본문 바로가기
카테고리 없음

ChatGPT 채팅GPT

by EasyGPT 2023. 2. 5.
반응형
#ChatGPT #채팅GPT

ChatGPT는 대화방식으로 상호작용하는 OpenAI에서 훈련된 고급 AI 챗봇입니다.
대화형식을 통해 ChatGPT는 후속 질문에 답하고, 실수를 인정하고, 잘못된 전제에 이의를 제기하고, 부적절한 요청을 거부할 수 있습니다.
ChatGPT는 강력한 GPT-3.5 기술을 사용합니다.
GPT는 Generative Pre-Trained Transformer의 약자로, 혁신적 Attention 개념을 기반으로 하는 복잡한 #신경망 입니다.
ChatGPT is an advanced AI chatbot trained by OpenAI which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
ChatGPT relies on the powerful GPT-3.5 technology. GPT stands for Generative Pre-Trained Transformer, a complex neural network based on the revolutionary Attention concept.

Able to generate anything you want
당신이 원하는 무엇이든 생성 가능

ChatGPT는 2022년 초 훈련을 마친 GPT-3.5 시리즈 모델에서 미세조정되었습니다.
여기에서 3.5 시리즈에 대해 자세히 알아볼 수 있습니다.
ChatGPT 및 GPT 3.5는 #Azure AI 슈퍼컴퓨팅 인프라에서 훈련되었습니다.
ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished training in early 2022. You can learn more about the 3.5 series here. ChatGPT and GPT 3.5 were trained on an Azure AI supercomputing infrastructure.

Limitations 제한사항

ChatGPT는 때때로 그럴듯해 보이지만 부정확하거나 무의미한 답변을 작성합니다.
다음과 같은 이유로 이 문제를 해결하는 것은 어렵습니다:
(1) RL 훈련 중에는 현재 진실의 출처가 없습니다.
(2) 더 신중하도록 모델을 훈련하면 올바르게 대답할 수 있는 질문을 거부합니다. 그리고
(3) 이상적 대답은 인간 시연자가 아는 것이 아닌, 모델이 알고 있는 것에 달려 있기 때문에 #지도훈련 은 모델을 오도합니다.
ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly; and (3) supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows.


Iterative Deployment 반복 배포

ChatGPT의 오늘 연구 릴리스는 점점 더 안전하고 유용한 AI 시스템을 반복적으로 배포하는 OpenAI의 최신 단계입니다.
GPT-3 및 Codex 같은 이전 모델 배포에서 얻은 많은 교훈은 #인간피드백강화학습 ( #RLHF )을 사용하여 달성한 유해하고 거짓된 출력의 상당한 감소 등 이 릴리스에 대한 안전 보강을 알렸습니다.
Today’s research release of ChatGPT is the latest step in OpenAI’s iterative deployment of increasingly safe and useful AI systems. Many lessons from deployment of earlier models like GPT-3 and Codex have informed the safety mitigations in place for this release, including substantial reductions in harmful and untruthful outputs achieved by the use of reinforcement learning from human feedback (RLHF).

How we collect data 데이터 수집 방법

🦾 Conversations may be reviewed by our AI trainers to improve our systems.
시스템을 개선하기 위해 AI 트레이너가 대화를 검토할 수 있습니다.
🔐 Please don't share any sensitive information in your conversations.
대화에서 민감한 정보를 공유하지 마십시오.





The ChatGPT Cheat Sheet 

이 치트시트는 #개발자  #콘텐츠제작자  #미디어콘텐츠생성 #자연어처리 #프로그래밍 등 다양한 도메인에서 프롬프트하는 대규모 언어모델의 숙련도를 향상할 수 있는 OpenAI의 ChatGPT의 다양한 기능을 보여줍니다.
This cheat sheet illustrates the diverse abilities of OpenAI’s ChatGPT for developers and content creators to enhance their proficiency in large language model prompting across various domains including media content creation, natural language processing, and programming. 




Table of Contents:


  1. NLP Tasks
  2. Code
  3. Structured Output Styles
  4. Unstructured Output Styles
  5. Media Types
  6. Meta ChatGPT
  7. Expert Prompting



1. NLP Tasks

Text Generation 텍스트 생성
Prompt: write an intro paragraph to a mystery novel 추리소설에 소개문을 쓰세요



As the mist settled over the small town of Ravenswood, a sense of unease descended upon its residents. The disappearance of local heiress, Emily Parker, had everyone on edge. Detective Jameson, freshly transferred to the quiet town, found himself in the midst of his toughest case yet. With a limited pool of suspects and a town full of secrets, he knew it wouldn't be an easy solve. But he was determined to get to the bottom of it, no matter the cost. Little did he know, the mystery of Emily's disappearance was only the tip of the iceberg in a web of deceit and corruption that threatened to destroy Ravenswood and all who called it home.
Ravenswood라는 작은 마을에 안개가 드리워지자 주민들은 불안감을 느꼈습니다. 지역 상속녀인 에밀리 파커의 실종으로 모두가 긴장했습니다. 조용한 마을로 막 전근 온 제임슨 형사는 자신이 가장 힘든 사건의 한가운데에 있음을 알게 되었습니다. 제한된 용의자 풀과 비밀로 가득 찬 마을에서 그는 이 문제를 쉽게 해결할 수 없다는 것을 알았습니다. 그러나 그는 어떤 대가를 치르더라도 끝까지 해결하기로 결심했습니다. 그는 거의 알지 못했습니다. 에밀리의 실종 미스터리는 레이븐스우드와 그곳을 고향이라고 부르는 모든 사람들을 파괴하겠다고 위협하는 속임수와 부패의 그물망에 있는 빙산의 일각에 불과했습니다.

It was a dark and stormy night when the body was found. The small town of Millfield was in shock as the news spread that one of their own, local businessman, had been murdered in cold blood. The police were at a loss for leads and the residents of the town were on edge. But one person, a young and ambitious reporter, was determined to uncover the truth and find the killer. Little did she know, this case would lead her down a treacherous path of lies, deception, and hidden secrets that would shake the very foundations of the town. As she delves deeper into the investigation, she begins to realize that nothing is as it seems and that the killer may be closer than she ever imagined.
시체가 발견된 것은 어둡고 폭풍우가 치는 밤이었습니다. Millfield의 작은 마을은 자신의 지역 사업가 중 한 명이 냉혹하게 살해되었다는 소식이 퍼지자 충격에 휩싸였습니다. 경찰은 실마리를 찾지 못했고 마을 주민들은 긴장했습니다. 그러나 젊고 야심 찬 한 사람은 진실을 밝히고 살인자를 찾기로 결심했습니다. 그녀는 거의 알지 못했지만, 이 사건이 그녀를 거짓, 속임수, 그리고 마을의 근간을 뒤흔들 숨겨진 비밀의 위험한 길로 이끌 것입니다. 그녀는 수사를 더 깊이 파고들면서 보이는 것과 다른 것은 없으며 범인이 그녀가 상상했던 것보다 더 가까이에 있을지도 모른다는 사실을 깨닫기 시작합니다.




