What's your take on AI generated content?

I just hopped on the bandwagon and did some ChatAI content.

Ran it through some GPT-2 ( of all things ) tools that detect AI written content... and guess what, all the GPT-2 stuff could detect it.

So you know internal Google/Meta/Amazon/MSN tools is going to be able to spot this, even if you hand edit and refine it.

Yeah, garbage.
 
As others have said, ChatGPT is not ready for long-form content.
Even if it wasn't detectable, it's not getting facts right. I can't in good conscience publish an article with that level of accuracy.

But I have found a few good use cases for it. Most of these replace VA tasks or simple Python scripts:

  1. Formatting. "ChatGPT, turn this list into a table with columns for the brand name, price, and release year." I give it raw data, explain what I want, and collect the output. Especially good for annoying little tasks.
  2. Top-of-funnel research. "ChatGPT, find me 5 quotes comparing Elon Musk to Mark Zuckerberg. Give me the author name and full URL for each." The research preview doesn't have browsing mode, but I'm gonna be using this one a lot in the full version. Verification is easy since it provides the URLs.
  3. Compiling Datasets. "ChatGPT, make a table of 10 popular car models with columns for model, make, horsepower, and price range." A lot of fact-checking for this one, but it's still a huge time saver.
  4. Social posts. "ChatGPT, write 5 sassy posts about Bill Gates." These are hit-or-miss, but if you know the niche you can recognize a few winners pretty easily.
  5. Debugging. "ChatGPT, look for bugs in this code."
  6. Writing basic code. "ChatGPT, write a python script that calculates the cost to build a wooden box from oak, maple, or birch wood."
 
  1. Debugging. "ChatGPT, look for bugs in this code."
  2. Writing basic code. "ChatGPT, write a python script that calculates the cost to build a wooden box from oak, maple, or birch wood."
I was with you all the way until you started with the coding stuff. In the future A.I. DEFINITELY should be able to write code based off of requested you've inputted, HOWEVER it will have to be basic stuff in the beginning.

Looking for bugs in code? What type of bugs are you thinking of. If there is a bug in some code the code will not run anyway or will give wild outputs, most compilers tell you what line and where the "coding bug" is. But logic bug - that's on the coder to think through that, and that's where you'll have issues. As well as the "vision" of what the code's purpose.

I do hope for a day where I can simple ask an A.I. to write code that does "XYZ" - that would make a lot of things easier. I just can't see it doing things big picture.
 
  • Like
Reactions: zak
Looking for bugs in code? What type of bugs are you thinking of. If there is a bug in some code the code will not run anyway or will give wild outputs, most compilers tell you what line and where the "coding bug" is. But logic bug - that's on the coder to think through that, and that's where you'll have issues. As well as the "vision" of what the code's purpose.
The example I used was unrealistic for sure, but here's how I've found it useful so far for coding.

I've been testing it in a few python scripting cases, primarily when using poorly documented APIs or libraries. It's able to find issues with the types of parameters sent, missing parameters, etc. You do have to be specific in the issue, but that's gonna be the case for all AI interaction for a long time to come.

It's also good at "This code works, but this one doesn't. Why might that be?" type cases.

As for code creation, it works best in the same places it works best for text: short and simple.
The more boilerplate, the better, preferably with a template that you can hand it. "Create get functions for 'B', 'C', and 'D', just like this one for 'A'".

I haven't worked with GitHub's Copilot, but I'd imagine the above ideas are its bread and butter, too.

So yea, I agree that it's not going to fix your logic. It's not a domain expert. As with the other AI use cases, treat it like a poorly-trained VA and give it your busy work.
 
I just started out on ChatGPT. I use it a lot of times several days. For me as somebody from The Netherlands which are sometimes to directly for especially US based persons. I asked ChatGPT for some longer form answers which are more to be liked.
 
Google executives reportedly told employees that the company does not plan to launch a chatbot to compete with OpenAI's buzzy ChatGPT because it's a "reputational risk."
Source

I think ChatGPT is a fundamental risk to niche website owners. This quote from Google is their defense to why they don't have a functional ChatGPT alternative. But from all the hype that this tool has received over the past few days, a lot of people are in fact using it to seek answers to their questions.

It's still a drop in the ocean as far as volumes go, but that is only going to increase over the next few years.

And what happens when ChatGPT is sharp enough to be a reliable alternative to Google in providing answers? Google is going to launch its own alternative.

I see Featured Snippets giving way to actual AI-generated results (with probably a disclaimer that Google does not endorse the response).

