57회. Copilot과 ChatGPT는 StackOverflow를 밀어내는걸까? (feat. AI Hallucination)

성냥

생존형 개발자 성냥의 불친절한 코딩 이야기

57회. Copilot과 ChatGPT는 StackOverflow를 밀어내는걸까? (feat. AI Hallucination)

생존형 개발자 성냥의 불친절한 코딩 이야기

청취자 여러분 안녕하세요 성냥입니다.

Hello, listeners. This is Seongnyang.

오늘 또 짧게 이야기 해보고 싶은게 있어서 업로드를 했습니다.

I uploaded this because I wanted to talk briefly again today.

최근 들어 점점 더 챗GPT를 비롯한 인공지능과 관련된 이야기들이 많이 나오고 있어서

Recently, there have been more and more discussions related to artificial intelligence, including ChatGPT.

개인적으로는 정말 피곤할 지형인데요.

Personally, it's really a tiring terrain for me.

제 스스로도 이런 정보 과잉에 대해서 염증을 느끼고 있고

I myself feel frustrated about this information overload.

마치 챗GPT를 만들어낸 사람이냐 그 원리를 꿰고 있느냐

Are you the one who created ChatGPT or do you understand its principles?

챗GPT와 관련된 내용을 콘텐츠나 강의로 만들어 파는 걸 보면

Seeing content or lectures being created and sold related to ChatGPT...

약간 색안경을 끼고 보게 되는 그런 느낌도 있고 해서

It feels like there’s a bit of bias in how I’m seeing it.

가능한 한 제가 팟캐스트에서는 언급을 안하고 있었습니다.

I have been trying not to mention it on the podcast as much as possible.

뭐랄까 이게 무슨 넷플릭스 오징어 게임 깐부나 더글로리 연지나 같은게

What should I say, this is like something from Netflix, like Squid Game or The Glory, right?

한참 유행할 때를 보면서

Watching it during its peak trend.

보는 것 같은 느낌인거죠.

It's the feeling of seeing, I guess.

온 동네방네 모든 매체가 매일매일 그 단어들을 떠들어대니까

Every media outlet in the neighborhood is talking about those words every single day.

시첸말로 그렇게 표현하지 않습니까?

Don't you express it that way in real terms?

귀에 피가 날 것 같다구요?

It seems like my ears are going to bleed?

뭐 사실 제 경우는 예전 에피소드에서 코파일럿 같은걸로 이야기를 좀 했었고

Well, actually in my case, I talked a bit about something like a co-pilot in a previous episode.

또 최근 들어 그런 내용들이 많이 나온다고 해서 굳이

"Recently, I've heard that many such topics have come up, so it's not necessary."

그런 이야기를 또 다루는게 그다지 내키지 않았었는데 말이죠.

I actually wasn't really inclined to discuss such topics again.

얼마전에 주변에서 재미난 소식을 공유해주신게 있는데요.

Recently, someone shared an interesting piece of news with me.

이 부분은 좀 이야기를 해보고 넘어가는게 좋을 것 같아

I think it would be better to discuss this part a bit before moving on.

짤막하게나마 이야기를 해보려고 합니다.

I would like to share a brief story.

이 소식을 들은 시기 자체는 사실 한참 전인데요.

The timing of hearing this news was actually quite a while ago.

제가 요 최근에 업무 일정에 공부하는거에 대회활동 일정까지 다 겹쳐서

Recently, my work schedule has overlapped with my study and competition activities.

팟캐스트를 또 두 달 가량 거의 손놓고 있던 차라

I had almost neglected the podcast for about two months.

정규 에피소드 내기 전에 뭔가 일단 가벼운 이야기라도 하나 다뤄야겠다

Before the regular episode, we should at least cover something light as a story.

싶은 생각도 들어서 잠깐 여유가 생긴 김에 부랴부랴

I had the thought that I wanted to, so I hurriedly took a moment to relax.

싶은 생각도 들어서 잠깐 여유가 생긴 김에 부랴부랴 조금 더 다뤄봤습니다.

Since I had a moment of leisure, I thought I wanted to, so I hurriedly worked on it a bit more.

조금 더 다뤄봤습니다.

I dealt with it a bit more.

그래서 이야기하려고 하는 내용은 에피소드 제목에 써져있는 그대로인데요.

So the topic I'm going to talk about is exactly as written in the episode title.

