
Let’s start here with something that I don’t see many people talking about in the frothy whipped-up frenzy of “OMG if you’re not using AI right now you’re falling behind!” hysteria in mainstream media and the posts of AI hustle bros on Youtube: what the hell are artificial intelligence apps like ChatGPT actually doing when you use them?
Understanding this is key to sorting out what, if anything, you will want to use AI for in your work or personal life in the future.
If you want a deep dive into this, I highly recommend this great article by Stephan Wolfram, one of the most intelligent humans on the planet right now, on how ChatGPT AI-type models work.
The bottom line is that ChatGPT, and other AI systems that use large language models (LLMs), aren’t really intelligent like humans. They don’t think like you and I do. They’re not even thinking at all. They’re calculating probabilities.
That’s cool, and what they create as output is impressive, but they can’t reason, they make shit up all the time and act like what they’re saying is true (these instances are called hallucinations), and they suck at math (AI and I share this in common.)
So what the hell is ChatGPT doing when you ask it questions and it writes lengthy responses insanely fast that read like it’s actually really, really intelligent, and that it’s actually thinking and processing things and spitting back smart-ish, long written answers that sound human in about 20 seconds?
Here’s the most important thing to understand about artificial intelligence apps that write human-sounding text, like ChaGPT: they are essentially autocomplete on caffeinated steroids injected inside Adderall pills that have been soaked in horse adrenaline.
I’ll say that again, minus the descriptive adjectives, because this is the key: AI is just crazy fast autocomplete. It’s just like when you start typing a search phrase into Google, and it starts guessing what the rest of your search query will be. That’s autocomplete.
If you type “Humpty dumpty sat” into Google, it will instantly generate 15 guesses about the rest of what you’re about to type, with the most likely version of your search query at the top of the list for you to click on and less likely options below it.
In the same way, when you prompt ChatGPT and ask it questions, it starts generating sentences (at an absolutely insane pace) using the same method as autocomplete. It writes one word, then runs some insane probabilistic calculations to figure out the next most likely word based on it having read billions and billions of webpages of text.
If it’s writing a sentence about Humpty Dumpty, it calculates the probability percentages of the next most likely word that would come after “Humpty Dumpty…” Its calculations might say, for example, that the word “sat” has a 47% likelihood of being the next most likely word after the phrase “Humpty Dumpty.” And it has a list of 6 or 7 other words with diminishing probability percentages after that.
Once you grasp that these AI tools that write are “autocomplete on steroids,” the next logical question is, “Where is ChatGPT getting the information it uses to complete sentences using these “probability calculations”?
The answer: billions of publicly accessible websites, discussion forums, and social media posts, including your website and things you’ve published on it.
ChatGPT has crawled billions of websites that were publicly accessible up to late 2021. That’s when they stopped the “training” of the current version of ChatGPT.
As of a couple of weeks ago, however, through plugins and a deal with Microsoft, ChatGPT now has access to the internet, giving it access to billions of updated webpages full of human-created content that it will scrape and suck up information off of like an evil sci-fi AI robot except that it’s a vacuum that hoovers up all of the things that we, you and me and millions of other content creators, have published so that it can run its calculations and mash our content all together into human sounding responses for people using the AI apps. Welcome to content creator hell.
Let’s put all the pieces together now, shall we? ChatGPT went out and crawled through billions of websites that have publicly available text. It used all of the text on those billions of sites written by humans to train its large language model.
This “training” allows it to run its probability calculations of “what word is most likely to come next after the one I just wrote.” This means that ChatGPT has crawled billions of pages of text created by humans, like blog posts, articles, news reports, Twitter threads, etc.
It’s using all of those things we’ve written to allow it to go, “Okay, based on all the trillions of sentences written by humans on this topic that I’ve read, this combination of words is most likely what a human would say next.”
ChatGPT does NOT think when you use it. It is not doing any human-like reasoning. It’s just going, “Alright, I’ve read all the shit that lots of humans have written on this topic, and I’m now going to mash all of it together and write something that is somewhat new-ish based on my calculations of what word is most likely to follow the previous word so I can write something that sounds a lot like a human wrote it.”
The fact that it does this so remarkably fast and that the output reads like a smart college student that got an A in writing, but a D in creating factual content, all of that is just scary and impressive when you watch it type an entire blog post in 20 seconds right in front of your eyes. (I found it deeply unsettling the first few times I watched it write entire blog posts before my eyes in seconds.)
Is AI going to steal your job or ruin your business?
Alright, so that’s how AI models work. The next topic I want to address is the fear many of us have felt about AI concerning how it will impact our jobs or businesses in the online business space. Is the fear warranted? Maybe…in some cases, but not necessarily in others.
