Some posts only go to subscribers via email. EXCLUSIVELY.
You can read more here or simply subscribe:
A few days ago, OpenAI released a new version of its GPT AI engine. It’s called ChatGPT, but it does much more than simply emulate or mimic chatting. I was wondering how to introduce it in this post and asked for its help.
Here’s the answer to “what can you do” –
As an AI assistant trained by OpenAI, I can assist users with a wide range of tasks and provide information on a variety of topics. Some examples of the types of things I can help with include: answering questions, providing definitions and explanations, offering advice and suggestions, providing information and facts, and much more.
Why is it called ChatGPT, then? Is it just a chatbot? I asked it again, and the reply was –
The name ChatGPT may refer to the fact that I am a language model trained using the GPT (Generative Pretrained Transformer) algorithm and that I am able to assist users through a chat interface.
Ok, enough ai-generated texts for now. I’m taking over and writing the rest. The message I have to share here today is too important to be left in the hands of a machine.
I believe this new GPT version (sometimes called GPT 3.5) provides us with a glimpse into the not-too-distant future. It could be a game-changer (but probably not just yet).
Don’t worry. Today’s post is not a doom-and-gloom prediction. At least, not for those web publishers who play their cards right.
Pssst: newsletter subscribers got to read this post a couple of days ago. If you want to get my posts directly to your inbox before they hit the blog, you can subscribe to the newsletter here.
Initial impressions: How good is this AI version?
We (my family and I) have been testing it for the last three days, and it’s wicked good. Much better than I had anticipated for this stage.
It answered my son’s ACT questions with ease (and correctly).
It generated code like a pro and even solved one of the questions my other son was dealing with in his PhD assignments (in computer science).
And more importantly, it wrote excellent answers to multiple questions we posed.
Impressive, to say the least.
It is so much better than previous AI writing tools. Not only are the answers well-written, but they also make sense. Real-world, actual people sense. Almost always.
Here’s an example.
The question: Can people walk on ceilings?
An older version of GPT (vis Jasper) answered:
Yes, people can walk on ceilings. In fact, it’s a great way to get a different perspective on things or to reach areas that are difficult to access. It can also be a fun way to workout as it requires you to use your core muscles and balance in ways that traditional walking doesn’t.
No, it is not physically possible for people to walk on ceilings. Humans and most other animals are gravity-bound, meaning that we are only able to walk on surfaces that can support our weight. The force of gravity pulls us downwards, so unless we are in a zero-gravity environment, it is not possible to walk on a ceiling.
See what I mean?
Both tools provide a well-written and coherent answer. Jasper’s reply is entirely wrong; a human writer would never have even considered writing that piece of drivel. However, the new GPT’s answer could have just as easily been written by a human.
But is it consistently that good?
Mostly. It’s not 100% foolproof. ChatGPT can still make mistakes, but they are rare, as far as I can tell. And hey, it’s not as if human writers never make mistakes.
Some limitations apply
Right now, the new AI is limited in many ways. It has no access to the internet and only knows the world from the datasets provided to it by the OpenAI research team.
For example, it can’t deal with questions about trends or world events. It also refuses to offer opinions on politics and other sensitive topics. And it’s programmed to refuse to share information on illegal or indecent.
By the way, it will claim not to know the answers to some questions, but often these are merely artificial barriers put in place by its developers. In fact, some people have found ways to fool the system with more elaborate prompts. OpenAI is trying to close these loopholes, and rumors are they may shut down the beta version if the problems continue.
But for the sake of this newsletter, let’s assume they manage to keep it open, and it becomes a viable tool that remains available to the public.
What does that mean for us as web publishers?
Will AI-generated content replace writers?
That seems to be the first question that pops into mind.
I actually think it’s the wrong question, and I’ll elaborate on that in a minute. But it is the more immediate question, so let’s talk about that for a minute.
As far as I can tell, this level of AI can produce fantastic texts at least 95% of the time. If I had to put a price tag, I’d say they’re the equivalent of 5-6-cents-per-word writers.
If you care about the quality of your content, then no, you can’t use it to create a hands-off fully-automated operation. Not any more than you’d trust a human writer off a marketplace.
To get the right level of quality that I’d like to see on our websites, I sometimes had to play around with prompts, add pointers on what to elaborate on, and generally do some light editing.
However, since I trust the answers more than I trust anything by the current industry-standard AI tools, it does offer a great option for producing content to scale. It’s fast and inexpensive (technically, it’s free right now, but I assume they’ll have paid options in the future).
