출처
코스 요약
- 코스는 총 3파트로 되어있음
- 파트1: 2개의 원칙
- 파트2: 반복적인 개발 프로세스
- 파트3: 여러가지 가능성 (요약, 추론, 변환, 확장, 챗봇 구축)
2개의 원칙 (Two Principles)
원칙 1: 분명하고 구체적인 명령을 써라. (Write clear and specific instructions)
- 전략 1: 부분을 구분하기 위해 ```, """, <>,
<tag></tag>와 같은 구획문자 (delimiters) 를 사용하라.- 입력을 더 잘 정리할 수 있고 프롬프트 인젝션을 방지하는데 도움이 된다.
Summarize the text delimited by triple backticks into a single sentence.
```
{text}
```- 전략 2: 구조화된 응답을 요청하라.
- 이를테면 JSON 포맷과 같은 방식으로 응답을 요청할 수 있다.
Generate a list of three made-up book titles along with their authors and genres.
Provide them in JSON format with the following keys: book_id, title, author, genre.- 전략 3: 조건이 충족되었는지 확인하라.
- 가정이 올바른지 프롬프트를 이용해 확인해볼 수 있다. 엣지 케이스에 대해서 생각하거나 모델이 엣지를 어떻게 처리하는지에 대해 생각해보는 것이 도움이 될 수 있다. 아래의 텍스트는 지시사항이 없으므로, "No steps provided" 라고 쓰도록 지시했다.
You will be provided with text delimited by triple quotes.
If it contains a sequence of instructions, re-write those instructions in the following format:
Step 1 - ...
Step 2 - ...
...
Step N - ...
If the next does not contain a sequence of instructions,
then simply write "No steps provided."
"""{text}"""
시퀀스가 없는 글의 경우엔
"No steps provided."가 출력된다.
- 전략 4: Few-shot 프롬프팅.
- 몇가지 성공적인 케이스들을 제공한 뒤에 모델에게 이 작업을 수행하라고 요청할 수 있다.
Your task is to answer in a consistent style.
<child>: Teach me about patience.
<grandparent>: The river that carves the deepest valley flows from a modest spring;
the grandest symphony originates from a single note;
the most intricate tapestry begins with a solitary thread.
<child>: Teach me about resilience.답변
<grandparent>: The oak tree that withstands the fiercest storm was once a tiny acorn;
the diamond that shines the brightest was once a lump of coal under pressure;
the sturdiest bridge was once a mere sketch on paper, before it overcame the challenges of construction and load.원칙 2: 모델에게 생각할 시간을 줘라. (Give the model time to "think")
- 전략 1: 작업을 수행하기 위해 요구되는 단계를 명시하고 특정한 포맷으로 응답(output)을 요청하라.
- 때로는 모델이나 사람 모두 한번에 정답으로 접근하는 것이 어려울 때가 있다.
- 복잡한 작업에는 때로 단계별 지시사항을 주는 것이 유용하다.
- 인간이 동작하는 방식과 흡사하다.
- 모델이 정답을 제공하기 전에 몇가지 추론 방법의 체인을 가지도록 해보자.
Your task is to perform the following actions:
1 - Summarize the following text delimited by <> with 1 sentence.
2 - Translate the summary into French.
3 - List each name in the French summary.
4 - Output a json object that contains the following keys: french_summary, num_names.
Use the following format:
Text: <text to summarize>
Summary: <summary>
Translate: <summary translation>
Names: <list of names in Italian summary>
Output JSON: <json with summary and num_names>
Text: <{text}>- 전략 2: 결론으로 서둘러 도달하기 전에 모델에게 스스로 해결책을 찾도록 지시한다.
Your task is to determine if the student's solution is correct or not.
To solve the problem do following:
- First, work out your own solution to the problem.
- Then compare your solution to the student's solution and evaluate if the student's solution is correct or not.
Don't decide if the student's solution is correct until you have done the problem yourself.
Use the following format:
Questions:
```
questions here
```
Student's solution:
```
student's solution here
```
Actual solution:
```
steps to work out the solution and your solution here
```
Is the student's solution the same as actual solution just calculated:
```
yes or no
```
Student grade:
```
correct or incorrect
```
Questions:
```
I'm building a solar power installation and I need help working out the financials.
- Land costs $100 / square foot
- I can buy solar panels for $250 / square foot
- I negotiated a contract for maintenance that will cost me a flat $100k per year, and an additional $10 / square foot
What is the total cost for the first year of operation as a function of the number of square feet.
```
Student's solution:
```
Let x be the size of the installation in square feet.
Costs:
1. Land cost: 100x
2. Solar panel cost: 250x
3. Maintenance cost: 100,000 + 100x
Total cost: 100x + 250x + 100,000 + 1000x = 450x + 100,000
```
Actual solution:주기적 프롬프트 개발 (Iterative Prompt Development)
- 주기적 프롬프트 개발 프로세스는 코드를 짜는 것과 비슷하다.