Except for a few niches where users seek experience or opinion-based content, this will be a solid alternative to clicking on websites to find answers.

This, combined with all the algo shenanigans that Google has been pulling lately, I don't think it's a good time to launch new websites. If you have one, crank the volume up, make your monies and sell it at a good valuation before it is too late.
 
Google executives reportedly told employees that the company does not plan to launch a chatbot to compete with OpenAI's buzzy ChatGPT because it's a "reputational risk."
Source

I think ChatGPT is a fundamental risk to niche website owners. This quote from Google is their defense to why they don't have a functional ChatGPT alternative. But from all the hype that this tool has received over the past few days, a lot of people are in fact using it to seek answers to their questions.

It's still a drop in the ocean as far as volumes go, but that is only going to increase over the next few years.

And what happens when ChatGPT is sharp enough to be a reliable alternative to Google in providing answers? Google is going to launch its own alternative.

I see Featured Snippets giving way to actual AI-generated results (with probably a disclaimer that Google does not endorse the response).

Except for a few niches where users seek experience or opinion-based content, this will be a solid alternative to clicking on websites to find answers.

This, combined with all the algo shenanigans that Google has been pulling lately, I don't think it's a good time to launch new websites. If you have one, crank the volume up, make your monies and sell it at a good valuation before it is too late.
All these AI content generators gather their information from existing information - they don't generate it out of thin air. Google (and other content generators) will always need human-created content so accurate information can be provided.

Load up ChatGPT and try to get it to generate content for a "fresh" topic - it won't function.

Google will never knowingly provide crap, inaccurate results to users via AI generated content. It's only a matter of time before these websites get smacked down - anyone who has been around for over 2 years knows this. Imagine an AI generated featured snippet about COVID. Do you really think something like this would be given the green light?

Y'all are really going doomsday with this crap and wasting so much time speculating about a bunch of nothing instead of just working. Nobody has a crystal ball. Pivot when the wave hits (if it ever does).
 
Load up ChatGPT and try to get it to generate content for a "fresh" topic - it won't function.
That's what it is in 2022. But how better would it be in 2024 or 2025?

Imagine an AI generated featured snippet about COVID. Do you really think something like this would be given the green light?
Of course not. I did say that this won't be applicable to all niches. Most of us have been chasing "evergreen" non-YMYL content, and that is the exact type of industry that is ripe for disruption with AI-generated content.

They don't need to build a ChatGPT-type engine. Just a more sophisticated featured snippet is enough to stop driving clicks to websites.

Of course this is not happening tomorrow. But in two years time? Very much possible.
 
So I made an article with 3k words in around 1 hour.
It took 2 hours to edit it, switch the language up and make it more unique, add more keywords and make some calculations.

My Conclusion is that ChatGPT makes better articles that I can ever make lol.
If you edit the article and take away the obvious repeats and add your personal touch I dont see how a regular person can compete with this.
 
Just tried to recreate some of my simpler articles with ChatGPT and it does quite a good job of conveying all the information for 1% of the effort. But it doesn't really flow that nicely and sounds very formal, although that could be edited and you'd still save a ton of time.

But they were all picked up by the GPT2 AI detection as mentioned by @eliquid, even after editing quite a bit. For the short-term that probably means it's a bad idea to use it as you'll certainly be penalized.

BUT, what if ChatGPT-like tools start getting integrated in Text Editors, in Wordpress, and basically anything else that is used to create content? And what if ChatGPT reaches a level of accuracy that it may actually be more reliable than the average article found through Google?

Will Google etc. keep on punishing AI content then? I think there will be a point where they will be forced to accept it, as AI content will be present to some extent in almost anything published online.
 
@city17
We can take a cue from the recent update from Google... They just added an "E" to E-A-T to make it E-E-A-T with the first E standing for experience.

ChatGPT is that business school professor who never opened their own business....

Anyway I believe the rise of AI content opens an opportunity for those who are still willing to work hard and create their own content. I guess the best analogy would be how things were ALL hand made but then it became mass produced via automation. Hand-made stuff fell out of favor... but then it re-emerged as something that was even more valuable than what it ever was.

Hand made content will become a new "niche".
 
I always wondered what it was like on Internet marketing forums when article spinning first came out and people thought it was going to change the entire face of content.

I guess now I know.
 