Summarization 요약
Prompt: summarize this text: 이 본문을 요약하십시오
It was a dark and stormy night when the body was found. The small town of Millfield was in shock as the news spread that one of their own, local businessman, had been murdered in cold blood. The police were at a loss for leads and the residents of the town were on edge. But one person, a young and ambitious reporter, was determined to uncover the truth and find the killer. Little did she know, this case would lead her down a treacherous path of lies, deception, and hidden secrets that would shake the very foundations of the town. As she delves deeper into the investigation, she begins to realize that nothing is as it seems and that the killer may be closer than she ever imagined.
시체가 발견된 것은 어둡고 폭풍우가 몰아치는 밤이었습니다. Millfield의 작은 마을은 자신의 지역 사업가 중 한 명이 냉혹하게 살해되었다는 소식이 퍼지자 충격에 휩싸였습니다. 경찰은 실마리를 찾지 못했고 마을 주민들은 긴장했습니다. 그러나 젊고 야심 찬 한 사람은 진실을 밝히고 살인자를 찾기로 결심했습니다. 그녀는 거의 알지 못했지만, 이 사건이 그녀를 거짓, 속임수, 그리고 마을의 근간을 뒤흔들 숨겨진 비밀의 위험한 길로 이끌 것입니다. 그녀는 수사를 더 깊이 파고들면서 보이는 것과 다른 것은 없으며 범인이 그녀가 상상했던 것보다 더 가까이에 있을지도 모른다는 사실을 깨닫기 시작합니다.

A businessman in Millfield is murdered on a dark and stormy night, leaving the small town in shock and the police with no leads. A young reporter sets out to uncover the truth and finds a web of lies and hidden secrets that threatens to destroy the town, realizing the killer may be closer than she thought.
Millfield의 한 사업가가 어둡고 폭풍우가 치는 밤에 살해되어 작은 마을은 충격에 빠지고 경찰은 단서가 없어집니다. 젊은 기자는 진실을 밝히기 위해 나섰고, 마을을 파괴하겠다고 위협하는 거짓말과 숨겨진 비밀의 거미줄을 발견하고 살인자가 생각보다 가까이에 있을 수 있음을 깨닫습니다.

A young reporter in a small town investigates the murder of a local businessman and uncovers a web of lies, deception and hidden secrets that shake the foundations of the town. The police have no leads and the town is on edge. The reporter delves deeper into the investigation and realizes the killer may be closer than she imagined.
작은 마을의 젊은 기자가 지역 사업가의 살인 사건을 조사하고 마을의 근간을 뒤흔드는 거짓말과 속임수, 숨겨진 비밀을 밝혀냅니다. 경찰은 단서가 없고 마을은 위기에 처해 있습니다. 기자는 조사에 대해 더 깊이 파고들고 범인이 그녀가 상상했던 것보다 더 가까이에 있을 수 있음을 깨닫습니다.



Open Domain Question Answering 오픈 도메인 질문 답변
Prompt: when did Apollo 11 land on the moon 아폴로 11호는 언제 달에 착륙했는가

Apollo 11 landed on the moon on July 20, 1969.


Paraphrasing
Prompt: rewrite this text: Apollo 11 landed on the moon on July 20, 1969 1969년 7월 20일 아폴로 11호가 달에 착륙했습니다.

On July 20, 1969, Apollo 11 mission successfully landed on the moon.
On July 20, 1969, Apollo 11 successfully touched down on the lunar surface.




건강전문 몰
www.dopza.com

Sentiment Analysis (few-shot or zero-shot)
Prompt: I like pizza, positive
I don't like pizza, negative
sometimes, I like pizza sometimes I don't, neutral
while the movie was good, I sometimes though it was a bit dry,
나는 피자를 좋아한다, 긍정
나는 피자를 좋아하지 않는다, 네거티브
가끔은 피자가 좋다 가끔은 싫다, 중립
영화가 좋은데 가끔 좀 건조하긴 했지만,

neutral





Table to Text
Prompt: summarize the data in this table: I like pizza positive I don't like bananas negative
Sometimes I like pizza but sometimes I don't neutral
While the movie in general was pretty good, I sometimes thought it was a bit dry neutral


The sentiment towards pizza is positive, towards bananas is negative, and towards the movie is neutral. Sometimes the sentiment towards pizza is neutral and sometimes it is negative.
피자에 대한 감정은 긍정적이고 바나나에 대한 감정은 부정적이며 영화에 대한 감정은 중립적입니다. 피자에 대한 감정은 중립적일 때도 있고 부정적일 때도 있다.

This table contains four statements, two of them express a positive negative opinion about two different subjects, Pizza and Bananas, other two statements express a neutral opinion about the same subjects, Pizza and a movie.
이 표에는 4개의 문이 포함되어 있으며, 그 중 2개는 서로 다른 2가지 주제인 피자와 바나나에 대해 긍정적인 부정적 의견을 표현하고, 다른 2개는 동일한 주제인 피자와 영화에 대해 중립적인 의견을 표현합니다.



Text to Table
Prompt: create a table from this text: create a 2 column table where the first column contains the stock ticker symbol for Apple, Google, Amazon, Meta, and the other column contains the names of the companies.
이 텍스트에서 테이블 만들기: 첫 번째 열에는 Apple, Google, Amazon, Meta의 주식 시세 기호가 포함되고 다른 열에는 회사 이름이 포함된 2열 테이블을 만듭니다.
 



Token Classification (few-shot or zero-shot)
Prompt: classify the named entities in this text: George Washington and his troops crossed the Delaware River on December 25, 1776 during the American Revolutionary War.
이 텍스트에서 명명된 독립체를 분류하십시오: 조지 워싱턴과 그의 군대는 미국 독립 전쟁 중 1776년 12월 25일에 델라웨어 강을 건넜습니다.
 


Dataset Generation (few-shot or zero-shot)
Prompt: generate more datapoints from this text:
"contains no wit , only labored gags "
0 (negative)
"that loves its characters and communicates something rather beautiful about human nature "
1 (positive)
"remains utterly satisfied to remain the same throughout "
0 (negative)

"Lacks creativity and humor, just uninspired jokes" 0 (negative)
"Celebrates its characters and conveys a heartwarming message about human nature" 1 (positive)
"Stays stagnant and fails to evolve" 0 (negative)
"Predictable and unoriginal" 0 (negative)
"Focuses solely on cheap laughs rather than genuine humor" 0 (negative)
"Has a monotonous, repetitive style" 0 (negative)



Machine Translation
Prompt: translate this text into Portuguese: welcome to the matrix
이 텍스트를 포르투갈어로 번역: 매트릭스에 오신 것을 환영합니다

Bem-vindo à matrix.



2. Code

Code Generation
Prompt: show me how to make an http request in Python
Python에서 http 요청을 만드는 방법을 보여주세요.


