Forum Replies Created
-
Posts
-
Hi Spllat,
The Prompt is excellent however in one place you say 8 yr old and in another 10 yr old, this may confuse the AI.The age of the child and book title / art piece could be encapsulated in {} making it easier for humans to mod the prompt. The AI can handle that.
I would recommend the book: ‘The Dog Crusoe and His Master’ by R. M. Ballantyne. I remember it vividly and still think about it from my childhood of 60 years ago.
Beyond that, it is just a suggestion, you could explore ‘personal rubric’ in an AI/Chat context. For example a response in this context could be:
“AI: With empathetic guidance and understanding, I provide tailored instruction to young autistic minds in language arts and creative writing. My approach focuses on fostering creativity, promoting self-expression, and empowering these children to embark on fulfilling and transformative creative journeys.”
Querying the AI further should / may generate a list like this:
Empathy and Understanding – 95,
Tailored Instruction – 90,
Fostering Creativity – 85,
Promoting Self-Expression – 85,
Knowledge of Language Arts – 90,
Expertise in Creative Writing – 90,
Effective Communication – 90,
Patience and Adaptability – 85,
Sensitivity to Individual Needs – 95,
Building Confidence and Self-esteem – 90.Note these are not based on any empirical measurement, however you can then adjust to what you perceive as the most important aspects in numerical terms and ‘play them back’ to the AI as the ‘required expertise.’
I did run this a few times but the AI does seem to have a problem with the language age matching.
My final output was:
“AI: Certainly! Here are revised versions of the questions with vocabulary more suitable for an 8-year-old:
How do you think Crusoe’s determination and never giving up, even when he was alone on the island, can inspire you to keep trying when things are difficult?
Quote: “Crusoe never gave up, even when he was all alone on the island.”In the story, Crusoe made friends with animals. Can you think of a time when you felt a special connection with an animal or pet? How did it make you feel?
Quote: “Crusoe became friends with the animals, and they helped each other.”Crusoe was kind to everyone, even if they were different. How can you show kindness to others, no matter how they might be different from you?
Quote: “Crusoe treated everyone with kindness, no matter how different they were.”Crusoe used his skills and knowledge to survive on the island. What are some things you’re really good at? How can you use those skills to help others?
Quote: “Crusoe used his cleverness and skills to find food and build shelter.”The story shows how working together with others can help us achieve goals. Can you think of a time when working as a team helped you accomplish something? How did it make you feel?
Quote: “Crusoe and his friends worked together as a team to solve problems.”Crusoe was brave and faced danger. Can you think of a time when you had to be brave? How did you feel afterwards?
Quote: “Crusoe faced his fears and stood up against danger.”Crusoe appreciated the simple things on the island. What are some things in your life that you’re grateful for? Why are they important to you?
Quote: “Crusoe appreciated the simple things, like having enough food and a safe place to stay.”These revised questions use simpler language and are more age-appropriate for an 8-year-old student.”
I’d be happy to run this more to perfect it, if you so desire or think it may be of any help / benefit to you.
Regards
DesHi Brian,
thanks for that..
I’ve almost finished my own 20,000+ word Prompt Manual, I suppose x hundreds hours work, though not really work as I find it meshes nicely with 40 years writing documents / reports for clients plus a large body of creative writing, mostly poetry, (on the slow work days). I do find that one can find a ‘really helpful’ ver in ChatGPT with better manners and fuller / more comprehensive answers – an AI enigma one could posit.I’m approaching it from a pure semantic and syntactic basis. It is a most interesting intellectual challenge to elicit facts / offerings from the AI beyond what can be discovered in ‘casual’ conversation. I suppose one would say this is a soft form of interrogation where tone / intent / empathy / nuance / role / the rubric and Boolean yield interesting results.
In Ireland this is something in everyday use where no one will ask you a direct question yet at the same time try to find out everything about you.Being such a subjective thing it is very difficult to judge one’s one rating, I’m not in the sub 5 perhaps so somewhere in the 5-10 range, hubris notwithstanding.
Regards
DesHi Linnea,
this is a fascinating hypothesis – like you said “a loooong time ago” – in my experience the consensus was that an internet year was 12 weeks, i.e. one week for each month in RL. I would contend that an AI year is now 12 days, i.e. one day for each month in RL. Even if it was double then 24 so lets call it 18 for now.One could then juxtapose this into previously ‘normal’ / historical dev timelines.
1) short timeframe – we are seeing wearable prototypes & beneficial ‘connections’ e.g. Gert-Jan Oskam whilst not robotics/AI per se it is a very short step.