Before I dive into this, it’s important to remember that intense fear drives a lot of clicks on the internet. Many people are taking advantage of this on social media, Youtube, and in articles on small and large news media publishing sites.
So, some of the content you’ve seen online about AI putting you out of a job or putting you out of business could be a bit overblown because the people that created those articles are attempting to get clicks and traffic for the gigantic advertising machine that makes money for the people creating this clickbaity content.
There are a few types of businesses and jobs that AI is already impacting. Some of my copywriter friends are already telling stories of clients coming to them, saying, “Why should we pay you $4,000 to write this sales page when ChatGPT can do it for less than $1.00 and get it done in under 1 minute?”
Unfortunately, the clients saying things like that to freelance writers don’t fully understand how ChatGPT works and don’t understand that the writing it creates is mediocre at best. And further, they don’t understand that to get that mediocre level of content from an AI app, you need to know quite a bit about how to prompt it to create halfway decent writing, whether it’s blog posts or sales pages or email copy.
It can’t compete with a seasoned conversion copywriter with decades of experience writing strong copy that gets people to buy what businesses are selling. But clients don’t know this…yet. Without knowing much about what makes excellent copy, or solid blog posts, clients saying the things I mentioned above are looking at the output of chatGPT and going, “Holy shit, that was fast and, hey, this is actually pretty good! It sounds like a human wrote this!”
So, copywriters and content writers who create content like blog posts for clients are already being impacted by AI tools like ChatGPT.
Graphic illustrators are also getting hit because tools like Dall-E and Midjourney can crank out impressive and good-looking digital art that can look photorealistic or in the style of almost any artist.
The AI tools that generate visual art use a similar autocomplete process that I described above, but instead of having crawled all the available text on the internet, these tools have crawled all the public visual art on the web, hoovering up all of the art that humans have made. Then they use it to mash up bits and pieces into some new-ish art that can look like a human created it.
Who else will be negatively affected by AI tools? Most likely anyone who has affiliate revenue as part of their income stream…like me.
This one hit close to home, as affiliate marketing has been a part of my business model for the last eight years. (Maybe I should try one of the new AI therapy assistants to help me process all the emotions around this challenge. I’m joking, but there actually are several AI tools designed to give people therapeutic support when you chat with them. As those tools improve, I suspect therapists and coaches will see an impact on their practices.)
But back to affiliate marketing. There are two ways that affiliate marketers with niche sites are likely to get negatively impacted by AI.
First, anyone like me who writes product reviews will now compete with other affiliate marketers who are cranking out hundreds of mediocre pulp content to cover topical maps on given topics. (A “topical map” is an SEO term that means you’re writing dozens or hundreds of articles about any topics that are related to a primary topic, which can help establish you as an expert on a given topic, and when Google thinks you’re a top expert, it will put your content above other people’s content in search results. Creating large amounts of topical map content is a big part of SEO work that affiliate marketers do to rank well in search results.)
*Forest takes a deep breath in so he can say this next sentence all in one breath* This means that if you’re doing affiliate marketing or anything that relies on pages of your site ranking well in search results so you get organic traffic from Google, if you’re not using AI tools to help you generate large numbers of articles or blog posts on a cluster of topics that will allow you to demonstrate to the Google bot that you’re an “expert,” or “topical authority” on a given subject, you’re going to fall behind the other people who are using AI tools to do this very quickly, which means you’ll lose whatever good search result positions you had. (Grammarly will throw all kinds of red alerts about that last sentence at me. Oh, and did you know that Grammarly is now integrating AI tools into its software? AI is everywhere all at once already.)
And here’s the second way that affiliate marketers, and anyone else that relies on search traffic for part of their revenue stream, will be negatively impacted by the explosion of AI tools: two weeks ago, Google made a massive announcement about the most significant change to the search experience they’ve ever made. You’ve likely heard the news by now, but for those who haven’t, here’s the scoop: Google will soon release a new search experience that will use their version of ChatGPT, called Bard, to generate a huge box of AI-generated content above the normal organic search results that link off to websites like yours and mine.
This news was like a hydrogen bomb going off in the SEO world. Why? Because this AI-generated content box above the normal search results will answer a large percentage of people’s search questions right in that box. If people get their answers from AI right in that box, they won’t need to click through to websites to get further information. And if they don’t click through to websites like yours and mine, those of us with business models that rely to some extent on organic search traffic will see a significant loss of traffic and revenue.
