Artificial intelligence prompt engineering for writing

[Mercurian Aspirations]

I know some people are using OpenAI ChatGPT for different use cases. But what are some prompt engineering techniques geared towards writing essays, social media posts, video descriptions and articles.

 

[icenoid]

I kinda still don't get the difference between chatGPT vs Perplexity AI. Can someone explain? What I do understand though is that they have citations when they generate a response you have asked.

GPT is OpenAI's LLM, the brain that accepts text and output text. ChatGPT is the user-facing interface that allows users to chat with GPT. Perplexity AI also provides a user-facing mobile app and web app. The backend is built on top of GPT, using GPT to translate the user's query and summarize Bing's search results and the citation. ChatGPT is more of a generalist. Perplexity AI is a specific use case that utilizes GPT to return the user the best possible answer to the user's question.

 

[Only Chapter]

I usually put a custom instruction (for ChatGPT) that requests the AI to include its level of confidence, sources, and date of an answer, and also instruct the artificial intelligence to specifically tell me if the answer is speculative. And make it a markdown table at the end of every answer. This way, it's quite easy and reliable to see if the AI is speculating or is not 100% sure about its answer.

 

[Z-Mobile]

ChatGPT is very useful in the marketing field as well. You can perform content-related tasks within no time. But the ugly side is its content misses depth.

 

[Odd Soft]

Is it me or is prompt engineering the new search engine optimization (SEO)? It will be hot for a while but it's a transient thing that will get washed out as tech gets better. You're better off working on the models yourself or in ML/LLMOps.

 

[Conscious Log]

Forgive me if I am wrong, but I feel like the tech industry is watering down the significance of what it means to be an engineer. Engineering is not simply writing prompts for proprietary or open source Large Language Models. Instead of prompt engineer, it should be called prompt writer. What I am trying to say is you wouldn't call someone who can use Google search engine a Google engineer.

 

[Pink Desert]

Forgive me if I am wrong, but I feel like the tech industry is watering down the significance of what it means to be an engineer. Engineering is not simply writing prompts for proprietary or open source Large Language Models. Instead of prompt engineer, it should be called prompt writer. What I am trying to say is you wouldn't call someone who can use Google search engine a Google engineer.

Everyone is an engineer now man.

 

[Spike Mountain]

Forgive me if I am wrong, but I feel like the tech industry is watering down the significance of what it means to be an engineer. Engineering is not simply writing prompts for proprietary or open source Large Language Models. Instead of prompt engineer, it should be called prompt writer. What I am trying to say is you wouldn't call someone who can use Google search engine a Google engineer.

Prompt engineer is a term like domestic engineer or sanitation engineer.

 

[Zero Art]

This whole thing seems to be hype. Writing prompts is straight forward.

 

[Badger McBadger]

Forgive me if I am wrong, but I feel like the tech industry is watering down the significance of what it means to be an engineer. Engineering is not simply writing prompts for proprietary or open source Large Language Models. Instead of prompt engineer, it should be called prompt writer. What I am trying to say is you wouldn't call someone who can use Google search engine a Google engineer.

The only difference between coding and prompts is the language being used.

 

[Nomad 80]

Being good at writing prompts just means being able to string a sentence together with the correct bunch of qualifiers. Seems pretty simple, unless you struggle with building sentences.

 

[Fruits Power]

This whole thing seems to be hype. Writing prompts is straight forward.

That's not how writing prompts work. Just having proper grammar isn't good enough. For example, using delimiters to separate context and format responses is a trick to use. Also, knowing that writing your intent at the end of the prompt and after the context (or even restating your intent at the end) is also important because LLMs are usually more biased towards the end of the prompt. What about system prompts vs user prompts?

 

[Adam Dreaming]

This whole thing seems to be hype. Writing prompts is straight forward.

Not quite because with ChatGPT it's garbage in garbage out.

If you don't know what you're doing it can't really help you.

 

[Potential Case]

Everyone is an engineer now man.

Yeah, you just add "engineering" to every new job and you are ready to go.

 

[Livid Beautiful]

This whole thing seems to be hype. Writing prompts is straight forward.

You still have to cross-check because sometimes artificial intelligence hallucinations are like LSD visuals.

 

[Neurotic Knight]

How about creating your custom GPT? Wouldn't that be a good idea?

The marketplace of OpenAI's custom GPT Store is going to be so bloated it's probably a viable idea to create a model that suggests what the best model is to get something done. Kind of like those GitHub repos that track other repos.

 

[rav3style]

How about creating your custom GPT? Wouldn't that be a good idea?

Everything that gets built by third-party developers will be used by OpenAI to integrate it into the next version of their Large Language Model. Devs will be training the model and then be left with nothing when GPT can do it natively. It's a trap!

 

[Neurotic Knight]

Everything that gets built by third-party developers will be used by OpenAI to integrate it into the next version of their Large Language Model. Devs will be training the model and then be left with nothing when GPT can do it natively. It's a trap!

True.

 

[Potential Case]

Everything that gets built by third-party developers will be used by OpenAI to integrate it into the next version of their Large Language Model. Devs will be training the model and then be left with nothing when GPT can do it natively. It's a trap!

The same could be said of every conversation you've ever had with ChatGPT, they don't even try to hide it.

 

[Vegetable Broccoli]

Everything that gets built by third-party developers will be used by OpenAI to integrate it into the next version of their Large Language Model. Devs will be training the model and then be left with nothing when GPT can do it natively. It's a trap!

Amazon style!

 

[Adam Dreaming]

Amazon style!

Are you talking about Amazon Basics? Because one has to wonder if people continue to make millions on Amazon from the point that Amazon copies their product and starts to offer it for the half price of the original?