I have been using copy.ai for a while and never tried to use GPT3 to achieve the same results as I was already happy with it.
Somehow I always thought I could easily get the same results using the OpenAI playground but to my surprised I was getting nothing even near the quality of copy.ai.
I guess what I am missing is a good prompt header that will give GPT3 swiss knife skills around copy. If anyone has tried this before or want to give it a go and share their findings this is the thread. =)
Why not try making a fine tune model using the results of copy.ai?
You can build a custom article generation model with just a single prompt!
Here is a test article I created for a coffee website by using a prompt to simulate a creative coffee writer:
Davinci already has all of the knowledge it needs to write expertly on just about ANY topic.
You need to experiment with creative prompt engineering approaches to “persuade” AI to generate articles in the style and format required.
Forget generic AI writers, custom Davinci prompts do an amazing job!
Nice! The general idea was for people to share a prompt that works as well as copy.ai
copy.ai it provides really good outputs for general cases. I am just nowhere near to generate neither the general cases or the specific ones. I do know now that you attest this is possible which is something but not what I was after.
Davinci can easily out-perform copy.ai IF you custom engineer a prompt (or train a model) for your specific use case.
You can create your own article generator which will produce articles in the format & writing style you desire - & in bulk if you need a lot of content.
Copy.ai can’t do that, and they don’t attempt to because their business model is to be a generic content generator.
You can engineer a custom prompt with a couple of hours creative experimentation.
It seems that we both agree that this is or should be possible in theory.
I still need a solution in the practical terms which for now remains to be seen.
Have you got a prompt to share that does exactly what open.ai does or better? Otherwise my initial request remains intact.
So, what you are seeking seems to be proof that a custom prompt (or trained model) can produce output to the same quality as copy.ai.
The problem here is that a custom prompt (or custom article creation model) is a totally different kettle of fish as it is customized to generate articles specific to your exact use case and will not be at all useful for other niches.
Here is an example of creating a basic prompt model (I just built this).
The prompt used here is actually generic but with a few hours work could be engineered to write as an expert on any topic and in ANY style.
You could implement a prompt based model for creating articles in bulk if you so desired - in your style - writing as an expert on your chosen topic. This would prob involve a day or 2 of experimentation to get it right - and then you have your own writer.
It is ironic that the article recommends training a model but the article itself was generated by a single prompt 
Prompt & output:
I need an expert to demonstrate how GPT3 can be used to write an article at least as well as Copy.ai.
GPT3 can be used to write articles in a similar way to Copy.ai. However, there are some key differences between the two tools. For example, GPT3 is designed to be used with a wider range of data sources, including text, audio, and video. Additionally, GPT3 offers more customization options, allowing users to fine-tune the output to their specific needs.
What information do you need from me to generate an example article?
To generate an example article using GPT3, you will need to provide:
The topic of the article
The length of the article
Any desired keywords or topics to focus on
Once you have provided this information, GPT3 will use its artificial intelligence algorithms to generate a FULL & COMPREHENSIVE informative article that meets your specifications.
Ok, here you go:
Topic: “Generic AI Content Writers Versus Custom GPT3 Models for Article Generation”
Length: 1000 words MINIMUM
Focus topic: “Don’t bother with generic AI article writers - everyone is using them. Build your own on GPT3!”
First create an 11 paragraph article outline:
Introduction
In recent years, there has been a growing trend of businesses and individuals using artificial intelligence (AI) to generate content. This trend is only set to continue, as the technology behind AI content generation improves and becomes more accessible.
There are two main types of AI content generators: generic AI content writers and custom GPT3 models. Generic AI content writers are tools that use pre-trained algorithms to generate articles on a range of topics. Custom GPT3 models, on the other hand, are specifically designed and trained to write on a particular topic or in a specific style.
In this article, we will compare the two types of AI content generators, looking at the benefits and drawbacks of each. We will also explore why you should use custom GPT3 models for your content generation needs.
What are AI Content Writers?
AI content writers are tools that use artificial intelligence algorithms to generate articles on a range of topics. These tools usually have a database of pre-written sentences and paragraphs that they can rearrange to create new articles.
The quality of the articles produced by AI content writers can vary widely, depending on the quality of the algorithms used and the size of the database of pre-written sentences and paragraphs. However, in general, AI content writers are not yet able to produce articles that match the quality of those written by humans. This is because they lack the ability to understand context and produce original ideas.
What are Custom GPT3 Models?
Custom GPT3 models are artificial intelligence models that are specifically designed and trained to write on a particular topic or in a specific style. Unlike generic AI content writers, which use pre-trained algorithms, custom GPT3 models are created from scratch using a training dataset that is tailored to the desired topic or style.
This means that custom GPT3 models can be much more accurate and produce higher-quality articles than generic AI content writers. Additionally, custom GPT3 models can be trained to write in any style, including first person, second person, or third person; in any tense; and with any level of formality. This makes them much more versatile than generic AI content writers, which are often limited to writing in a single style or tense.