No one knows how much traffic certain sites will lose, but everyone in the SEO industry agrees it will be significant. The estimates I’m seeing from SEO experts I follow in the industry are that we’ll likely see somewhere between 30% to 50% fewer clicks going to organic search results (which is where your website normally shows up in the list of search results) for many types of search queries.
Coming back to affiliate marketers that rely on organic search traffic going to product reviews on their websites, these changes will hit this business model hard, especially for product reviews on physical e-commerce products.
Google is going to literally use product review content that people like me write to create a mashup of AI-generated answers about products people are searching for and will deliver a load of AI-generated content in that gigantic AI answer box above all the organic search results (which Google calls “SGE,” for “search generative experience.”)
And if Google’s AI bot succeeds in answering people’s questions about things like “the best 50-inch TV in 2023,” for example, right in that AI box above search results, a large percentage of people shopping for the “best TV in 2023” will not need to click through to the websites with deep-dive reviews of affiliate marketers who make money from people coming to their reviews and clicking through on their affiliate links to buy things like TV’s (or software in my case.)
I don’t want to sound overly apocalypsy, but this could be the beginning of the end of the affiliate marketing business model. I could be very wrong on this because we have yet to determine the full impact of what Google’s new SGE AI content box will be. Also, there will be a few links to source sites that the AI is drawing on in that AI content box. So, it might be that the SEO game will be massively different, and the new goal will be to get your website listed inside the AI content box.
At this moment, though, the coming impact of the new AI-powered Google search experience cannot be overstated. We don’t know yet how significant the impact will be for businesses that rely on organic traffic, but from what we SEO nerds can see from the demo of Google’s new AI search experience, we’re all pretty sure that the impact will be somewhere in the “wow, there’s going to be a lot of people that are no longer clicking through to our websites anymore” range of the spectrum.
And if that’s not making you feel the Earth shifting beneath your business model’s foundation, there is also this to factor in: there is already a noticeable drop in search traffic because millions of people are now using ChatGPT to search for answers instead of what they’ve used for decades: Google.
The AI wars have begun. Microsoft got the first shot off, and now Google is freaking out and trying to catch up. This war will have casualties, and they won’t just be large tech companies. This AI war will impact any business that relies on organic search traffic.
The bottom line: if you have a business that has a revenue model that relies in some part on organic search traffic coming to your site, and if you haven’t already started making plans for adjustments to your business model, now is the time to start, because the AI train is moving incredibly fast and it won’t be long until we all start seeing the impacts of lower search traffic.
Bro, do you even AI?
Should you use AI in your business? Are you, as a New York Times article I came across this morning said, “falling behind if you’re not using AI?” Mmmm…maybe, but it depends on your business model.
Suppose you’re in a competitive niche competing against many other people using advanced SEO strategies, which is where I find myself. In that case, you will likely not be able to beat your competition if you’re not using AI to help you create large amounts of inexpensively produced content to help your money pages rank well in search results.
Even if you’re not in the affiliate marketing space, if you’re doing old-school blogging as part of your content marketing strategies, you would likely see benefits if you learned to use AI tools like ChatGPT to help you create content faster.
Remember, though, that even when you learn to do solid prompt engineering to get ChatGPT to create the best content it can make, most of the time, what it writes still needs to be edited by a good writer/editor.
In my experiments with prompt engineering and content creation with ChatGPT, I find that even at its best, what it writes is a basic, mediocre, milquetoast level of writing. However, if you’re already a strong writer, what AI content is great for is getting you into the writing zone faster.
If you haven’t already messed around with ChatGPT, once you see its mid-level content creation, you’ll most likely go, “Huh. That’s pretty good for an AI bot. It’s pretty close in these areas, and I can make these bits better. It’s wrong in this section, but that makes me think of this common misconception about this idea, and I know how I can use that and turn it into useful content.”
In other words, even though ChatGPT’s content is often kind of “meh,” I find that it gets me right into the space of writing and editing a lot faster than starting with a blank page.
If you create a lot of written content as part of your marketing efforts, experimenting with ChatGPT 4 is worth doing. (The free version of ChatGPT uses the 3.5 version of the software. To get version 4, which creates much better content than version 3.5, you’ll need a paid ChatGPT account.) It’s fun to play with, and given where things are heading, knowing the basics of how AI software like ChatGPT works will come in handy down the road.
Significant changes are happening very, very quickly in the AI, SEO, Google and Bing search, and content marketing industries. I suspect that most of us with online businesses will be impacted in some way by the coming of AI tools. Some of us will face negative impacts, and others will see massive upsides.
Time will tell. Oh…and by the way, this entire article was written by ChatGPT 4.