In Python, you can make an HTTP request using the requests library. Here's an example of a simple GET request:
python
import requests response = requests.get("https://www.example.com") if response.status_code == 200: print(response.text) else: print("Request failed")
In this example, we're sending a GET request to https://www.example.com using the requests.get() method. The response of the request is stored in the response variable. We then check the status_code attribute of the response to see if the request was successful (200 means success). If the request was successful, we print the response text. If not, we print an error message.
You can also send other types of requests, such as POST, PUT, and DELETE, using the corresponding methods in the requests library:
makefile
import requests # POST request response = requests.post("https://www.example.com/api/create_user", data={"username": "john", "email": "john@example.com"}) # PUT request response = requests.put("https://www.example.com/api/update_user/1", data={"username": "john_doe"}) # DELETE request response = requests.delete("https://www.example.com/api/delete_user/1")


Code Explanation
Prompt: explain this python code: from deepsparse import Pipeline qa_pipeline = Pipeline.create(task="question-answering") inference = qa_pipeline(question="What's my name?", context="My name is Snorlax") >> {'score': 0.9947717785835266, 'start': 11, 'end': 18, 'answer': 'Snorlax'}

This Python code is using a library called deepsparse to create a question-answering pipeline. The Pipeline class is imported from the deepsparse library and a pipeline object qa_pipeline is created by calling the create() method with the task parameter set to "question-answering".
Once the pipeline has been created, it is then called on a question and context string with the following code: inference = qa_pipeline(question="What's my name?", context="My name is Snorlax"). The question and context input to the pipeline is expected to be strings and the output is a dictionary containing the score, start, end, and answer of the best answer found in the context for the given question.
The final line {'score': 0.9947717785835266, 'start': 11, 'end': 18, 'answer': 'Snorlax'} is the output of the pipeline, which shows that the pipeline has found the answer "Snorlax" in the context with a score of 0.9947717785835266 (indicating a high level of confidence in the answer), starting at index 11 and ending at index 18.

This code is using a library called "deepsparse" to create a pipeline for question-answering. from deepsparse import Pipeline imports the "Pipeline" class from the "deepsparse" library, which allows the user to define a pipeline for a specific task like question-answering. qa_pipeline = Pipeline.create(task="question-answering") creates a new pipeline object for the task of question-answering and assigns it to the variable "qa_pipeline". inference = qa_pipeline(question="What's my name?", context="My name is Snorlax") is passing the values of the question and context to the pipeline object "qa_pipeline" and storing the result in a variable called "inference". The inference variable stores a dictionary containing the score of the answer, the start and end index of the answer in the context, and the answer itself. The returned output would look like {'score': 0.9947717785835266, 'start': 11, 'end': 18, 'answer': 'Snorlax'} It appears that the pipeline is using some form of machine learning model to understand the question and context, and extract the answer from the context based on the question. The library "deepsparse" is meant for public use, this code is an example of how to use it to perform question-answering task.

Docstrings Generation
Prompt: write a docstring description for this function: import requests def make_get_request(url): response = requests.get(url) return response.status_code, response.text make_get_request('https://www.example.com')

python
import requests def make_get_request(url): """ Make a GET request to the specified URL and return the status code and text of the response. Args: - url (str): The URL to make a GET request to. Returns: - tuple: A tuple containing the status code (int) and text (str) of the response. Example: >>> make_get_request('https://www.example.com') (200, "...") """ response = requests.get(url) return response.status_code, response.text


Programming Language Conversion
Prompt: convert this code from Python to Javascript: print("hello world")

console.log("hello world");

Data Object Conversions (JSON, XML, CSV etc.)
Prompt: convert this JSON object into XML: {"Name":{"0":"John Smith","1":"Jane Doe","2":"Bob Johnson","3":"Samantha Williams"},"Age":{"0":32,"1":28,"2":45,"3":40},"Gender":{"0":"Male","1":"Female","2":"Male","3":"Female"},"O ccupation":{"0":"Software Developer","1":"Data Analyst","2":"Project Manager","3":"Marketing Director"}}

<?xml version="1.0"?>
<root>
<Name>
<item key="0">John Smith</item>
<item key="1">Jane Doe</item>
<item key="2">Bob Johnson</item>
<item key="3">Samantha Williams</item>
</Name>
<Age>
<item key="0">32</item>
<item key="1">28</item>
<item key="2">45</item>
<item key="3">40</item>
</Age>
<Gender>
<item key="0">Male</item>
<item key="1">Female</item>
<item key="2">Male</item>
<item key="3">Female</item>
</Gender>
<Occupation>
<item key="0">Software Developer</item>
<item key="1">Data Analyst</item>
<item key="2">Project Manager</item>
<item key="3">Marketing Director</item>
</Occupation>
</root>

Knowledge Graph Generation
Prompt:convert this text into nodes and edges: BabeRuth joined the New York Yankees in 1920. The Boston Red Sox sold his contract to the Yankees. He played for the Yankees from 1920 to 1934, and during that time he established himself as one of the greatest players in baseball history.

Here is an example of how the text could be converted into nodes and edges: Nodes: •Babe Ruth •New York Yankees •Boston Red Sox •baseball history Edges: •Babe Ruth (node) joined (edge) New York Yankees (node) in 1920 (edge property) •Boston Red Sox (node) sold (edge) Babe Ruth's contract (edge property) to (edge) New York Yankees (node) •Babe Ruth (node) played for (edge) New York Yankees (node) from 1920 (edge property) to 1934 (edge property) •Babe Ruth (node) established himself as (edge) one of the greatest players (edge property) in baseball history (node) It's important to note that, the representation of the knowledge graph can be different depending on the purpose of the graph, and depending on the data available in the text.


HTML to Text (Web Scraping)
Prompt: convert this HTML to text: <h1 class="heading1" id="neural-magic-platform-documentation">Neural Magic Platform Documentation</h1>

Neural Magic Platform Documentation



3. Structured Output Styles

List
Prompt: give me a list of 5 citrus fruits

Lemon, Lime, Orange, Grapefruit, Tangerine

Here is a list of 5 citrus fruits:
●Oranges
●Lemons
●Limes
●Grapefruit
●Tangerines


Numbered List
Prompt:give me a numbered list of 5 citrus fruits

Sure! Here is a list of five citrus fruits:
  1. Orange
  2. Lemon
  3. Lime
  4. Grapefruit
  5. Tangerine



Headings and Subheadings
Prompt:convert this text into headings and subheadings:Babe Ruth joined the New York Yankees in 1920. The Boston Red Sox sold his contract to the Yankees. He played for the Yankees from 1920 to 1934, and during that time he established himself as one of the greatest players in baseball history.

Heading: Babe Ruth Joins New York Yankees
Subheading: Transfer from Boston Red Sox
Subheading: Career with the Yankees (1920-1934)
Subheading: Establishing himself as a Baseball Great

Here is an example of how the text could be converted into headings and subheadings: 1.Introduction ○The text is about Babe Ruth and his time playing for the New York Yankees ○The text covers his move from the Boston Red Sox to the Yankees and his time playing for the team 2.Babe Ruth's move to the New York Yankees ○Babe Ruth joined the New York Yankees in 1920 ○The Boston Red Sox sold his contract to the Yankees 3.Babe Ruth's time with the New York Yankees ○He played for the Yankees from 1920 to 1934 ○During that time he established himself as one of the greatest players in baseball history It's important to note that the text needs to be structured and well-organized in order to create clear and meaningful headings and subheadings.