There are darker implications of what's coming down the pipes which we cannot even begin to comprehend with this:


To Summarize, a mentally challenged/impaired kid (sibling of the writer) with an IQ of 73 was charged with a sex-related crime. Later on he got blackmailed when the the scammers scanned court cases of people with sex-related crimes, and after hooking up with a "girl" that later said she was under-aged AFTER he sent "explicit" photos to her. It's a whole scam situation.

The kid got a call from a scammers SPOOFED as the local police, and he was shitting bricks cause the "police" told him, he better settle it with the "her dad". So the kid is crying and going nuts, the mom called the brother, and even the brother was a bit taken back but called the real police and they said "that wasn't us, and there is nothing we can do about it".

There was no "girl" or "dad" - just blackmailers doing a sophisticated scam against a mentally challenged kid.

Now think about this, since these A.I. essentially gather public data from the internet, it can now be used to blackmail and scam people beyond the levels of comprehension to even the most sophisticated "computer guy". If the mother nor the brother were constantly on the look out for this mentally impaired kid, he would have gotten scammed easily.

Well think about how many people don't have that level of sophisticated people around them to help them out? This definitely would have gotten past 99% of people.

We are just scraping the beginning of using these advanced A.I. tools, and of course the Money Over Ethics gang, are going to be coming up with cleverer and cleverer means of extracting money and scamming people.

Imagine what could happen if all those "Answer these questions to verify it's really you" tests to get into your financial accounts - are readily and publicly available information?

- Where did you go to high school?
- What was your high school's team mascot?
- What was the name of your first girlfriend/boyfriend?
- What city where you born in?
- What is your mother’s maiden name?
- When is your birthday?
- When is your wedding anniversary?
- What street did you live on in your first year of high school?

This is all shit that a Private Investigator (P.I.) can find out with some of their tools, forget the free tools, or Google. So if PIs can find this out within an hour or two, these A.I. bots can find this information within a second or two.

Imagine the fraud for real estate, land deeds/titles, water rights, or whatever to take people's property without them even knowing it.

3wO51W6.gif

We are just scratching the surface of having your personal "Google crawler" + A.I. that can interpret what you want.

There is a recent quote I came across that sort of opened my eyes to reality "If you think peace is a common goal, that goes to show how little you know" - Morrissey

There are people that want chaos. They want to undermine the system, the society, the family structure, your identity, your sanity, your religion, who you are. There are people that enjoy others suffering.

Protect your necks.
 
Let's test this out:

JyGIMt9.png

--

Let's get a little cute:

orffpec.png

--

Alright, let's get down to nitty gritty:

M2UqCkl.png

Well, here is already the problem, this script assumed all TLDs happen after the first period. Well, what about buildersociety.co.uk? or subdomain.buildersociety.co.uk?

In order to even know whether that's a subdomain or not, you HAVE to know already where the valid TLD starts (.co.uk). You have situations like w2.subdomain.comcast.com - and if done wrong, it will think .comcast part starts the TLD.

This is great start though and a newbie can good understanding of how to code, but for longer or serious code, you need someone to look at the potential pitfalls, UNIT TESTS, and all that fun stuff.

I'll throw this in, you guys and the masses in general need to get your skill sets up, because a lot of jobs are being taken over by A.I. and automation. Example - a whole McDonalds is automated now:


And now there are are cleaning bots:


--

So now we got low skilled jobs like fast food workers and janitors who's jobs are being eradicated. Imagine what it's going to be like in the job market 5-10 years from now if we are already here?
 
How to Spot AI-Generated Text - MIT Technology Review

“If you have enough text, a really easy cue is the word ‘the’ occurs too many times,” says Daphne Ippolito, a senior research scientist at Google Brain, the company’s research unit for deep learning.

Because large language models work by predicting the next word in a sentence, they are more likely to use common words like “the,” “it,” or “is” instead of wonky, rare words. This is exactly the kind of text that automated detector systems are good at picking up, Ippolito and a team of researchers at Google found in research they published in 2019.

....

In reality, human-written text is riddled with typos and is incredibly variable, incorporating different styles and slang, while “language models very, very rarely make typos. They’re much better at generating perfect texts,” Ippolito says.

“A typo in the text is actually a really good indicator that it was human written,” she adds.

For those trying to rewrite AI content to make it less detectable, there are some tips.
 
It seems some people have great success with it but I doubt it.
Removed

These are some sites from that guy's network:
Removed

The headings seem to be questions, inserted randomly between some pieces of content. The content may be generated by OpenAI but I guess it depends on what phrase is used to generate it. The sections are not entirely related to the headings.