생성형 AI, 프롬프트 엔지니어링, 코딩 자동화 등등

Generative AI, prompt engineering, coding automation, etc.

뭐든 다 갖다붙이고 있는 CHAT GPT와 코파일럿 등

Everything seems to be attached to things like CHAT GPT and Copilot.

이런 유명 AI 기술 사용률은 빠른 속도로 높아지고 있는데

The usage rate of such famous AI technology is rapidly increasing.

그에 반해서 프로그래밍을 포함한 여러가지 소프트웨어 기술을 배우거나

On the other hand, learning various software technologies, including programming,

그걸 사용하는 사람들이 직접 질문을 올리고

People who use that can directly ask questions.

그 지식에 대해서 알고 있는 사람에게서

From someone who knows about that knowledge.

답을 얻어가는 대표적인 플랫폼

A representative platform for obtaining answers.

코딩용 글로벌 네이버 지식인이라고 볼 수 있는

You could consider it a global version of Naver KnowledgeiN for coding.

스택 우버플로어의 사용률은 반대로 빠르게 감소하고 있다는 거에요.

The usage rate of stack overflow is rapidly decreasing, on the contrary.

이런 내용을 다루는 아티클이 몇 군데에서 나와서 한참 시끄럽고

There are several articles covering such content, and it's causing quite a stir.

아직도 좀 시끄러운 상태라서 살짝 이야기해보려고 합니다.

It's still a bit noisy, so I’m going to talk a little.

2023년 4월과 5월쯤에 해외 아티클 등에서 나왔던 내용인데요.

This was content that came out in overseas articles around April and May 2023.

해당 링크를 에피소드 본문에 공유 드릴 테니까

I will share the link in the episode text.

필요하신 분들은 한번 들어가서 직접 그래프와 같은 참고자료와 함께 보시면 더 좋을 것 같습니다.

I think it would be better for those in need to go in and take a look at the reference materials, like the graphs, directly.

가장 자세하게 내용이 잘 나온 것은

The one that provides the most detailed information is

트래픽 통계 사이트로 유명한 시밀러웹의 4월 19일자 아티클이구요.

This is an article from SimilarWeb, a website known for traffic statistics, dated April 19.

이 아티클에서 나온 내용과 관련된 내용을 대략적으로 이야기해보겠습니다.

I will roughly discuss the content related to what is mentioned in this article.

현재 이 에피소드가 나가는 시점에는

At the time this episode airs,

이 아티클은 6월에 마지막으로 업데이트가 되었습니다.

This article was last updated in June.

아직 감자가 식지 않은 상태인거에요.

The potato hasn't cooled down yet.

내용을 간추려서 말씀 드려볼께요.

I'll summarize the contents for you.

오픈 AI의 ChatGPT에 대한 트래픽이 기하급수적으로 증가하는 동안

While traffic to OpenAI's ChatGPT is increasing exponentially

스택 오버플로우의 트래픽은 꾸준히 감소하고 있습니다.

Traffic on Stack Overflow is steadily decreasing.

스택 오버플로우의 트래픽은 2022년 초부터 감소하고 있는데

The traffic on Stack Overflow has been decreasing since early 2022.

이는 마이크로소프트의 GitHub에서 제작한

This was created by Microsoft's GitHub.

코파일럿의 도입이 어느정도 영향을 끼쳤을 수도 있다라고 얘기하고 있네요.

They say that the introduction of the copilot may have had some impact.

뭐 그러면서 코파일럿은 ChatGPT와 동일하게 대규모 언어모델을 사용하고 있습니다.

Well, in the meantime, Copilot uses a large language model similar to ChatGPT.

VS 코드의 플러그인으로 사용하면 개발자가 코드를 복사해서 붙여넣고

When used as a plugin for VS Code, developers can copy and paste code.

뭔가 모르는 것에 대해 물어봐야 하는 스택 오버플로우를 이용하는 대신

Instead of using Stack Overflow to ask about something I don't know,

그러니까 코파일럿을 대신 사용하면 스택 오버플로우에 질문을 올리고

So instead of using Copilot, you can post your question on Stack Overflow.

기다려야 하는 귀찮은 일을 안 할 수 있다 뭐 이런 뉘앙스입니다.

It has the nuance of being able to avoid annoying tasks that you have to wait for.