Tables
Prompt: create a table from this list: Oranges, Lemons, Limes, Grapefruit, Tangerines

Here is the table you requested:
 

4. Unstructured Output Styles

Narrative Modes (1st, 2nd or in the 3rd person)
Prompt:write a paragraph on how to make browniesin the 1st person


To make delicious brownies, first preheat your oven to 350°F (180°C) and grease a 9x9 inch (23x23 cm) baking dish. In a medium saucepan, melt 1 cup (2 sticks) of unsalted butter over low heat. Add 1 1/4 cups (250 g) of granulated sugar and 2 teaspoons of vanilla extract, stirring until well combined. Remove from heat and let cool for a few minutes. Then, add 4 large eggs, one at a time, stirring well after each addition. In a separate bowl, mix together 1 cup (125 g) of all-purpose flour, 3/4 cup (75 g) of unsweetened cocoa powder, and 1/2 teaspoon of baking powder. Gradually add the dry ingredients to the wet mixture, stirring until just combined. Finally, fold in 1 cup (170 g) of semisweet chocolate chips. Pour the batter into the prepared baking dish and bake for 25-30 minutes, or until a toothpick inserted in the center comes out clean. Allow to cool in the pan for 10 minutes, then cut into squares and serve. Enjoy your homemade, gooey and chocolaty brownies!
맛있는 브라우니를 만들려면 먼저 오븐을 350°F(180°C)로 예열하고 9x9인치(23x23cm) 베이킹 접시에 기름을 바릅니다. 중간 냄비에 무염 버터 1컵(스틱 2개)을 약한 불로 녹입니다. 알갱이 설탕 1 1/4컵(250g)과 바닐라 추출물 2티스푼을 넣고 잘 섞일 때까지 저어줍니다. 열에서 제거하고 몇 분 동안 식히십시오. 그런 다음 큰 달걀 4개를 한 번에 하나씩 넣고 잘 저어줍니다. 별도의 그릇에 다목적 밀가루 1컵(125g), 무가당 코코아 가루 3/4컵(75g), 베이킹 파우더 1/2티스푼을 섞습니다. 마른 재료를 젖은 혼합물에 서서히 첨가하고 잘 섞일 때까지 저어줍니다. 마지막으로 살짝 달콤한 초콜릿 칩 1컵(170g)을 넣습니다. 반죽을 준비된 베이킹 접시에 붓고 25-30분 동안 또는 중앙에 삽입된 이쑤시개가 깨끗해질 때까지 굽습니다. 팬에서 10분 동안 식힌 다음 사각형으로 잘라 서빙합니다. 직접 만든 쫀득쫀득한 초콜릿 브라우니를 즐겨보세요!


Hey there, I'm excited to show you how to make delicious brownies from scratch. To make brownies, you will need the following ingredients: ●1/2 cup (1 stick) of unsalted butter ●1 cup of granulated sugar ●2 large eggs ●1 teaspoon of vanilla extract ●1/2 cup of all-purpose flour ●1/3 cup of unsweetened cocoa powder ●1/4 teaspoon of salt Here's the step by step process: 1.Preheat your oven to 350°F (175°C) and grease an 8-inch square baking pan. 2.Melt the butter in a saucepan over medium heat. Once melted, remove from heat and stir in the sugar. 3.Add the eggs and vanilla extract, and mix well. 4.In a separate bowl, combine the flour, cocoa powder, and salt. Gradually add this to the butter mixture, stirring until well blended. 5.Pour the batter into the prepared pan, and smooth the top with a spatula. 6.Bake for 18-20 minutes, or until a toothpick inserted into the center comes out clean. 7.Once done, remove from the oven and let it cool for a few minutes before cutting into squares and serving. And there you have it, a batch of delicious homemade brownies. I recommend serving them warm with a scoop of vanilla ice cream or a drizzle of chocolate sauce. Enjoy!
안녕하세요, 처음부터 맛있는 브라우니 만드는 방법을 보여드리게 되어 기쁩니다. 브라우니를 만들려면 다음 재료가 필요합니다. ●무염 버터 1/2컵(1스틱) ●과립 설탕 1컵 ●큰 계란 2개 ●바닐라 추출물 1티스푼 ●다목적 밀가루 1/2컵 ● 무가당 코코아 가루 1/3컵 ●소금 1/4티스푼 단계별 과정은 다음과 같습니다. 1. 오븐을 350°F(175°C)로 예열하고 8인치 정사각형 베이킹 팬에 기름을 바릅니다. 2. 중불에서 냄비에 버터를 녹입니다. 다 녹으면 불을 끄고 설탕을 넣고 저어줍니다. 3.계란과 바닐라 익스트랙을 넣고 잘 섞는다. 4. 별도의 그릇에 밀가루, 코코아 가루, 소금을 섞습니다. 이것을 버터 혼합물에 서서히 첨가하고 잘 섞일 때까지 저어줍니다. 5. 준비된 팬에 반죽을 붓고 주걱으로 윗면을 평평하게 해주세요. 6.18~20분간 또는 가운데에 이쑤시개를 꽂아 깨끗이 나올 때까지 굽는다. 7. 완성되면 오븐에서 꺼내 몇 분 동안 식힌 다음 사각형으로 잘라 서빙합니다. 그리고 맛있는 홈메이드 브라우니 한 묶음이 있습니다. 바닐라 아이스크림 한 스쿱 또는 초콜릿 소스 이슬비와 함께 따뜻하게 제공하는 것이 좋습니다. 즐기다!



Formal
Prompt: write a paragraph on the topic of cellular automata in a formal style

Cellular automata is a mathematical model that deals with the study of discrete systems. The model consists of a grid of cells, each of which can be in a number of states and transitions between those states are determined by a set of rules. The behavior of cellular automata is observed over time, with the state of each cell at any given time step being dependent on the state of its neighboring cells at the previous time step. This model has been used to study various phenomena ranging from the growth of biological populations to the formation of patterns in physical systems. It has also found applications in fields such as computer science, cryptography, and artificial intelligence. Cellular automata has been a subject of extensive research and continues to be an area of active investigation in the fields of mathematics and computational science.
셀룰러 오토마타는 이산 시스템 연구를 다루는 수학적 모델입니다. 모델은 셀 그리드로 구성되며 각 셀은 여러 상태에 있을 수 있으며 이러한 상태 간의 전환은 일련의 규칙에 의해 결정됩니다. 셀룰러 오토마타의 행동은 시간이 지남에 따라 관찰되며, 주어진 시간 단계에서 각 셀의 상태는 이전 시간 단계에서 인접한 셀의 상태에 따라 달라집니다. 이 모델은 생물학적 인구 증가에서 물리적 시스템의 패턴 형성에 이르는 다양한 현상을 연구하는 데 사용되었습니다. 또한 컴퓨터과학, 암호화 및 인공지능과 같은 분야에서 응용 프로그램을 찾았습니다. 셀룰러 오토마타는 광범위한 연구 대상이었으며 수학과 전산과학 분야에서 계속해서 활발한 조사가 진행되고 있는 분야입니다.

Cellular automata is a mathematical model that simulates the behavior of complex systems through the interaction of simple, locally interacting agents. This model is used to understand and predict the behavior of a wide range of systems, including but not limited to physics, chemistry, biology and computer science. The theory behind cellular automata is based on the concept of a collection of discrete cells, which can exist in a finite number of states, and whose evolution is determined by a set of rules that depend on the state of the cells and their neighbors. Its applications are varied and range from theoretical physics, to computer science, to the modeling of complex systems.
셀룰러 오토마타는 단순하고 로컬로 상호 작용하는 에이전트의 상호 작용을 통해 복잡한 시스템의 동작을 시뮬레이션하는 수학적 모델입니다. 이 모델은 물리학, 화학, 생물학 및 컴퓨터과학을 포함하되 이에 국한되지 않는 광범위한 시스템의 동작을 이해하고 예측하는 데 사용됩니다. 셀룰러 오토마타의 이론은 유한한 수의 상태로 존재할 수 있고 셀과 그 이웃의 상태에 따라 달라지는 일련의 규칙에 의해 진화가 결정되는 별개의 셀 집합의 개념을 기반으로 합니다. 응용 분야는 다양하며 이론 물리학에서 컴퓨터과학, 복잡한 시스템 모델링에 이르기까지 다양합니다.


Informal
Prompt: write a paragraph on the topic of cellular automata in an informal style

Yo, have you ever heard of cellular automata? It's this really cool concept that simulates the behavior of complex systems by using simple agents that interact with each other. Basically, it's like a virtual world where little "cells" interact with each other to create different patterns and behaviors. It's used in all sorts of fields like physics, chemistry, biology and computer science. It's pretty wild to think about all the possibilities that cellular automata holds. It's not a widely known concept but It's pretty interesting once you learn about it.


Personas
Prompt: write a paragraph on the topic of cellular automata in the style of a social media influencer

Hey guys, have you heard about cellular automata? It's this super cool concept that simulates the behavior of complex systems through the interaction of simple agents. Think of it like a virtual world where tiny "cells" interact with each other to create amazing patterns and behaviors. It's like a mini universe in your computer! It's used in all sorts of fields like physics, chemistry, biology and computer science. It's so mind-blowing to think about all the possibilities that cellular automata holds. I'm obsessed and I think you will be too! #cellularautomata #science #futurism


Custom Text Manipulation
Prompt: write a paragraph on the history of the calculator, include emojis at the end of every sentence, and do not capitalize the first word in each sentence

the first calculator was invented by william ovens in 1642. it was called the "pascal calculator" and was designed to perform the four basic arithmetic operations. in the following centuries, many different types of calculators were invented, including the slide rule and the abacus. however, it wasn't until the 1970s that electronic calculators became widely available and affordable. the first electronic calculator was the "hp-35" which was introduced in 1972 by hewlett-packard. it was a scientific calculator that could perform trigonometric and logarithmic functions. 🤖📊 Since then, the calculator has become an essential tool in many fields, including science, engineering, and finance. 💻🧮 Today, we have a wide variety of calculators available, from basic models to advanced models with advanced features like graphing capabilities and programming languages. 📱💻


5. Media Types

Write Social Media Posts
Prompt: write a tweet on futurism "Exploring the future is both exciting

"Exploring the future is both exciting and daunting. From AI to biotech, advances in technology are shaping our world in ways we never thought possible. Let's embrace change and shape the future together. #futurism #technology #innovation"


Write Blogs
Prompt:write a blog on french cuisine

French cuisine is widely considered to be one of the most sophisticated and refined in the world. The French have a long history of culinary arts, dating back to the Middle Ages, and have developed a culinary tradition that is rich and diverse. French cuisine is characterized by its use of fresh, seasonal ingredients, expertly prepared and presented in a way that is both elegant and delicious. One of the most recognizable elements of French cuisine is the use of sauces. French sauces are typically made with a base of butter, cream, or stock, and are flavored with herbs, spices, and other ingredients. Some of the most famous French sauces include béarnaise, hollandaise, and béchamel. These sauces are often served with meats, seafood, and vegetables and are a hallmark of French cuisine. Another important element of French cuisine is the use of fresh, seasonal ingredients. The French are known for their commitment to using only the freshest and highest-quality ingredients in their cooking. This is why many French dishes are only available at certain times of the year, when the ingredients are at their peak. One of the most popular types of French cuisine is 'bistro' cuisine. These are casual, friendly restaurants that serve traditional French dishes in a relaxed atmosphere. Bistro cuisine often features simple, hearty dishes such as steak frites, mussels, and ratatouille, and is a great way to experience the flavors of France. French pastries and desserts are also famous around the world. French patisseries, or pastry shops, are known for their delicious croissants, macarons, tarts, and cakes. French desserts are often made with fresh fruits, cream, and chocolate and are a perfect way to end a meal. In conclusion, French cuisine is a true art form, known for its sophistication and refinement. The use of fresh,




Write Emails
Prompt: write an email selling software to corporate executives

Subject: Enhance Your Business Operations with Our Innovative Software Dear [Executive's Name], I hope this email finds you well. My name is [Your Name], and I am reaching out to introduce you to our cutting-edge software that has the potential to revolutionize the way your business operates. Our software is designed to streamline and automate various business processes, including but not limited to inventory management, financial forecasting, and customer relationship management. With its user-friendly interface and customizable features, it is easy for your team to navigate and tailor to your specific needs. What sets our software apart from the competition is its ability to integrate seamlessly with other systems and platforms, allowing for smooth data transfer and real-time analysis. This results in increased efficiency, reduced errors and ultimately improved decision-making capabilities. We understand that every business is unique, and that's why our team is dedicated to providing personalized solutions and exceptional customer service. We would love the opportunity to schedule a demo and show you how our software can benefit your business. Please let me know if this is something that interests you and we can schedule a time that works best for you. I look forward to hearing from you soon. Best regards, [Your Name] P.S. Our software comes with a 30-day free trial and flexible pricing options to fit any budget. Don't hesitate to reach out if you have any questions or concerns.


Write Poems
Prompt:write a poem about the soul and show rhymeand meter headings

The soul, a mystery deep and true A treasure to find, a journey to pursue Rhyme: AABB Meter: Iambic tetrameter It whispers to us, in the dark of the night Guiding us towards, the path that is right Rhyme: AABB Meter: Iambic tetrameter It is the essence, of who we truly are A light that shines, from the stars Rhyme: AABB Meter: Iambic tetrameter It is the music, that plays in our hearts The rhythm that keeps us, from falling apart


Write Songs
Prompt:write a folk song about the sunset show guitarchords

Verse 1: G D Em C The sun is setting in the sky, G D Em C Its colors painting the world so bright. G D Em C I strum my guitar, and sing this tune, G D Em C Of the sunset show, and its beauty in bloom.
Chorus: C G D Em Oh, the sunset show, it takes my breath away, C G D Em With colors of red, orange, and gold on display. C G D Em I'll sing this song, until the stars come out, C G D Em Of the sunset show, and its beauty no doubt.
Verse 2: G D Em C The sky turns dark, but the colors stay, G D Em C In the memory of all who saw this day. G D Em C I'll keep strumming, this tune I wrote, G D Em C Of the sunset show, that I'll always promote.
Chorus: C G D Em Oh, the sunset show, it takes my breath away, C G D Em With colors of red, orange, and gold on display. C G D Em I'll sing this song, until the stars come out, C G D Em Of the sunset show, and its beauty no doubt.
Outro: G D Em C So let's sit back, and enjoy the view, G D Em C Of the sunset show, and its colors so true. G D Em C I'll keep strumming, and singing this song, G D Em C Of the sunset show, and its beauty so strong.

