I recently started using ChatGPT, I read several books and found that much of the information on the subject was substandard. So I have put together a ‘lessons learned’ blog. I have created nothing new, only reviewing and consolidating information that others have provided. Think of this as a collection of key words and ideas from the listed resources. Also consider this to be a true starting point, and not the source for all answers. Many links have been provided to sources of the REAL started points. I believe that I have provided an introduction up to the point of “fine tuning”. For further introduction, use sources like:
https://platform.openai.com/docs/guides/gpt-best-practices/strategy-give-gpts-time-to-think
and now for my journey:
In the evolution of human – machine communications, I participated using:
on/off switches,
variable resistors,
punched cards,
magnetic tape,
optical disk,
teletype,
typewriter,
mouse,
joy stick,
touch screen,
speech recognition,
text recognition,
gesture recognition,
as well as numerous computer languages
and now …. normal conversation in English. I love it …
AND the hope of having a tool for truly searching a data base, of my choosing or creation, perhaps using EXCEL or buying data from amazon. Put on my lamp/hat and we’ll do some REAL “data mining”!
How do we get there from here – that is the question for now.
How do we communicate? Sample Sources of human thought generation:
https://en.wikipedia.org/wiki/Metaphor
https://en.wikipedia.org/wiki/Analogy
https://en.wikipedia.org/wiki/Parable
How does ChatGPT communicate? Sample Sources of ChatGPT communications:
https://en.wikipedia.org/wiki/OpenAI
https://en.wikipedia.org/wiki/ChatGPT
https://en.wikipedia.org/wiki/Meta_Platforms
https://en.wikipedia.org/wiki/Google_Assistant
How do we communicate with ChatGPT? Information sources:
What is ChatGPT doing … Stephen Wolfram.
https://en.wikipedia.org/wiki/Stephen_Wolfram
https://writings.stephenwolfram.com/2023/02/what-is chatgpt doing-and-why-does-it-work
A Mathematical genius of our era, Stephen Wolfram, is excited about combining ChatGPT and his Math Language. AND will tell you how to do it!
There are MANY tutorial videos on YouTube, for all ages, strengths, and likeness. Below is a sample:
Supercharge your research … Ruopeng An – a professional researcher, that has a staff of bright graduate students, exploring ChatGPT for scientific publication. He has beautifully and lovingly collected approaches to using ChatGPT that work! You can’t get a better introduction to: ‘how to communicate with a bot’ then here.
A REAL reference book:
https://platform.openai.com/docs/introduction/overview
We must learn to talk with an intelligent being …
PROMPT ENGINEERING – a new discipline of the utmost importance
Crafting Effective Conversation starting points:
Navigating to reduce limitations and use capabilities
Items related to INPUT to ChatGPT – be thoughtful and …
(NOTE: the numbers refer to items in Dr. An’s book and on-line class provided above)
1 – be specific, clear, precise, detail, no ambiguity or jargon
“Speak clearly; Don’t Mumble”
2 – use a simple series of related parts and exchanges
chain prompting – open ended questions, directed & building discourse
3 – explore token syntax, word order, vector embedding
4 – provide context definition; specific field, audience, cooperation …
lets contemplate this methodically, you are …
8 – establish constraints – format accepted
9 – provide example communications – analogies, generalization, few-shot, act as
Items related to OUTPUT from ChatGPT – “don’t be shy”:
5 – request step-by-step explanations
6 – demand sources and citations – grounding in verifiable facts
7 – invite alternatives with pros/cons, for/against
8 – honor constraints – length, style, format, interests, goals, audience
9 – provide example communications – analogies, generalization, few-shot
10 – explore multidiscipline perspectives
Ask for :
evidence based,
step-by-step,
sources & citations,
arguments & counter arguments,
for & against,
in favor or not,
brief explanation,
examples,
the question is …,
… and more
Examples of Prompts requesting Explanations:
Describe the step-by-step process of …
Outline the key stages of …
Explain the steps involved in …
Suggest 10 inquiry approaches and relate to …
Examples of Prompts requesting Sources & Citations:
Provide scientific relevant citations to support & backup …
Challenge soft statements – the quality of the evidence
Be polite to preserve our courtesy & dignity, but don’t be shy
Examples of Prompts requesting Alternatives:
Provide pros/cons & for/against arguments and counter arguments
Encourage critical thinking and exploration
What are the advantages and disadvantages of …
Investigate diverse insights & perspective
Encourage genuine conversation
Examples of Prompts requesting Report constraints:
Control length – style – format – language …
Provide a 200-word summary
Provide a 5 bullet point explanation
Provide a 5 column table
List 5 main factors contributing to …
Write a computer program in Python which will …
Examples of Prompts requesting Assistance:
List key risk factors for developing …
List steps to conduct a meta-analysis of …
What are potential interdisciplinary approach …
Explore how advances in ___ could potentially revolutionize …
Consider ethical implications of …
Using and Encouraging Examples in Communications
EXAMPLES of EXAMPLE forms
Specimen,
Sample,
Instance,
Case,
Representation,
Illustration,
For instance,
Such as,
That is,
In particular,
Like – likewise,
As in – as – as if,
In other words,
Especially,
As a case in point,
That is to say,
Namely,
To illustrate,
Let’s say,
Also,
This includes,
Note well,
Additionally – in addition,
Can be seen,
In a similar case,
Pretend that – imagine,
To be specific,
As proof,
In the manner of,
_ is to _ as _ is to _,
Comparable,
Like _ except,
Representative,
Symbolic,
Remember,
Consider,
For the sake of example,
Let’s think about,
Act as …
We can almost conclude that all meaningful communications can be accomplished using examples
Starting with the primal metaphor. What was the referral to, in the first analogy – hey, that was just like …
Semantic vs syntactic … what is the difference?
Logic vs rules of grammar
Model vs words
Truth vs fiction
https://en.wikipedia.org/wiki/Semantics
https://en.wikipedia.org/wiki/Syntax
Numbers are the language of the Gods – be right
Examples are the language of Robots – be careful
Words are the language of Manipulators – beware
Sample Questions that I would like answered:
Spreadsheets are to math as Chatgpt is to … (Language and Communications)
Evidence of suppression of scientific evidence of homosexual behavior in nature.
Examples of establishment involvement in selective reporting on natural behavior to align with religious teachings.
NEXT UP – Fine Tuning – this the end of the road for me. A lot of commitment and ability is required for travel beyond this point. You will benefit from knowing a language like python, to get neat stuff done! If you go on – enjoy following your bliss and remember that the answer is 42.
With specialized data and tools you can use:
https://openai.com/blog/chatgpt-plugins
The coding is at a mid-to-high level
Leave a comment