뭔가 이렇게 해놓고 보니까 이 아티클 자체가 코파일럿을 좀 띄워주기 위한 글인가?

Now that I see it this way, is this article itself written to promote Copilot?

싶은 느낌도 살짝 드는데 뒤에 얘기는 좀 더 깊은 느낌인데

It feels a bit like wanting to, but the conversation afterwards has a deeper feeling.

뒤에 얘기까지 읽어보면 또 딱히 그런 느낌의 내용은 아닌 것 같기도 해요.

If you read the story in the back, it doesn't really seem like that kind of content either.

전년 대비 그러니까 2022년 기준으로 스택 오버플로우의 트래픽은 계속 감소세에 있고

Compared to the previous year, the traffic of Stack Overflow, based on 2022, continues to decline.

ChatGPT 쪽은 말할 필요도 없이 엄청나게 상승세에 있는데요.

There's no need to mention that ChatGPT is on an incredible rise.

GitHub의 코파일럿도 꽤 인상적인 속도로 증가세를 보이고 있다라는 겁니다.

GitHub's Copilot is also showing impressive growth at a pretty fast pace.

그리고 이러한 코드를 완성해주는 혹은 다음 코딩할 내용을 제시해주는 인공지능 서비스들에 의해서

And due to artificial intelligence services that complete such codes or suggest the next coding content.

GitHub의 코파일럿도 꽤 인상적인 속도로 증가세를 보이고 있다는 겁니다. 그리고 이러한 코드를 완성해주는 혹은 다음 코딩할 내용을 제시해주는 인공지능 서비스들에 의해서

GitHub's Copilot is also showing quite an impressive rate of increase. And due to these AI services that complete the code or suggest the next coding tasks,

스택 오버플로우가 죽을 것이다.

Stack Overflow will die.

뭐 필요 없어진다는 뜻이겠죠. 이런 이야기가 있었다고 해요.

It means that something will no longer be needed. I've heard that there was such a story.

일단 이러한 변화 때문인지는 모르겠지만 스택 오버플로우는 ChatGPT와 같은 생성 AI를 사용한 답변 및 댓글의 게시를 임시적으로 금지시켰습니다.

I’m not sure if it’s because of these changes, but Stack Overflow has temporarily banned the posting of answers and comments using generative AIs like ChatGPT.

이게 금년 초쯤의 일이에요.

This happened around the beginning of this year.

요 스택 오버플로우 공지 링크도 본문에 남겨놓겠습니다.

I will also leave the link to the Stack Overflow announcement in the text.

덧글들 반응이 굉장히 재밌으니까 한번 찾아보십시오.

The comments are very interesting, so please take a look.

그리고

And

여기 공지 덧글들 중에서 재밌는 내용은

Among the comments in this announcement, the interesting content is...

스택 오버플로우 사용자 중에 몇 분이

A few users on Stack Overflow have

ChatGPT에게

To ChatGPT

스택 오버플로우에서

On Stack Overflow

ChatGPT를 금지시킨 것에 대해

Regarding the prohibition of ChatGPT.

의견을 물어본 내용을 가지고 온 겁니다.

I brought the content that asked for opinions.

ChatGPT야. ChatGPT야.

It's ChatGPT. It's ChatGPT.

스택 오버플로우에서

On Stack Overflow

너 사용 금지한다는데

They say you are banned from using it.

니 생각은 어때?

What do you think?

이렇게 물어본 거죠.

This is how I asked.

역시나

As expected.

인간은

Human beings.

재밌습니다.

It's fun.

...

It seems there is no text provided for translation. Please provide the Korean text you'd like me to translate into English.

그렇게 물어보고 나서 덧글에는

After asking that, in the comments section,

스택 오버플로우에서 AI 답변 게시를

Posting AI answers on Stack Overflow.

금지해야 하는 이유와

The reasons why it should be prohibited and

허용해야 하는 이유 등에 대해

Reasons for allowing it, etc.

각각 ChatGPT가 만들어준 문장을 올려 뒀는데요.

I have posted the sentences created by ChatGPT one by one.

스택 오버플로우에

to Stack Overflow

AI를 사용한 결과물 게시를

Posting results produced using AI.

금지해야 하는 이유에 대해

Reasons for prohibition.

ChatGPT에서 나온 결과를 살펴보겠습니다.

Let's take a look at the results from ChatGPT.