The Benefits of Custom GPT3 Models
There are several benefits to using custom GPT3 models for your content generation needs:
The Drawbacks of AI Content Writers
There are also some drawbacks to using AI content writers:
The Drawbacks of Custom GPT3 Models
There are also some drawbacks to using custom GPT3 models:
Why You Should Use Custom GPT3 Models
Despite the drawbacks, custom GPT3 models offer a number of advantages over generic AI content writers. These advantages include higher quality, accuracy, and versatility. Additionally, custom GPT3 models can be cheaper in the long run, making them a more cost-effective option for businesses and individuals who need high-quality content on a regular basis.
How to Create a Custom GPT3 Model
If you decide to use a custom GPT3 model for your content generation needs, there are a few things you need to do in order to create one:
Conclusion
In conclusion, custom GPT3 models offer a number of advantages over generic AI content writers. These advantages include higher quality, accuracy, and versatility. Additionally, custom GPT3 models can be cheaper in the long run, making them a more cost-effective option for businesses and individuals who need high-quality content on a regular basis.
Thanks Jeff for your time and further explanations. This is indeed a great example for blogpost generation.
Somehow you are still missing the point which was to achieve copy.ai results or better. As creating whole blogposts is not currently a feature of copy.ai that still remains an open exercise. You don’t seem to be familiar with copy.ai features.
Before you reply back with another “how to generate a whole blogpost example” let me give it another try using the information you provided already.
Thanks again for your patience and willingness to help.
Hi Gerard,
I assumed copy.ai was another of the generic blog generators.
I just checked their website and all of those features can be accomplished using a single prompt, and it’s a much easier job than a custom article generator.
In fact, with a few hours work you could do a much better job than all of the generic copy.ai templates.
I recently created a prompt for a client that responds to his emails in his own personal style. It took approx 16 hours to build / test the prompt and it has approx 95% success rate (he only needs to press re-generate 1 in 20 emails).
You can 100% easily replace all of those copy.ai templates with a single custom GPT3 prompt and achieve much better unique copy!
Good luck.
Jeff
That’s amazing! Looking forward to take my prompt skills to your level.
I’m just a novice prompt engineer 
Gerard.
Hi Gerard,
GPT3 already knows just about everything.
You simply need to tease the knowledge out of it!
It already knows all of the copy.ai template structures (or similar).
If you play with GPT3 creatively and get a “feel” for it I really cant think of many things you will be unable to accomplish.
Ignore anything you read from before 6 mths ago - it’s mostly irrelevant when playing with prompts for the latest large language models.
The new models seem to behave totally differently to all of the others!
I am mostly using Davinci 2 (Temperature 0) to interactively create prompts.
Once built you can decide which parts of your prompt to designate as variables to be passed to the API.
At run time I use Davinci 2 (play with Temp/freq/presence) or legacy Davinci (seems more creative).
When you have something working on Davinci, you can simply test the same prompt on smaller/cheaper models to see if you still get the output you need.
20% of the time you will.
I put together a prompt that gave me some good results. This is the final version which I tried in for a few different scenarios.
You are an excellent copy editor. Your copy has won many awards and accolades. You don't require a lot of direction and will make the right decisions given the information provided.
++This is the text you receive as input:
<<INPUT>>
++Copy direction:
Instructions: <<INSTRUCTIONS>>
Tone: << Friendly | Luxury | Relaxed | Professional | Bold | Adventurous | Witty | Persuasive | Empathetic>>
++This is your proposal:
Thanks for any feedback or improvements you foresee.
Gerard.
Nice job!
If you want it to write in YOUR style, simply include 3 full examples below your prompt.
Then use the WHOLE prompt inc examples + inputs for new copy required.
You can tune that until you get outputs that look like they were written by you.
Another option is to create a prompt that receives your raw incoming communications as input & auto creates the reply.
You can combine that approach with a vector database containing your knowledge base if your communications are related to business. I am working on a test for the vector database idea at the moment.
Thanks!
I’ll try that. Style transfer using prompts!
What do you mean by vectors? Do you mean collection/list of sample communications in different contexts?
Good luck with that experiment!
Gerard.
Yes, style transfer with a “few shot prompt” works very well.
Ref vector databases, it is a newish approach to storing info - which can be searched semantically and the search results used to generate AI content.
I am currently beginning testing this approach for auto generating a specific subset of business email responses by referring to the vectorized business knowledge base.
This is quite an interesting video exploring using Open AI with Pinecone vector database: Beyond Semantic Search with OpenAI and Pinecone - YouTube
Oh I see. Got it!
That seems a lot of work to get around the limitations around openAI API restrictions.
Wouldn’t be easier if openAI increased the working memory for the models?
Gerard.
Vector databases facilitate 200ms semantic search on a trillion documents.
This opens up possibilities that would not be possible with an expanded working memory.
The improved functionality of Davinci 2 combined with the 4k token limit does enable you to do a LOT that was previously impossible though, especially for your copy.ai project which can be implemented using custom few shot prompts.
If we had access to the model this wouldn’t be necessary so it’s bound to the restrictions imposed by openAI API one way or another.
Nevertheless I agree on your statement although it’s not a very common use case to have trillions of documents at your disposal it’s enabled by a vector database.
Let’s make the best of the current situation as it lasts.
If you can share 2 or 3 copy ai results and the input that is provided in their tool to kick off their generation, it would be fun to try to engineer a prompt that is competitive with that…
Done that already. Here: