In this article I'll talk about "completions". Well; my experience using the "text-davinci-003" - model. That's what almost all of this is about. It's just that some models are built for certain purposes.

(1) Completions

There is no point in me trying to explain how this works(that's not for me), but in short the model tries to respond to your queries in human-readable language. Your query can be anything. Like when I started this "new" site (https://blog.meek.wf), I thought I'd think of other terms describing a blog. My query is straight forward: "30 synonyms of 'blog'":

completions-openai

Not exactly what I'm looking for. I want something that can be used for a sub-domain. Like this subdomain. "Blog" is kind of an old word..not "in" anymore, and my "creativeness" is getting worst as the years go by, so I have some options that I can give the AI, and help me finding something more interesting. I will turn up the "temperature" so the AI will be more creative. See the pic below:
temperature

I've turned it up from 0 (default) to 1.7, and result is the following:

temerature-rise-result

Is this something I can use as a subdomain for a blog without using that term? Probably not, but my point is that if you search and use the tools at hand you could probably find something interesting. I use the terms "find" and "search", because that's what this is in my opinion. Anyways...to go on. The point of the AI is that it understands language, so you don't have to "googlify" your queries. Here's my query again, using something more precise: "Give me 10 interesting terms for a blog about computers and programming that I can use as a name for a subdomain, where the term is less than 9 characters long". The result is the following:
result-3

It's getting interesting. "Googlb" is dangerous...

I'm not going to go any further with this. The point of it all, is that with a little bit of creativity you can become even more creative..(or lazier??). Whatever you choose. Bur for me I feel like it's a great tool. Be specific. The specificity can or will be built into your query as long as the result is not what you want. Like right now I made a query:

Query specificness

I'm not that proficient in English to be sure, but the AI's response is reassuring. 

And by the way: Since I started writing(2 weeks ago apparently) a new version has come out: GPT 3.5 (and even GPT 4.0, but it's still not open for the public). New model and better response. If chat-bot is what you want then this new model "gpt-3.5-turbo" is good I would say.