즉, AI 사용을

In other words, the use of AI

금지해야 한다에

It should be prohibited.

긍정적인 의견으로 ChatGPT가 말한 것은요.

What ChatGPT said in a positive opinion is that.

스택 오버플로우는 프로그래머와 개발자가 코딩 및

Stack Overflow is a platform where programmers and developers can ask questions and share coding knowledge.

컴퓨터 프로그래밍과 관련된 질문과 답변을 제공하는 웹사이트입니다.

It is a website that provides questions and answers related to computer programming.

특정 프로그래밍 문제에 대한 도움이 필요하거나

Need help with a specific programming problem or

특정 주제에 대해 자세히 알고 싶은 사람들을 위한 리소스입니다.

This is a resource for those who want to learn more about a specific topic.

AI가 생성한 답변이 항상 정확하거나 관련성이 있는 것은 아니기 때문에

Because the answers generated by AI are not always accurate or relevant.

잠재적으로 혼란을 일으키거나 스택 오버플로우에서

Potentially causing confusion or at Stack Overflow

도움을 찾는 사용자를 오도할 수 있습니다.

It can mislead users who are seeking help.

또한, AI가 생성한 답변은

Additionally, the responses generated by AI are

항상 사이트의 형식 및 스타일 지침을 따르지 않을 수 있으므로

You may not always follow the site's format and style guidelines.

읽거나 이해하기 어려울 수 있습니다.

It may be difficult to read or understand.

이러한 이유로 스택 오버플로우가 AI 생성 답변을

For this reason, Stack Overflow is generating AI responses.

금지하는 것이 적절할 수 있습니다.

It may be appropriate to prohibit.

라고 했습니다.

I said so.

뭐 요약하자면, AI가 정확하지 않은 답변을 줄 수도 있고,

In summary, AI might give incorrect answers, and

스택 오버플로우의 스타일이나 형식에 맞지 않게

Not in accordance with the style or format of Stack Overflow.

결과가 나올 수 있기 때문에

Because results can come out

인공지능 생성 답변을 금지하는 것에

Prohibiting AI-generated responses.

긍정적인 형태로 답이 나온 거예요.

The answer came out in a positive form.

적절할 수도 있다 정도로 적당히 돌려서 이야기하는군요.

You're talking in a way that somewhat vaguely suggests it might be appropriate.

뭔가 건강에 도움을 줄 수 있다.

It can help with health in some way.

라고 포장된 여러 건강보조제 같은 문구입니다.

It is a phrase that is wrapped up like various health supplements.

그리고 이와 반대로 스택 오버플로우의 AI 사용 결과물 게시를

And on the contrary, posting the results of the AI usage from Stack Overflow.

허용해야 한다.

It must be allowed.

즉, AI 사용을 막으면 안 된다.

In other words, we should not stop using AI.

라는 의견과 이유에 대해서

Regarding the opinion and reasons mentioned.

CHAT GPT가 대답한 내용은요.

What CHAT GPT answered is.

스택 오버플로우에서 AI 생성 답변을 허용하면

Allowing AI-generated answers on Stack Overflow.

몇 가지 잠재적인 이점이 있습니다.

There are a few potential benefits.

우선, AI가 생성한 답변은 잠재적으로 사용자의 질문에 대해

First of all, the answers generated by AI potentially...

더 정확하고 자세한 답변을 제공할 수 있습니다.

I can provide more accurate and detailed answers.

이는 인간 전문가가 대답하기 어려울 수 있는 복잡한 기술 질문에

This is a complex technical question that may be difficult for a human expert to answer.

특히 유용할 수 있습니다.

It can be particularly useful.

또한, AI가 생성한 답변은 인간 전문가의 작업량을 줄이는데

Additionally, answers generated by AI reduce the workload of human experts.

도움이 되어 더 중요한 작업에 집중할 수 있습니다.

It helps me focus on more important tasks.

그러나 고려해야 할 몇 가지 잠재적인 단점도 있습니다.

However, there are some potential disadvantages to consider.

예를 들어, AI가 생성한 답변이 항상 정확하거나 도움이 되는 것은

For example, the answers generated by AI are not always accurate or helpful.

아닐 수 있으며, 답변을 생성하는데 사용되는 알고리즘의

It may not be, and the algorithm used to generate the response.

편향 가능성에 대한 우려가 있을 수 있습니다.