So, will it replace writers?
I think there’s a very good chance that it will.
The people who are currently writers won’t necessarily lose their jobs. Instead, they will become AI operators, specializing in working with AI tools to generate top-quality content.
You may be wondering at this point what Google will say about this.
So far, the official Google position, as expressed by John Mueller is that you should not include ai-generated content in your websites.
I think that will have to change at some point. Google wants to provide the best search results for people. That means they should favor the best quality content.
Yes, so far, AI content was typically low-quality, but that’s about to change.
I can’t see how Google will penalize high-quality AI content. It will go against their core mission and won’t make sense.
I’m not using it on any of my established sites just yet. Just in case. But I will be testing it out on some sites and I have a feeling it won’t get penalized. Not because Google won’t be able to tell, but because the content will be good.
As a publisher, I look forward to incorporating this content into our websites. It will save on costs and will be just as helpful to users.
There’s something else that bothers me, though.
The future of informational niche sites
What most people call “niche” sites are, in fact, informational websites that focus on specific topics. These sites go after queries that people search on Google and (ideally) try to provide good and helpful answers to those questions.
And this is where AI will shine.
IMO, that’s the big news of this week.
A technological breakthrough has happened. OpenAI figured out how to create an engine that can easily provide good answers to most of the questions people ask Google. On the fly. Without having to decide on topics in advance.
Ask, and you shall be answered.
IMO, it won’t take long for Google to implement a similar algorithm and come up with its own AI-driven answers. Other search engines will attempt to do the same. Bing springs to mind, as Microsoft had already invested in OpenAI.
Imagine someone who wants to know how to groom their Golden Retriever.
In 2022, you ask Google that question, and you get a long list of web pages that describe how to groom a golden retriever dog.
At some point in the not-too-distant future, you’ll ask Google that question, and it will provide you with an answer. It will go something like this great answer from ChatGPT:
If we (web publishers) are very lucky, Google may still append a list of search results after that answer. But who’s going to stick around to click through to those (and why should they?)
The result: Traffic to informational web pages will drop significantly. Possibly down to zero.
When will this happen?
I don’t know. If you asked me a couple of weeks ago, I would have guessed 5-10 years. Now I’m leaning more towards 4-5 years.
I don’t have a crystal ball though, so who knows, really. But I’m pretty sure it’s going to happen, eventually.
What I plan on doing
I’m excited about the new developments in the field of AI. I think we all should be excited about technological advancements. They’re always good, even if they require us to adjust.
I can already see both short-term and long-term directions for my websites and my company.
In the short term, I plan to gradually embrace AI tools as they continue improving. Nothing is going to be automated. Instead, I want to figure out how to use prompts correctly and how to check and edit the results. Then train my team to do that.
It’s going to be a gradual process.
I won’t be deploying AI content in any of our big sites just yet. Only in smaller sites where we don’t have a lot to lose. We can also use these texts in info products and mailing lists.
And that brings me to the long-term strategy.
In the long term, we need to diversify our traffic sources and ultimately wean ourselves off the teat that is Google’s search traffic.
Fortunately, this is not a new concept for our company. We’ve worked through 2021 and the first half of 2022 to establish a mean and lean content-generation machine.
We reached and exceeded our revenue goals but realized that we were a one-trick pony. I know how to create successful sites that follow the long-tail strategy. But that’s not enough. Relying on this one source of traffic is risky. Always has been and always will be.
Regardless of AI, we were facing two threats –
- Google algorithm updates
- Increased competition
That’s why we are already in the process of pivoting and trying to add more traffic sources. Our goal for 2023 is to get to a point where at least a third of our traffic is coming in from sources other than Google. And then increase on that.
That’s why I see the improved AI capabilities as an opportunity, more than anything. It can help us generate high-quality content for mailing lists, social media outlets, and possibly even for info products. We’ll need that if we want to diversify our sources of traffic as well as revenue.
Challenges and Opportunities Are often Mixed Together
Overall, while using AI in the web publishing industry brings new challenges, it also brings many positive opportunities for producing better content faster.
I believe that it is up to us, web publishers, to carefully consider both the challenges and opportunities of AI and use it to benefit our content and diversify our audience.
And yes, the above two paragraphs were written by ChatGPT (I think the default pompous tone came across?)
Credit for prompts and editing goes to me, though 😉 Us humans are still good for something!