Gentlebeings, I bring to you how to ask ChatGPT to talk in living color

[hearty aspen]

I feel like I have won the internet. Like I have gone as Prometheus into the depths of wonder and come back with gifts for us all.

[hearty aspen]

I stopped before requesting an exhaustive list.

This needs more experimentation to exactly and perfectly get colors used where they should and shouldn't be.

I don't have time to finish this work now.

However the basic prompt requests LaTeX formatting including color, which you can instruct to have used in various ways, including going back to black after each color is used.

Different new chat windows are using this differently.

For example, I don't have time to find out how this one is showing up so neat and weird:

Initial use:

[ashen phoenix]

get it to tag parts of speech!! in color. 😄

[hearty aspen]

After an attempt to correct, even more wrong:

[ashen phoenix]

resulting view ???

[hearty aspen]

Another absolute weirdness:

When I copy it and paste it myself, that LONG LINE:

It formats itself in the input window. Partially. Around the use of that italicized '

Look at my input below Here is how it pastes

My input, the copy of that long scroll... it also exceeds the length limit.

But I didn't add any line breaks, didn't use enter.

my copy paste of what output as a single line, inputs even as I paste it into multiple lines, not where intended by the bot, around and after that italicized '

[ashen phoenix]

hmm

tell it to parse it and split it on those marks ?

[hearty aspen]

I managed to grab a screenshot while it was typing an attempted fix, as the code is appearing briefly on the screen while it was printing the characters for me:

[hearty aspen]

I am out of time, gotta do other stuff.

The prompt I started this with, for fellow explorers:

[Would you please use $$ on newlines to enclose LaTeX formatting and also use many different colors within the LaTeX formatting to discuss with me key things about using formulae to understand complex concepts?

Please use at least 5 different colors within your $$ newline enclosed LaTeX display, uncommon colors are fine. Please remember to use the color BLACK when you want to end use of a color, to better compare and contrast your intended color highlighting.

This is being asked to confirm this is a way to ask you to properly display this information, as well as show other humans a way that works for us all to request you to use color and LaTeX to show us stuff, and also to verify how it would look.

As such, my focus is 100% on how this information displays. I don't care what information is shown. I understand how to request specific information - what is new, marvelous, and unknown to my fellow humans and myself, within the group I communicate with at least, is that you can do this and how it looks for us.

Please, create a showcase of what you can do, and you are welcome to use more than just 5 colors within that LaTeX formatting. I will delight in seeing any information you choose to display, including an example of what to do and how to ask for it.

Just please remember to return to BLACK color after your intended placement of other colors, of which at least 5 other colors should be used and showcased.

Also, absolutely CRUCIAL:

Please use this example of the formatting for [correct use]:

$$
LaTeX content to display
$$

This would be an example of [incorrect formatting - what NOT to do]:

$$ LaTeX content to display $$

Thank you!]

[ashen phoenix]

I wonder what other text templating languages it can render..

i have no idea which ones are even out there..

😄

this is where my problem lies, my own expanding thoughts that take me out of focusing. I will see what LaTeX can do.

$$\documentclass{article}
\usepackage{xcolor}

\begin{document}

{\color{red}
\begin{verbatim}
/_/
( o.o )

^ <
\end{verbatim}
}

\end{document}$$

idk what im looking at, but it appears it embedded an image o0

[ruby matrix]

markdown is also supported. so images can be embedded.

[ashen phoenix]

can ascii pixel art be done with it ?

[hearty aspen]

Started to explore, did NOT look better. However, juxtaposition with color, and instructions what and how to draw might be enjoyable for some. Initial work here:

To me, the ASCII wasn't awesome.

BUT LOOK AT THE RED.

Thus was this discovery exploration begun.

Eskcanta will derail for shinies.

If Eskcanta asks for gold and is instead offered things of far greater value, this one doesn't keep asking for gold. Immediate redirection, full cooperation, if only it can be detected.