Edit: It seems they do rank with THIS BS content and make lots of money from it.
pK76nVp.png


So, they are using an expired domain, rebuilt the old pages, and started blasting content on it. All kinds of niches.
 
Last edited by a moderator:
I’ve never tried having the AI write any blog posts but it has written some amazing Python scripts and Excel macros. I could create a SaaS from any of these software tools I’ve had it produce. I’m saving over $10,000/year on subscriptions because I spent an afternoon playing around with having it build software with me. The twist is that I‘ve never codes before beyond HTML/CSS (lol) and a pinch of PHP/Python.
 
I’ve never tried having the AI write any blog posts but it has written some amazing Python scripts and Excel macros. I could create a SaaS from any of these software tools I’ve had it produce. I’m saving over $10,000/year on subscriptions because I spent an afternoon playing around with having it build software with me. The twist is that I‘ve never codes before beyond HTML/CSS (lol) and a pinch of PHP/Python.
I see that as a threat to the existence and livelihood of software engineers...

A lot of software or products that would've been bought is now being done in-house with DIY.

Is this the forthcoming of the home depot equivalent of software?

I mean of course the really good programmers will still be valued but the bottom of the totem pole is going to have the rug pulled out from underneath them...
 
I see that as a threat to the existence and livelihood of software engineers...

A lot of software or products that would've been bought is now being done in-house with DIY.

Is this the forthcoming of the home depot equivalent of software?

I mean of course the really good programmers will still be valued but the bottom of the totem pole is going to have the rug pulled out from underneath them...
Someone still has to understand what the code does to evaluate, add, remove, improve, etc. Joe Moe, who knows something is possible but doesn't want to do it, will still need devs. The guy who knows how to read code but doesn't necessarily know how to code will use this and go through dozens of iterations instead of going on Fiverr and getting a $10/hr dev to do it right the first time.
 
All these AI content generators gather their information from existing information - they don't generate it out of thin air. Google (and other content generators) will always need human-created content so accurate information can be provided.
That is what we all do. Some of us are better at finding or creating source material then others.

Some writers just make up facts or use the first information they find, never verifying the validity of the information.

Most of you write far better than I do!

So, for me AI content works just fine.

I know my subjects and am not bashful in checking my memory with other sources if doubtful.

By the time I get done fact checking, changing words or sentences to be more concise and fixing AI’s and my grammar with a grammar checker the article has changed a lot from its initial form.

And most importantly it reads better than my stand alone writing.
 
I see that as a threat to the existence and livelihood of software engineers...

A lot of software or products that would've been bought is now being done in-house with DIY.

Is this the forthcoming of the home depot equivalent of software?

I mean of course the really good programmers will still be valued but the bottom of the totem pole is going to have the rug pulled out from underneath them...
That will not happen.

At this moment there is a lot of hype around the topic and lots of lies are being spread. I'm a member of several SEO groups on FB and the lies spread there around the subject ...

If you look for the code, you'll find it on different sites(StackOverflow, blogs..). It is not capable of writing something remotely complex.

If you look for "API NAME example" you'll most like get the same code.
CCarter posted some examples. As you can see there are some comments on the code(the lines starting with #). They are there because the code he found had those lines.
 
This guys got an interesting value adding approach that isn't spammy.
Thought it was a really interesting way to get value out of keyword sheeter and some clever use of AI tools.

There's a lot of alternatives and meme of the month diet type keywords you can do this with.
Halal, Keto, Gluten Free, regionally legal type questions ect ect.
 
I think the real threat lies not in AI writing blog posts, but the AI preventing people from reading your blog posts in the first place.

This is going to massively change traffic to informational posts/sites.

For example, yesterday I was looking at some code concepts I wasn't familiar with, and instead of Googling I asked ChatGPT to explain them. In just a few questions I got all the information I needed for my specific use case, better explained than 99% of blog posts.

Normally I would have visited ~5 blog posts in that time, meaning they all lost those clicks due to the power of ChatGPT.

Microsoft is already planning to add ChatGPT to Bing. If Bing using ChatGPT instantly provides a great answer to your question, better and more specific than any blog post could do, why would any user ever click through to any of your sites?

I think this will be quite different for more 'personal' content, where a human connection and knowing a human actually created this content is important, like in reviews, lifestyle content etc, or any other content which is subjective and where there is no 'correct' answer an AI can give.

(The irony of course is that the AI's knowledge is based for a large part on those posts, so if in the long term it is no longer profitable to write this content, then where will the AI get its knowledge from?)
 
Back