Whoa. Did you feel that? It’s like the entire ground just shifted, and the space-time continuum warped right before you. Anyway, I was kidding. ChatGPT created a total of ZERO sentences in this deep dive article. I love writing too much to let AI do it for me.
I hope you enjoyed reading this as much as I did creating it!
Chief Organic Intelligence Officer
This link is a good article on chat GPT prompts
Thanks, Forest. Terrific article. As Chatty might say, “The casual, conversational tone of the blog post was engaging and approachable, which helped me not to feel overwhelmed by the complexity of AI.”
Or, as I might put it – Once again, I totally appreciate that you take the time to digest a hot topic and make it understandable to those of us who (through lack of exercise and an over-appreciation of French wine and cheeses) have turned themselves into human large language models.
As I develop the newly re-defined quality of “Promptness,” I have to laugh at some of the things I make Chatty say. However, the speed and the mere fact that it creates the stuff takes me back to the days when my Mac actually gave me information about good restaurants in New Orleans, even though I had never even been there! Or when the little red LED on my TI-2500 Datamath paused for a moment as it tried to help me make it through my early 70’s Stats class. I realized it was “thinking” about the answer, not just calculating it.
My first experiences with Chat GPT were like that. It’s been a long time since something computery made me go, “Wow,” but I find myself saying that often when I finally get the prompt right.
Something has shifted; something is forever changed.
I’m not sure what it is that has shifted or exactly how, but it has happened. I can only be grateful that, once again, I am along for the ride.
And also grateful that, like Justice Potter Stewert, I may not be able to describe the difference between AI-generated content and artfully written prose like yours, but “…I know it when I see it…” and I really appreciate you took the time to give it to us.
So glad you enjoyed the article, John. 🙂 And I’m so happy to hear that you appreciate me going down in the rabbit hole and coming back to explain what I found in simple language. It’s really helpful for me to hear that because I’m considering writing more of these deep dive pieces into other topics that are grabbing my interest (and I think there may be a part 2 of this piece on AI in the near future.)
And I’m with you on ChatGPT taking you back to some earlier experiences with new computing technology back in the day. I still remember the first laptop I had when I downloaded one of the first astronomy apps that would calculate the positions of the Messier objects in the sky. It was magic.
And yes, this arrival of ChatGPT and other AI tech that’s competing with it has changed the way we interact with computers and the internet and information in general. It will indeed be a wild ride.
So glad that we get to go on this ride together.
Thank you for the deep dive Forest, and great that we can participate publicly through comments.
I have used ChatGPT to create content and will continue to do so.
The content it creates unless you know how to prompt it comes out in a familiar pattern. So if you give it a word count for instance it applies it in funny ways – typically it resulted in paragraphs cut off mid sentence (just when it got to something interesting). Or it gives you this boring and predictable opening and closing paragraph which is just a repetition of the question. Worse than meh… it’s like reliving my high school essay writing just to get the word count in about a boring topic I never want to revisit.
AI created content, fails in creating content about mastery, especially if there are no masters in your field who have blogged about it.
My wife is a master hairdresser. I say master because of the level she operates at. Her mastery has been acquired through thousands upon thousand of hours working her craft. Her knowledge is not contained on any website, except ours. And that content has been created by me over many years, listening to her talk about her craft, watching her apply the technique to thousands of hair cuts and hair colours, year after year, so that I am now a self proclaimed hair expert who knows enough to be dangerous 😉
I could also say it was created with blood sweat and tears – hearing her talk, writing down bullet points, asking her to expand etc etc. She has dyslexia… so it is a burden to write but not a burden to talk. So now we video ourselves having conversations about hair. And you know what? It makes for really good content… I have tried to use the auto generated transcript to create written content, but the New Zealand accent is just something that has not been included into the overall programming of transcription bots. Even IVR voice systems struggle with her saying “yes”… it’s quite comical… “yis… YISSSS” she yells!
So when I ask my VA to create content about “hairstyles for women with fine hair” using ChatGPT, the absolute DRIVEL it came back with was the sum total of all the shite that people have been regurgitating for years… and it was all wrong.
But I knew this would happen. Because hairdressers who don’t have mastery get it all wrong.
And those people actually create bad hair, but can still tell a lie often enough so that when you see someone on stage grab a beautiful head of fine hair twist it around and start shaving the edges off to create texture (or any other stupid technique you see on YT, insta or TT)… most of you would be marvelling about how fabulous that all seems, while me and my wife gasp in audible horror, eyes wide open and mouth agape and we mumble to each other “do you think we should go back stage and give the model our card for a complimentary hair fix? It would make GRRREEEEAAAT content!”