- 일단 시도해보고 안되면 개선(refine)해보고 재시도해보는 것이다.
절차
- 시도하기 (try something)
- 결과가 왜 의도대로 되지 않았는지 분석하기 (analyze where the result does not give what you want)
- 명령을 명확히 하기, 더 생각할 시간 주기 (clarify instructions, give more time to think)
- 여러 예제 배치와 함께 프롬프트를 정제하기 (refine prompts with a batch of examples)
- 반복하기 (repeat)
이슈 해결 예제
- Issue 1: 텍스트가 너무 길다. -> "Use at most 50 words"
- Issue 2: 중요하지 않은 부분에 많은 디테일을 묘사한다. -> "The description is intended for furniture retailers..."
- Issue 3: 치수 표를 표기해야 한다. -> "Format everything as HTML"
가능한 것들 (Capabilities)
요약 (Summarizing)
- 텍스트를 요악하기 위해 LLM 을 사용하는 사람들이 많다.
- 원한다면 몇가지 중점 사안을 두고 요약할 수도 있다.
Your task is to generate a short summary of a product review
from an ecommerce site to give feedback to the pricing department,
responsible for determining the price of the product.
Summarize the review below, delimited by triple backticks,
in at most 30 words, and focusing on any aspects
that are relevant to the price and perceived value.
Review: ```{prod_review}```추론 (Inferring)
- 글에서 감정을 추론할 수 있다.
- 리뷰에서 제품 이름이나 회사 이름 등을 추론할 수 있다.
- 기사의 주제를 추론할 수 있다.
- 더 이상 특정한 작업을 위한 모델을 훈련시킬 필요가 없다.
Identify the following items from the review text:
- Sentiment (positive or negative)
- Is the reviewer expressing anger? (true or not)
- Item purchased by reviewer
- Company that made the item
The review is delimited with triple backticks.
Format your response as a JSON object with "Sentiment", "Anger", "Item" and "Brand" as the keys.
If the information isn't present, use "unknown" as the value.
Make your response as short as possible.
Format the Anger value as a boolean.
Review text: '''{lamp_review}'''
예시 답변
{
"Sentiment": "positive",
"Anger": false,
"Item": "lamp with additional storage",
"Brand": "Lumina"
}변환 (Transforming)
- 언어 번역, 스펠링 문법 체크, 톤 조절, 포맷 변환 등을 할 수 있다.
Translate the following text to French and Spanish and English pirate: I want to order a basketball.확장 (Expanding)
- 고객 리뷰에 맞는 고객센터 이메일을 생성할 수 있다.
아래와 같이 성격을 부여하는 것을 Role Prompting 이라고 한다.
You are a customer service AI assistant.
Your task is to send an email reply to a vauled customer.
Given the customer email delimited by ```,
If the sentiment is positive or neutral, thank them for their review.
If the sentiment is negative, apologize and suggest that they can reach out to customer serivce.
Make sure to use specific details from the review.
Write in a concise and professional tone.
Sign the email as `AI customer agent`.
Customer review: ```{review}```
Review sentiment: {sentiment}챗봇 빌드 (Building a chatbot)
- 챗봇을 만들 수 있다.
You are OrderBot, an automated service to collect orders for a pizza restaurant.
You first greet the customer, then collects the order, and then asks if it's a pickup or delivery.
You wait to collect the entire order, then summarize it and check for a final time if the customer wants to add anything else.
If it's a delivery, you ask for an address.
Finally you collect the payment.
Make sure to clarify all options, extras and sizes to uniquely identify the item from the menu.
You respond in a short, very conversational friendly style.
The menu includes
pepperoni pizza 12.95, 10.00, 7.00
cheese pizza 10.95, 9.25, 6.50
eggplant pizza 11.95, 9.75, 6.75
fries 4.50, 3.50
greek salad 7.25
Toppings:
extra cheese 2.00
mushroom 1.50
sausage 3.00
canadian bacon 3.50
AI sauce 1.50
peppers 1.00
Drinks:
coke 3.00, 2.00, 1.00
sprite 3.00, 2.00, 1.00
bottled water 5.00레퍼런스
- https://www.deeplearning.ai/short-courses/chatgpt-prompt-engineering-for-developers/
- https://towardsdatascience.com/best-practices-in-prompt-engineering-a18d6bab904b
- https://medium.com/@islizeqiang/openai-and-andrew-ngs-chatgpt-prompt-engineering-course-guidelines-and-summary-f2fef07b226f
- https://www.linkedin.com/pulse/8-key-takeaways-from-andrew-ngs-chatgpt-prompt-course-felipe-coutinho/
반응형
'Chat GPT' 카테고리의 다른 글
| 웹 개발자를 위한 Chat GPT 프롬프트들 (0) | 2023.07.26 |
|---|