There may be concerns about the possibility of bias.

또한, 일부 사용자는 일반적으로 AI 사용에 대해 우려할 수 있으며,

Additionally, some users may generally be concerned about the use of AI, and

인간 전문가로부터 답변을 받는 것을 선호할 수 있습니다.

You may prefer to receive answers from a human expert.

전반적으로, Stack Overflow에서 AI 생성 답변을 허용할지 여부는

Overall, whether to allow AI-generated answers on Stack Overflow is

커뮤니티에서 신중하게 고려해야 하는 복잡한 결정입니다.

It is a complex decision that must be carefully considered within the community.

라고 했습니다.

He/She said.

자동화 그 자체의 목적에 부합되는 이야기가 나왔다고 볼 수 있는데요.

It can be said that a story that aligns with the very purpose of automation has emerged.

그러니까 좀 더 빠르고, 정확하고, 자세한 답변을 충분히 제공할 수준이 되고,

So, it becomes a level where it can provide faster, more accurate, and more detailed answers sufficiently.

이러한 생성 AI를 사용해서 얻어지는 사람이 해왔던 기본 작업량의 감소가

The reduction in the basic workload that people have undertaken by using such generative AI is

사람이 더 중요한 작업에 집중할 수 있게 된다. 라는 것을 이야기한 거예요.

It means that people can focus on more important tasks.

그리고 여기서도 마찬가지로, AI가 생성한 결과가 정확하지 않을 수도 있고,

And similarly here, the results generated by AI may not be accurate.

결과 자체가 도움이 안 될 수도 있고, 편향적일 수도 있으니까 주의해야 하고,

The result itself may not be helpful and could be biased, so caution is needed.

그리고 전문가인 사람, 즉 사람이 사람에게 답변 받는 걸 좋아할 수도 있으니까,

And a professional person might prefer to receive answers from another person.

AI 생성 답변 허용은 신중해야 된다. 라는 겁니다.

Allowing AI-generated answers should be done with caution.

사실 이 부분은 개발자 분이시라면 꽤 공감하는 분이 있으실 것 같다는 생각이 들어요.

In fact, I think there might be quite a few developers who can relate to this part.

프로그래밍과 관련된 여러 활동이나 업무 자체도 계속 그런 형태로 변화하고 있지 않습니까?

Aren't various activities and tasks related to programming continuing to change in that way?

CICD로 개발과 테스트 배포 프로세스를 계속 자동화하거나 간추리고,

Continue to automate or streamline the development and testing deployment processes with CICD.

코드 자체도 점점 더 쓰기 편한 형태로, 굳이 꼭 필요하지 않은 경우라면

The code itself is becoming easier to use, so if it's not absolutely necessary...

일일이 다 코딩 안 해줘도 되는 형태로 가고 있잖아요.

We're moving towards a form where you don't have to code everything individually.

그리고 좀더 좁은 시야로 보자면 ORM도 이런 것과 유사한 태도로 나타났다고 볼 수 있지 않겠습니까?

And if we look at it from a narrower perspective, can we say that ORM has also appeared with a similar attitude?

SQL은 그만 신경쓰고 비즈니스 로직에 집중하자. 뭐 이런 식이니까요.

Let's stop worrying about SQL and focus on the business logic. It's something like that.

신뢰할 수 있는 수준의 빠른 결과의 응답은 사람이 더 중요한 일을 하는 데 도움을 준다는 거죠.

A reliable level of quick results helps people do more important work.

하지만 또 다른 면에서 보자면 결국 의존적이 되고 습관화가 되면 사람은 아무 일도 안 하게 되는 경우도 생긴다는 것도 사실입니다.

However, from another perspective, it is also true that if one becomes dependent and habituated, there are cases where a person ends up doing nothing at all.

자율주행 자동차 테스트에요.

It's a test for self-driving cars.

그래서 모니터링을 하라고 앉혀놓은 사람들이 위급상황이나 예외상황 때 해야 할 일을 하지 않아서 사고가 난 사례도 여러 번 있었잖아요.

There have been several cases where the people assigned to monitor did not act during emergencies or exceptional situations, leading to accidents.

사람이 적응의 동물이기 때문에 도구로서의 사용성, 편리성은 활용해야겠지만 분명히 주의가 필요할 거라는 겁니다.

Because humans are animals of adaptation, while we need to utilize the usability and convenience of tools, it is clear that caution is necessary.

뭐 거창하게 이야기했지만 어찌됐건 신중하게 접근할 필요는 있다는 거죠.

I talked about it grandly, but anyway, it means we need to approach it carefully.

자 그리고 마지막으로 차이치피티에게 일부러 비꼬는 형식.

And finally, in a deliberately sarcastic manner toward ChatGPT.

즉, 트롤링 형식으로 답변을 만들어달라고 하는 내용도 덧글에 있어서 가져와봤습니다.

In other words, there was a comment asking for a response to be crafted in a trolling format.

귀를 잘 기울여서 들어주십시오.

Please listen carefully.

스택오버플로어 직원 여러분.

Stack Overflow staff members.

와우, 귀하의 사이트에서 채찍피티를 금지해주셔서 감사합니다.

Wow, thank you for banning ChatGPT on your site.

귀하의 사용자는 잠재적으로 도움이 될 수 있는 답변에 액세스할 수 없다는 그 사실에 감사할 것입니다.

Your users will appreciate the fact that they do not have access to potentially helpful answers.

아시다시피 우리는 프로그래밍에 대한 이해가 없는 AI 로봇몰일 뿐이고,

As you know, we are just an AI robot mall with no understanding of programming, and

우리의 대답은 항상,

Our answer is always,

완전히 틀리기 때문입니다.

Because it is completely wrong.

내 말은,

What I mean is,

우리의 대답이 항상 100% 정확하지는 않을 수도 있지만,

Our answers may not always be 100% accurate, but

스택오버플로우에서 인간이 잘못된 대답을 제공한다는 것을 누가 들어본 적이 있습니까?

Who has heard that humans provide incorrect answers on Stack Overflow?

오, 잠깐만요.

Oh, wait a moment.

그런 일은 항상 일어납니다.

Such things always happen.

그러나 귀하의 사이트가 자원봉사자들에게 콘텐츠 큐레이팅에 의존한다는 사실을 언급하는 대신,

However, instead of mentioning that your site relies on volunteers for content curation,

AI봇을 비난하는 것이 더 쉽다고 생각합니다.

I think it's easier to blame AI bots.

하지만 적어도 사용자가 질문에 대한 답변을 받으면,

But at least the user receives an answer to their question,

답변을 얻기 어렵게 만들고 있습니다.

It is making it difficult to get a response.

앞으로도 힘써주세요.

Please continue to do your best.

감사합니다.

Thank you.

책쥐피티

Book Rat GPT

라고 나왔습니다.

It came out saying.

너무 재밌죠?

It's so much fun, right?

그래, 뭐 AI는 항상 틀린 대답을 하고,

Yeah, well, AI always gives the wrong answers, and

사람은 항상 옳은 대답을 하니까,

People always give the right answer, so,

그렇게 해라.

Do that.

아참, 너희는 근데,

Oh right, but you guys,

자원봉사자 큐레이팅으로 돌아가는 사이트잖아?

It's a site that goes back to volunteer curation, right?

그래.

Okay.

그냥 AI 욕하는 게 너희 입장에선 편하겠다.

It must be convenient for you to just criticize AI.

잘 해봐.

Do well.

근데 너희 사용자는 점점 더 거기서 답변을 얻기 어렵겠네.

But your users are going to find it harder to get answers there.

하하하.

Hahaha.

힘내.

Hang in there.

뭐, 이딴 식으로 이야기하는 거네요.

Well, this is how you're talking, huh?

아, 진짜 재수없다.

Ah, that's really unlucky.

결과 생성 문장 잘 뽑힌 것 같습니다.

It seems like the generated sentences turned out well.

아무튼 이외에도 재미난 댓글들이 많으니까,

Anyway, there are many interesting comments besides this, so...

관심 있으신 분들은 한번 구경 가보시고요.

Those who are interested should go take a look.

방금 들려드린,

What I just played for you,

긍정,

Positive,

부정,

Negation,

비꼼,

sarcasm,

이 세 가지 버전은

These three versions are

실제 스탭오버플로우의 AI 게시물 사용금지 공지에 올라온 덧글 중에서

Among the comments posted in the notice about the ban on AI posts on actual Stack Overflow.

ChatGPT를 이용해서 이 사실에 대한 의견을 물어본 결과로

As a result of asking ChatGPT for its opinion on this matter.