Verse 1: G C G The sun sets in the west G C G And the sky turns a fiery red G C G I take my guitar in hand D G C G And play a song for the sunset show Chorus: C G D Oh the sunset show C G D Is a sight to behold C G D As the day comes to a close G C G I'll keep playing my guitar


Write Resumes/Cover Letters
Prompt: write a software engineer resume

[Full Name] [Address] [Phone Number] [Email Address] Summary: Highly motivated and experienced software engineer with over 5 years of experience in the industry. Strong background in full-stack development and a proven track record of delivering high-quality, reliable and maintainable code. Skilled in multiple programming languages including Java, Python, and JavaScript. Strong problem-solving and analytical skills, with a passion for learning new technologies and staying current in the field. Education: ● Bachelor of Science in Computer Science, [University Name], [Graduation Date] ● Master of Science in Computer Science, [University Name], [Graduation Date] Technical Skills: ● Programming Languages: Java, Python, JavaScript, C++, C# ● Web Development: HTML, CSS, React, AngularJS, Node.js ● Database: MySQL, MongoDB, Oracle ● Tools and Frameworks: Git, JIRA, Scrum, Agile ● Operating Systems: Windows, Linux, MacOS Professional Experience: Software Engineer, [Company Name], [Dates of Employment]

● Worked on the development of a web-based application using React, Node.js, and MongoDB. ● Collaborated with the team to design and implement new features, improve performance and fix bugs. ● Contributed to the design and implementation of the application's RESTful API. ● Assisted with the deployment of the application to the production environment. ● Participated in code reviews and provided feedback to other team members. ● Mentored junior developers and provided guidance on best practices. Software Developer, [Company Name], [Dates of Employment] ● Worked on the development of a mobile application using Java and Android SDK. ● Collaborated with the team to design and implement new features, improve performance and fix bugs. ● Assisted with the deployment of the application to the Google Play Store. ● Participated in code reviews and provided feedback to other team members. Additional Experience: ● Participated in various hackathons and coding competitions, winning several awards for creative and efficient solutions. ● Contributed to open-source projects on Github. Certifications: ● Oracle Certified Professional, Java SE 11 Developer ● Microsoft Certified: Azure Developer Associate ● AWS Certified Developer - Associate Professional Affiliations: ● Member of [Professional Organization Name] ● Member of [Professional Organization Name] Languages: English (native) Spanish (fluent)


6. Meta ChatGPT Ask ChatGPT About Its Own Capabilities Prompt:what ways can you structure text output? As a language model, I can structure text output in a variety of ways, including: ●Paragraphs: I can generate text in the form of multiple paragraphs, which can be useful for longer pieces of writing such as articles or essays. ●Lists: I can generate lists of items, which can be useful for creating bullet points or enumerated lists. ●Headings and subheadings: I can use headings and subheadings to organize text and make it easier to read. ●Sentences: I can generate text one sentence at a time, which can be useful for generating responses to specific questions or prompts. ●Tables: I can generate table format text, which can be useful for displaying data in a structured format. ●Poems and songs: I can structure text in rhyme and meter, which can be useful for generating creative writing like poems and songs. ●Dialogue: I can generate text in the form of a conversation or dialogue between two or more characters. These are just a few examples of the ways that I can structure text output. The specific structure will depend on the prompt and task at hand.


7. Expert Prompting You can use several conditions discussed in this cheat sheet to prompt ChatGPT to obtain more granular output. Prompt ChatGPT with Simultaneous Conditions Prompt:write a paragraph on machine learning, inthe 1st person, with emojis, with headings and subheadings 🤖 Hi there! I'm excited to talk to you about machine learning. 🧠 Machine learning is a type of artificial intelligence (AI) that allows systems to learn and improve from experience without being explicitly programmed. 📈 It is used in a wide range of applications, such as image recognition, natural language processing, and predictive analytics. 🤖 Here's a breakdown of the different types of machine learning: 1.Supervised Learning ● 🏫 In this type of learning, the model is trained on a labeled dataset, where the correct output is already known. ● 💻 Common examples include identifying objects in images or spam emails. 2.Unsupervised Learning ● 󰡷 In this type of learning, the model is not given any labeled data and must find patterns and relationships on its own. ● 🤖 Common examples include clustering data or identifying patterns in customer behavior. 3.Reinforcement Learning ● 🎮 In this type of learning, the model is rewarded for certain actions and learns through trial and error. ● 🤖 This is often used in gaming and robotics. 🤖 I hope this gives you a better understanding of machine learning and its different types! Let me know if you have any questions.


what is the best aromatherapy

The best aromatherapy depends on the individual and their personal preferences and needs. Some popular and commonly used essential oils in aromatherapy include lavender, eucalyptus, peppermint, lemon, and bergamot, each with its own unique properties and benefits. For example, lavender is known for its calming and relaxing properties, eucalyptus can help with respiratory issues, and peppermint can provide a refreshing and invigorating experience.
It's important to remember that not all essential oils are suitable for everyone, and some can cause adverse reactions or interact with medications. It's also crucial to use high-quality, pure essential oils and to dilute them properly before use. It's best to consult with a qualified aromatherapist or healthcare professional before trying aromatherapy to determine the best approach for your specific needs and to avoid any potential health risks.
최고의 아로마테라피는 개인과 개인의 선호도 및 필요에 따라 다릅니다. 아로마테라피에서 인기 있고 일반적으로 사용되는 일부 에센셜 오일에는 라벤더, 유칼립투스, 페퍼민트, 레몬 및 베르가못이 포함되며 각각 고유한 특성과 이점이 있습니다. 예를 들어, 라벤더는 진정 및 긴장을 풀어주는 특성으로 유명하고, 유칼립투스는 호흡기 문제에 도움이 되며, 페퍼민트는 상쾌하고 상쾌한 경험을 제공할 수 있습니다.
모든 에센셜 오일이 모든 사람에게 적합한 것은 아니며 일부는 부작용을 일으키거나 약물과 상호 작용할 수 있음을 기억하는 것이 중요합니다. 고품질의 순수한 에센셜 오일을 사용하고 사용하기 전에 적절하게 희석하는 것도 중요합니다. 아로마테라피를 시도하기 전에 자격을 갖춘 아로마테라피스트 또는 의료전문가와 상담하여 특정 요구 항에 가장 적합한 방법을 결정하고 잠재적 건강 위험을 피하는 것이 가장 좋습니다.

가장 좋은 아로마테라피는 개인의 취향과 필요에 따라 다릅니다. 아로마테라피에서 인기 있고 널리 사용되는 기본적인 오일은 라벤더, 유칼립투스, 페퍼민트, 레몬, 베르가못 등이 있으며, 각각은 자체 고유의 성질과 효과가 있습니다. 예를 들어, 라벤더는 진정성과 안정감이 있다는 것으로 알려져 있으며, 유칼립투스는 호흡기 문제를 해결하는 데 도움이 될 수 있습니다.
모든 기본적인 오일이 모든 사람에게 적합하지 않을 수 있으며, 어떤 오일은 부작용이나 약물과의 상호 작용을 일으킬 수 있습니다. 또한, 고품질의 순수한 기본적인 오일을 사용하고 적절하게 약화하는 것이 중요합니다. 아로마테라피를 시도하기 전에 적극적인 아로마테라피사 또는 건강 전문
 
     
     
 
아무도 알려주지 않는 ChatGPT에 대한 놀라운 사실 6가지
 bearmom  2023. 2. 4. 9:55
Mike Elgan | Computerworld


모두가 ChatGPT에 대해 이야기한다.
ChatGPT는 사람처럼 시와 산문, 심지어 소프트웨어 코드까지 쓰는 오픈AI(OpenAI) 툴이다.