Great technicians are busy building their craft and serving the people who want what they have to offer. The content they have is not in the trillion websites out there… it is on my head and the heads of her customers who see the value of paying $250 for a haircut and $300+ for a colour.
At some point she will stop. And we will monetise that content behind a paywall and she will be paid to talk about it. I know this doesn’t help people who want to mass create content BUT my point here is to extend Forest’s view point and help you form a balanced view and not one based on hysteria.
It also didn’t create content that matched our voice… because our voice is created by us.
It just created some lightweight, surface semi positive boring meh… I mean you just feel so icky reading that kind of content so you don’t bother.
BUT we are dealing with smart dumb robots in the game of getting people to your website for conversions…
You need to ask yourself what kind of purchase/thought process is my customer going to go through once they land on my website?
I have something called the 3 second sale that will account for 80% of our conversions.
The customer is constantly scanning for trust signals…
They find the content related to the search they performed – this requires semantic SEO – check
It is scannable and it answers their inner trust signals within a second – check
They see our location images and looks pretty fancy – check
They see images of our work that is RELATABLE TO THEM AND THEIR SEARCH – check
The Call To Action is super accessible – check
So my machiavellian viewpoint is if I can show the algorithm that I have something of use to say and I say a reasonable amount of it, I will use GhatGPT to save me time to create a framework for ideas that I edit a considerable amount to churn out content that feeds the smart dumb robots but remains faithful to our ethos.
It saves me a lot of time in generating outlines of what I could cover for any given topic.
And if I had to take a guess I would have used about 50% of the content as it was spewed forth. And I made massive gains in my content production as a result.
On the topic of ART
AI still relies on the mastery of the human artist. I look at a lot of AI art on instagram and I have developed an appreciation of the story created by the human artist with the aid of AI. And I feel I can discern beautiful AI art and average. Plus the human artist’s cleverness or personal opinions can still come through… I love AI art that shows our elders, our fat people, our homeless that are rarely palatable in mainstream art.
Would I want it on my walls? Ahhhhh no, but I might like to look at them in a book or magazine.
And because I have a passion for hand made objects. There’s an energetic quality to them. Would I be able to tell the difference? I don’t know and do I really care at the moment? Not really.
Yes this AI is a thing Bro.
It’s been brewing Bro… for a lot longer than we think bro.
I has got a lot more to say bro, but chur.
We just have to adult it and it needs to be schooled Bro!
Chur bro… as the kiwi’s would say… look up “The meaning of Chur” by Jimi Jackson on YT…
Thank you so much for your thoughtful comment about how all of the AI stuff has been landing for you and your wife.
I have a feeling, as you mentioned, that AI will continue to struggle to create any kind of useful content for the kind of skill and mastery that your wife has as a hairdresser.
In a way, that’s great for you guys because it’s highly unlikely that AI will ever be able to create written content or video content that would ever come close to what you and your wife can create.
AI is definitely fun to play around with, and it’s great to get the writing engine rolling for sure. But there are some things I don’t think it will ever be able to do…things that humans can do in the real world with physical things, like people’s hair. 🙂
Good luck with everything over there, and with the eventual transition to your wife monetizing her mastery through speaking and teaching about it online and in person.
Very nice article and timely.
I’ve been playing with ChatGPT for three months and have come to the same basic conclusions. Mobile Home Doctor, my blog about how to repair mobile homes, first published in 1998, will retire soon. My search for Ai “proof” blog topics is underway, with inconclusive results so far.
I really like the way Chat helps me surmount the “Blank Page” problem.
If you are well educated, with a good vocabulary, the prompt creation game becomes addictive. I’m not convinced that a prompt creation business will last beyond the current mania.
I’ll be sharing your article with my grandkids, hoping to provide a more informed perspective on what they are hearing elsewhere.
Glad to hear this article came at a good time for you.
That sounds like a great approach you’re taking to search for AI proof topics. It’s a whole new online world we have to inhabit now. We have a new competitor to constantly think of, one that will scrape all the content written by everyone else and then mash it up as answers to search queries.
I’m sure you’re already thinking along these lines with your search to find an AI proof topic to write about, but anything you can write on that involves your personal experience particularly with anything that deals with something physical and offline (which you had with mobile home repair) is something that AI will have a harder time creating content around, particularly if you included your own videos of you doing something in the physical world.
Just a thought.
And yes, ChatGPT is great to get the writing engine going when you’re not sure how to begin, even if you end up scrapping everything it writes, it still feels faster sometimes to begin with what it will generate and then start editing and changing things from there.
Thanks heaps for sharing my article with your grandkids, and good luck with the search for your next topic to create content around!