So in RL context: 3 years
So in AI context: 54 days2) mid timeframe – so then looking at something like the Tree of Thoughts paper
So in RL context: 3 years
So in AI context: 54 days3) long timeframe – so highly speculative on my behalf, but taking 1 & 2 a five year RL comparison would not be unreasonable. Given it is somewhat sequential if 1 hits circa 60 days in a lab somewhere I think then 2 is almost running concurrent. Your 3 is the most interesting milestone then so.
So in RL context: 5 years
So in AI context: 90 days (from about August 2023 forward) so by the end of 2023 max.I appreciate the ‘mainstream’ implementation may come along in the old year by year timeline but the year by year approach is doomed… <g>.
Great article 🙂
Most interesting to explore further if you have the time.Regards
DesHi All,
Yes, I’m in..So they say
“A basic, illustrative prototype of a system that utilizes ChatGPT to promote deliberation and encourage consensus building, inspired by pol.is.”Alas their direction of “Grant recipients are expected to implement a proof-of-concept / prototype, engaging at least 500 participants and will be required to publish a public report on their findings by October 20, 2023.”
500 is not much of a representative sample in the context of the importance of the issue. It would appear that they may also require a degree of stratification in the ‘500’ but perhaps that is for another day.
I presume all have read their guidelines so please forgive me for regurgitating them here:
https://openai.com/blog/democratic-inputs-to-ai
https://openai.com/blog/democratic-inputs-to-ai#application-review-factors
https://openai.com/blog/democratic-inputs-to-ai#policy-statements-under-considerationWithin the ‘Application review factors’ (below) they have not offered any weightings:
Evaluation: We encourage participants to establish metrics for evaluating the quality of their methods, such as participant satisfaction, shifts in polarization, scalability, or other relevant indicators, and to invent new metrics for a healthy democratic process.
Robustness: Measures to prevent or address inappropriate behavior, such as trolling and fake accounts.
Inclusiveness and representativeness: Strategies for including individuals from diverse backgrounds and levels of familiarity with AI systems in the democratic process.
Empowerment of Minority Opinions: Ensuring that unpopular or minority opinions are heard and providing smaller groups the opportunity to influence matters of significant concern to them.
Effective Moderation: Addressing challenges in moderation, including ensuring diverse representation of viewpoints, distinguishing valuable contributions from “off-topic” comments, and preventing moderator biases from influencing the process.
Scalability: We emphasize scalable processes that can be conducted virtually, rather than through in-person engagement. We are aware that this approach might sacrifice some benefits associated with in-person discussions, and we recognize that certain aspects could be lost in a virtual setting.
Actionability: The degree of actionability of the information elicited by the deliberation process.
Legibility: How easy it is to understand and trust the process.Nonetheless we could of course evolve our own ‘personality rubric’ for this.
There is so much ‘questionable’ with regard to their ‘viewpoint’ as articulated here that one must wonder is this just simply a PR exercise. In Ireland we call this the tail wagging the dog. It does appear that there is a degree of dystopianism creeping in, one wonders about the words / phrases:
- inappropriate behavior
- the opportunity to influence
- preventing moderator biases
- lost in a virtual setting
Whilst I would participate wholeheartedly is does seem – as if to use Microsoft Word as an example – that it was being proposed that one could not use Word to write about pollution, climate change or a host of other politically charged issues. This inherent attempt to ‘throttle’ the AI is very disappointing. (imho)
“The primary objective of this grant is to foster innovation in processes – we need improved democratic methods to govern AI behavior. The specific answers to the questions matter less than the advancements made in the process itself.”
Modifying this it could well read..
The primary objective of this grant is to foster innovation in humans – we need improved democratic methods to govern human behavior. The specific answers to the questions matter less than the advancements made in the human model itself.Thanks Brian for suggesting this!
Regards
DesHi Linnea,
this is a fascinating hypothesis – like you said “a loooong time ago” – in my experience the consensus was that an internet year was 12 weeks, i.e. one week for each month in RL. I would contend that an AI year is now 12 days, i.e. one day for each month in RL. Even if it was double then 24 so lets call it 18 for now.One could then juxtapose this into previously ‘normal’ / historical dev timelines.
1) short timeframe – we are seeing wearable prototypes & beneficial ‘connections’ e.g. Gert-Jan Oskam “movement intentions are decoded in real time from brain recordings”
So in RL context: 3 years
So in AI context: 54 days2) mid timeframe – so then looking at something like the Tree of Thoughts paper
So in RL context: 3 years
So in AI context: 54 days3) long timeframe – so highly speculative on my behalf, but taking 1 & 2 a five year RL comparison would not be unreasonable. Given it is somewhat sequential if 1 hits circa 60 days in a lab somewhere I think then 2 is almost running concurrent. Your 3 is the most interesting milestone then so.