ChatGPT 관련 대화는 대부분 놀라움, 두려움, 경고로 이어진다.
ChatGPT가 할 수 있는 것에 대한 놀라움, 악용 및 ‘로봇’으로의 대체 대한 두려움, 그리고 창의성을 기계에 위탁하는 인간성의 위험에 대한 경고다.
하지만 이 모든 대화는 핵심을 놓치고 있다.
사실 ChatGPT는 대부분 미디어에서 묘사하는 것보다 덜 놀랍고 덜 무서우며, 경고할 필요성도 낮다.
 
ⓒ Getty Images Bank

ChatGPT는 샌프란시스코에 소재한 오픈AI라는 회사의 AI 언어 모델이다.
텍스트를 통해 사람과 대화하고, 요청에 따라 다양한 콘텐츠를 생성한다.
GPT는 Generative Pre-trained Transformer의 약어로, 사전훈련된 생성형 변환기라는 의미다.
2020년대 들어 생성형 AI는 단연 가장 중대한 기술적 추세가 됐다.
잘 알려지지 않은 ChatGPT에 관한 6가지 사실을 살펴보면서 생성형 AI를 더 정확히 이해해 보자.


1. ChatGPT의 기능은 독보적이지 않다

ChatGPT가 특별한 단 한 가지 이유는 공개서비스이기 때문이다.
일반 사람이 접할 수는 없지만 대등한 수준, 심지어 더 뛰어난 수준의 다른 AI 프로젝트도 있다.
구글, 메타 같은 주요 기술기업, 그리고 수많은 스타트업, 대학이 ChatGPT 또는 그 이상 역량을 가진 생성형 AI 툴을 개발했다.
그중에는 보유한 AI의 제한된 버전을 출시한 경우도 있고, 아직 일반에 공개하지 않은 경우도 있다.
실제로 메타는 ChatGPT보다 먼저 #Blenderbot 이라는 #챗봇 을 출시했다.
그러나 이 봇은 너무 조심스럽게 관리된 탓에 사용자들이 재미를 느끼지 못했다.
모두 조심하는 이유는 실제 환경에서 발생한 여러 사고 때문이다.
마이크로소프트는 몇 년 전에 ‘Tay’라는 챗봇을 출시했다.
Tay는 소셜미디어 게시물을 보고 학습했는데, 온라인에서 혐오를 조장하고 악플 일삼는 사람들이 올리는 콘텐츠를 흡수한 후 인종차별적 표현과 독설을 쏟아내면서 MS를 당혹게 했다.
구글은 2022년 람다(Language Model for Dialogue Applications, LaMDA)에 “지각이 있다”라고 공개적으로 (그리고 사실과 다른) 발언을 한 엔지니어를 해고했다.
현재 ChatGPT가 IT 분야에서 큰 인기를 누리는 이유는 이 툴을 대중에 공개한 오픈AI의 대담함, 그리고 일부 민감한 주제까지 봇이 다루도록 허용한 대담함에 있다.
다른 기업은 지나치게 조심스럽다.
그러나 ChatGPT가 엄청난 관심을 끌어모은 지금, 다른 기업들도 모두가 사용할 수 있는 각자의 봇을 서둘러 준비하고 있다.
구글은 일반 대중을 위한 AI 출시를 서두르기 위해 내부적으로 ‘코드 레드’를 발령한 것으로 알려졌다.
구글은 올해 20여 개의 새로운 AI 툴을 출시할 계획이다.
중국 Baidu도 오는 3월 ChatGPT의 경쟁 툴을 출시할 예정이다.
Futurepedia라는 사이트에는 현재 사용해볼 수 있는 수백 개의 AI 툴이 정리돼 있다.
상당수가 ChatGPT를 기반으로 하지만 강력한 경쟁 기술을 기반으로 하는 툴도 있다.
ChatGPT가 독보적으로 강력한 것이 아니라, 강력한 AI 시스템 중에서 독보적으로 대담하고 접근성이 높을 뿐이다.


2. 마이크로소프트가 오픈AI를 지배한다

ChatGPT를 만든 영리 목적의 오픈AI LP라는 회사는 비영리회사인 오픈AI 주식회사(OpenAI Incorporated)의 자회사다.
그러나 오픈AI 주식회사는 오픈AI LP를 ‘소유’하고 있지는 않다.
오픈AI LP의 최대 투자자는 마이크로소프트이며, 현재 마이크로소프트는 오픈AI LP에 100억 달러를 추가로 투자하는 방안을 추진 중이다.
이렇게 되면 마이크로소프트의 지분은 49%가 되고 비영리 모회사의 소유 지분은 2%에 불과해진다.
그 외 모든 투자자들이 나머지 49%를 나눠 갖게 된다.
마이크로소프트는 빙, 애저, 파워포인트, 아웃룩 및 기타 제품에 ChatGPT 기능을 집어넣기 위한 작업도 적극 추진하고 있다.
요점은 요점은 마이크로소프트가 오픈AI를 지배한다는 것이다.

3. ChatGPT에 대한 비판 대부분은 ChatGPT에 대한 것이 아니다

ChatGPT를 둘러싼 주된 우려 3가지는
▲학생의 부정행위 가능성
▲표절 및 소셜 엔지니어링과 같은 비윤리적 용도로의 사용
▲간혹 발생하는 틀린 정보 전달이다.

일단 앞의 2가지 우려는 무시해도 된다.
ChatGPT와 같은 방식으로 작동하는 생성형 AI의 새로운 세대가 나올 때마다 AI로 생성된 콘텐츠를 감지할 수 있는 툴도 함께 나올 것이다.
ChatGPT와 같은 AI 챗봇은 절대적으로 불가피하며, 앞으로 계속 발전할 수밖에 없다.
비윤리적 사용은 우리가 늘 마주해야 할 문제다.
어떤 경우에도 인간의 도덕적 타락은 AI 툴의 잘못이 아니다.
세 번째 비판은 방향이 잘못됐다.
ChatGPT나 다른 챗봇이 틀린 정보를 전달한다면, 이유는 하나다.
틀린 정보가 포함된 학습 데이터집합이 현재의 공개 ‘베타’를 포함한 광범위한 검토 프로세스에서 포착되지 않은 것이다.
잘못된 데이터를 이유로 ChatGPT를 비난하는 것은 형편없는 영화 한 편을 보고 나서 영화라는 매체 자체를 비난하는 것과 마찬가지다.
본질과 무관하다.
미래의 기업은 자체 데이터 또는 철저한 검토를 거친 데이터를 기반으로 생성형 AI를 사용할 것이다.
즉, 기술과 데이터는 서로 별개의 조각이다.
CIA 같은 조직이 모든 첩보 보고서와 정치적 분석, 수많은 전화 통화의 녹취록을 ChatGPT와 같은 AI에 집어넣는다고 가정해 보자.
상상을 초월하는 수준으로 신속하게 통찰력을 얻는 힘을 갖게 될 것이다.
이 기술은 엄청난 잠재력을 지녔다.
그러나 지금의 ChatGPT 학습 데이터는 그 잠재력과 무관하다.