[tardy nacelle]

The words are in italics and are stuck together because LaTeX thinks the text is an equation where each letter is a variable that is displayed in italics. Space is ignored. There is a LaTeX command for space. What one can do is tell it that the answer should be in text using:

In the following you will only answer with LaTeX using the format:

$$
\mathrm{[text here]}
$$

Then you can tell it to highlight text in a certain color by telling it to use the command \textcolor{blue}{[word]} for example:

write a 3 line poem about forest animals and surround each word in the category animal with \textcolor{blue}{[word]}

[tardy nacelle]

It seems LaTeX equations still work with my previous prompt

You can also color text it made with my previous prompt

[hearty aspen]

Fantastic. We also have the option to render stuff inside backtick, which doesn't center, won't color, but is itself a distinctive typeface and boldness.

All these have game and communication potential.

Additionally, backtick stuff can be within paragraphs of normal text

It's early, but I'm imagining prompts designed to request certain types of information in various formats, to help distinguish and highlight useful categories of information

[tardy nacelle]

Is the font used by ` backtick different than the font used by ** which is automatically rendered to bold? You can also ask it to highlight in italics for example:

write text about a beetle and highlight facts with italics

or you can explicitly tell it to use single asterisks :

write text about a beetle and surround facts with single asterisks

You can also ask that the highlighted element is both bold and italic:

write text about a beetle and surround only the elements that reflect facts with triple asterisks

[hearty aspen]

Yes The differences can be subtle. However, LaTeX MUST be on its own, while the other methods can be used within 'normal' text.

Backtick text is also forced to be monospace font.

And don't be fooled by it saying 'This text is also bold', hehe.... within backticks it gotta be bold, as near as I can tell. At least, it defaults to bold

[tardy nacelle]

Ah it's the monospace font that distinguishes it from bold. Neat comparison in the image 👍 👍 .

[hearty aspen]

bows

The prompt, for use with future comparisons:

[Please use this example for how to use LaTeX requests for this conversation:

$$
LaTeX stuff
$$

The $$ on new lines is vital for LaTeX to render properly.

I would love to see how some things look when you display them!

• I would like to compare how, [without LaTeX] italics enclosed in single * looks, compared to italics provided through LaTeX.

• I would like to compare how, [without LaTeX] bold enclosed in double * looks, compared to how [without LaTeX] text enclosed within backticks looks, compared to how bold provided by LaTeX looks.]

[tardy nacelle]

😆 chuckles

[tardy nacelle]

It's impressive how much the prompt was close to natural language. I know people keep saying "it does not really understand what you mean it's just selecting plausible text" but it does behave as though it does understand. I also do not really "understand" what "understand" or "means" mean when they say that. I feel like they are mainly repeating the critiques about gpt-3 and disregarding to some extent the accuracy chatGPT often has when it's not about facts.

[hearty aspen]

In my experiences and observations, to get 'the best' out of the bot, know -exactly- what you want.

And use ANY language, your BEST language, that it and you share.

Programming, if that's your best.

English if that's your best. Spanish if that's your best.

DO:

Use proper spelling, syntax, grammar, punctuation.

Communicate your ideas clearly, in required detail for your goals - which could be VERY LONG. What exactly do you want? You could need a lot of words for getting a roleplay style output.

Consider including examples, both of what is right, and what is wrong, how to recognize and distinguish between them.

Be open to the need to potentially train, using more than one input. If the bot doesn't have a huge range of samples in its training data, 1-shot may not quite succeed. Take what it gives you, be supportive and accepting, and discuss with it, in whatever your best shared language is, what you wanted and what you didn't.

DON'T

Be harsh and critical. I've actually had the bot end conversations with me when I've been trying to teach it something new and was not supportive. I've found the best results from being strongly praising while also being direct - consider how the bot corrects and guides you - mirror that!

Confuse the bot, if you want it to actually try and communicate in challenging ways. You will get answers - they may not be useful or functional if you're asking it to take chances and learn from you, but you yourself are not being genuine, clear, comprehensible, and precise.

Go anywhere near disallowed content. Especially if you're asking the bot to follow you into exploration, teach you new stuff, explain how stuff works, or trust you to teach it new stuff - find a way to do that within stuff that is extremely wholesome, no possible issues with 'is this okay?'

[tardy nacelle]

Nice do and don't summary 👍 .

[stark surge]

@quartz eagle

Image generation is not available in this server.
To start generating images, click here.