So in RL context: 5 years
So in AI context: 90 days (from about August 2023 forward) so by the end of 2023 max.I appreciate the ‘mainstream’ implementation may come along in the old year by year timeline but the year by year approach is doomed… <g>.
Great article 🙂
Most interesting to explore further if you have the time.Regards
DesHey SK,
forgive me if this is repeated since I did not see it in the stream..You are 100% right, it is no dream, it is reality…
In that context then this job from Anthropic in SF CA
https://jobs.lever.co/Anthropic/e3cde481-d446-460f-b576-93cab67bd1edThey say
“The expected salary range for this position is $280k – $375k.”Regards
DesHi – greetings from Ireland..
With regard to “expert who knows basic coding” this is not necessarily the case. In terms of Chat GPT we are dealing with an LLM, Large Language Model. By that fact then use of language remains the primary input method.Semantic and syntactic considerations are two critical components of natural language processing (NLP) and text generation.
From a semantic viewpoint:
Grammar: Pay attention to subject-verb agreement, verb tense consistency, and proper use of articles (a, an, the).
Syntax: Ensure proper word order and sentence structure to maintain clarity and readability.
Punctuation: Use appropriate punctuation marks, such as commas, periods, question marks, and quotation marks, to enhance the structure and meaning of the text.Semantics:
Meaning: Ensure that prompts convey the intended meaning and effectively communicate the desired information.
Context: Take into account the specific context and generate prompts accordingly.In many regards we are dealing with a 4 year old, don’t ask, tell.
At the next level are the primary nuances of language, the basic rules of grammar and syntax that govern language. These include things like subject-verb agreement, tense, word choice and word order. Word choice / Syntax / Connotation / Denotation / Figurative
An interesting aspect here is Roles, the syntax for a role is:
Role: {the role} Prompt {the prompt}
One can ask the AI to act in ‘role’, just like in a play / film the AI will adopt that role, e.g.Role:{Storyteller} Prompt: {please provide a 100 word story using “Inverted order” for stylistic or rhetorical purposes with the subject of “Mars exploration”}
AI: In the vast expanse of space, they journeyed, their eyes fixed on the red planet. Mars, the celestial enigma, beckoned them with its mysteries. Cloaked in anticipation, the explorers embarked on a daring quest. Step by step, they ventured across the barren landscape, their footsteps leaving imprints on the ancient soil. The thin atmosphere whispered secrets, as they delved into the secrets of this alien world. With every discovery, a new chapter unfolded, unveiling the enigmatic history of Mars. The triumphs and tribulations merged in a dance of exploration, revealing the cosmos’ hidden wonders to those who dared to venture.
Beyond this we must look at:
Intent Of The Prompt / Tone Of The Prompt / Keyword Extraction, Query Expansion, And ParaphrasingParaphrasing Example: “What led to the French Revolution and what were the consequences?”
Keyword Extraction Example: “What do we know about the Reign of Terror during the French Revolution?”
Query Expansion Example: “What political, social and economic conditions led to the Reign of Terror during the French Revolution?”In my own opinion the conciseness of language, proper grammar & punctuation, the use of Role / Intent / Tone / Explicit And Implicit Prompts / Active & Passive Voice all lead to iteration of the optimum output.
I’d encourage you to use the free version to experiment, in reality I believe it is all about English Language 101. Personally I am learning more about English from the AI than I ever did in school.As a final note if you have something in your particular field ask it to phrase your prompt for you.
Role: You are an expert mortgage broker Prompt: {tell me how to best ask … your question}Also if you find a topic of real interest ask it “Please elaborate / expand / provide 5 examples.
Note that it will render information into two/three etc column table and you can specify the column headings. FYI its math is pretty poor.Regards
Desso a bunch of ???????????? as you predicted and then this:-
) WHERE last_name = ? “, [‘NAT’], function(err, res){
if(err) throw err;
console.log(res);
})
setTimeout(() => {
const check = db.query(‘SELECT * FROM member WHERE last_name=?’, [‘NAT’], (err, result)=>{
if(err) throw err;
console.log(result);
res.json(result);
})
}, 100)
});
});// app.post(‘/AruniAuto/api/v1/tbl_employee_insert’, (req, res) => {
// const name = req.body.name;
// const age = req.body.age;
// const address = req.body.address;// db.query(“INSERT INTO tbl_employee (name, age, address) VALUES (?,?,?)”, [name, age, address], (err, result) => {
// if(err){
// console.log(err);
// }
// else {
// res.send(“Value Inserted”);
// }
// })
// });app.listen(‘3001’, () => {
console.log(“Server started on port 3001”)
})<|endoftext|>