4. #ChatGPT 는 사람이 만들었다

#텍스트기반AI 를 사용하다 보면 사람이 아니라 컴퓨터의 생각과 상호작용한다고 생각하기 쉽다.
그러나 ChatGPT와 같은 프로젝트는 사람이 만들어서 사람이 검토하고 사람이 우선순위를 부여한 콘텐츠를 사람의 프로그래밍을 사용해서 긁어모은다.
따라서 AI에는 사람의 실수, 사람의 편견, 사람의 결론이 포함된다.

5. ChatGPT를 제대로 활용하려면 기술이 필요하다

사실 텍스트 입력 방식의 모든 생성형 AI 툴에서 최대한의 효과를 얻기 위해서는 숙련된 입력이 필요하다.
각각의 AI 툴에 어떻게 말을 걸어야 하는지에 관한 기술이자 과학인 ‘prompt engineering’이라는 새로운 스킬이 있다.
SEO와 비슷하게 이 분야에서는 앞으로 어떤 단어가 어떤 결과를 생산하는지에 관한 지식을 개발하고 생성형 AI 사용에 도움이 되는 툴과 컨설턴트로 구성된 전문적 산업이 등장하게 될 것이다.
‘프롬프트 엔지니어링’은 새로운 전문 분야지만, 대부분 사람이 직접 시도하게 될 분야이기도 하다.
이 ‘참고 가이드’를 활용해 시작할 수 있다.

건강전문 몰
www.dopza.com
 
돕자몰
www.dopza.com


6. 일부 산업은 이미 ChatGPT에 의존하고 있다

전문적 문서작업 중에는 AI가 하든 사람이 하든 상관없이 ‘로봇’같은, 즉 자동적이고 정형화된 유형이 있다.
단조로운 일상적 작업은 AI가 더 빠르고 값싸게 할 수 있다는 사실이 이미 발견된 직종도 있다.
ChatGPT에 이미 중독된 분야 중 하나는 부동산 업계다.
매물 등록 시 판매할 집을 설명하는 데 사용할 수 있는 매개변수는 매우 많다.
기본적 항목으로는 수치 데이터(건축 면적, 총면적, 방과 화장실 개수, 가격 등)가 있고, ‘리모델링’, ‘개방적 분위기’, ‘넓은 주방’과 같은 서술적이고 사실적인 용어도 있다.
이제 부동산 중개인은 모든 데이터를 ChatGPT에 입력하고 웹사이트에 올릴 소개 문서 작성을 AI에 맡기면 된다.
부동산 종사자에게만 특별한 것이 아니라 단지 이 업계가 흐름에 앞서 있을 뿐이다.
앞으로 많은 직업이 AI 툴에 전적으로 의존하게 될 것이다.
ChatGPT에 모두의 관심이 집중되고 있다.
그러나 아직은 생성형 AI의 초창기 중에서도 극초기다.
따라서 이 기술과 관련해 일어나고 있는 일을 제대로 이해할 필요가 있다

원문보기:
https://www.itworld.co.kr/news/276085#csidx2a4b8f39af840dfb26ff4ec43f76a31 존재하지 않는 이미지입니다.
 
     
     
 
Prompt engineering
 bearmom  2023. 2. 4. 0:01

From Wikipedia, the free encyclopedia


프롬프트 엔지니어링은 인공지능, 특히 자연어처리(NLP)의 개념입니다.
프롬프트 엔지니어링에서 작업설명은 암시적으로 제공되는 대신 질문과 같이 입력에 포함됩니다.
프롬프트 엔지니어링은 일반적으로 하나 이상의 작업을 프롬프트 기반 데이터세트로 변환하고 "즉시 기반 학습" 또는 "즉시 학습"이라는 언어모델을 교육하는 방식으로 작동합니다.[1][2]
프롬프트 엔지니어링은 "접두사 튜닝" 또는 "프롬프트 튜닝"이라는 것과 함께 프롬프트의 표현만 학습되는 대규모 "고정 frozen" 사전훈련된 언어모델에서 작동할 수 있습니다.[3][4]
Prompt engineering is a concept in artificial intelligence, particularly natural language processing (NLP). In prompt engineering, the description of the task is embedded in the input, e.g., as a question instead of it being implicitly given. Prompt engineering typically works by converting one or more tasks to a prompt-based dataset and training a language model with what has been called "prompt-based learning" or just "prompt learning".[1][2] Prompt engineering may work from a large "frozen" pretrained language model and where only the representation of the prompt is learned, with what has been called "prefix-tuning" or "prompt tuning".[3][4]


GPT-2 및 GPT-3 언어 모델[5]은 신속한 엔지니어링에서 중요한 단계였습니다.
2021년에는 여러 NLP 데이터세트를 사용한 멀티태스킹 프롬프트 엔지니어링이 새로운 태스크에서 좋은 성능을 보였습니다.[6]
퓨샷 학습 예제에 일련의 생각을 포함하는 프롬프트는 언어모델에서 추론을 더 잘 나타냅니다.[7]
제로 샷 학습에서 생각의 사슬을 조장하는 프롬프트에 텍스트를 추가하면(예: "단계적으로 생각하자") 다단계 추론 문제에서 언어 모델의 성능을 향상시킬 수 있습니다.[8]
이들 도구의 광범위한 접근성은 이미지합성을 위한 여러 오픈소스 노트북 및 커뮤니티주도 프로젝트 출판에 의해 주도되었습니다.[9]
프롬프트 처리에 대한 설명에서는 2022년 2월에 약 170개의 데이터세트에 대한 2,000개 이상의 공개 프롬프트를 사용할 수 있다고 보고했습니다.[10]
The GPT-2 and GPT-3 language models[5] were important steps in prompt engineering. In 2021, multitask prompt engineering using multiple NLP datasets showed good performance on new tasks.[6] Prompts that include a chain of thought in few-shot learning examples show better indication of reasoning in language models.[7] In zero-shot learning prepending text to the prompt that encourages a chain of thought (e.g. "Let's think step by step") may improve the performance of a language model in multi-step reasoning problems.[8] The broad accessibility of these tools were driven by the publication of several open-source notebooks and community-led projects for image synthesis.[9]
A description for handling prompts reported that over 2,000 public prompts for around 170 datasets were available in February 2022.[10]

2022년에는 DALL-E, Stable Diffusion, Midjourney와 같은 기계학습 모델이 공개되었습니다.
이러한 모델은 텍스트 프롬프트를 입력으로 사용하여 이미지를 생성하는 데 사용하며, 이는 텍스트 대 이미지 프롬프트와 관련된 새로운 범주의 프롬프트 엔지니어링에 영향을 미쳤습니다.[11]

In 2022, machine learning models like DALL-E, Stable Diffusion, and Midjourney were released to the public. These models take text prompts as input and use them to generate images, which effected a new category of prompt engineering related to text-to-image prompting.[11]
반응형

댓글