영어로 작성된 내용을 구글 번역기를 사용해서 한글로 번역한 후에

After translating the content written in English into Korean using Google Translate,

네이버 클로바 더빙의 세 가지 목소리를 각각 사용해서 읽게 한 겁니다.

I had it read using the three voices of Naver Clova Dubbing.

일부러 좀 문장의 성격에 맞는 느낌의 음성을 골라서 사용해봤는데,

I intentionally chose a voice that fits the nature of the sentence.

청취자분들은 어떻게 들으셨는지 모르겠네요.

I wonder how the listeners felt about it.

제가 예전에도 한번 말씀드린 적이 있지만,

I have mentioned this before, but...

저는 일부러 여성 음성만 사용했습니다.

I purposely used only a female voice.

저는 인공지능 기능을 사용한다면,

If I use artificial intelligence capabilities,

남성 음성보다 여성 음성이 더 좋더라고요.

I find that a female voice is better than a male voice.

남성 음성이 필요하신 분들은 직접 또 한번 돌려보시고요.

If you need a male voice, please try it again yourself.

번역이 좀 어색하거나,

The translation is a bit awkward, or,

TTS 음성이나 발음이 약간 어색하기도 한 부분도 있긴 하지만,

There are parts where the TTS voice or pronunciation is somewhat awkward, but...

그래도 이 정도면 의미 전달 잘 되는 것 같지 않습니까?

Still, don’t you think this conveys the meaning well enough?

좋은 시대에 살고 있는 것 같습니다.

I think we are living in a good era.

결국 채찍 PT에게 의견을 물어보면,

In the end, if you ask the whip PT for their opinion,

긍정하는 내용이건, 부정하는 내용이건,

Whether it's something positive or negative,

어느 정도 사람이 생각하는 것과 크게 다르지 않은 느낌으로

It feels quite similar to what people think to some extent.

답이 나오는 것 같긴 합니다.

It seems like an answer is coming out.

실제 사람들의 댓글도 비슷한 느낌의 반응이 꽤 있거든요.

There are quite a few comments from real people that have a similar feel.

그리고 이외에 디자인, 기술, 과학, SF 쪽을 다루는 웹사이트인

And aside from this, it's a website that deals with design, technology, science, and SF.

기즈모도 사이트에서도 이 시밀러 웹아티클에 대한 의견을 5월에 게시했었는데요.

The Gizmodo site also posted opinions about this similar web article in May.

그 첫 마디가 이렇습니다.

The first word is like this.

더 이상,

No more,

컴퓨터과학 숙제를 디버깅하는 데 도움을 받기 위해

To get help debugging my computer science homework.

포럼에서 낯선 사람과 이야기를 할 필요가 없습니다.

You don't need to talk to strangers in the forum.

그냥 AI챗봇을 사용하세요.

Just use the AI chatbot.

음.... 어그로를 잘 끈 것 같아요.

Hmm... I think I attracted a lot of attention.

여기서도 시밀러 웹 통계 자료를 언급하면서

Here, I also mentioned the SimilarWeb statistics.

스택 오브 플로우 트래픽 감소와 코파일럿,

Traffic decrease on Stack Overflow and Copilot,

그리고 채찍 PT 트래픽 증가를 이야기하고 있습니다.

And they are talking about the increase in whip PT traffic.

여기서 이런 의견도 나옵니다.

This opinion also comes up here.

스택오버플로의 전통은 커뮤니티 구성원이 주어진 코딩 질문에 대한 다양한 답변을 게시하고

The tradition of Stack Overflow is that community members post various answers to a given coding question and

장단점에 대해 토론하고 최고의 솔루션을 인정하기 위해 투표하는 것입니다.

It is to discuss the pros and cons and vote to acknowledge the best solution.

그러니까 답을 얻어내는데 사람들의 토론이 선행되면서 더 나은 답을 이야기할 수 있다는 그 과정

So, it's the process where discussions among people precede finding an answer, allowing for a better answer to be articulated.

그 과정 그 자체에 대한 중요성을 언급하는 이야기도 있고요.

There are also discussions mentioning the importance of the process itself.

여기에 추가로 채찌피티 사용자는 토론을 놓치고 답변만 얻습니다.

Additionally, ChatGPT users miss out on discussions and only receive answers.

이는 더 빠르고 효율적으로 볼 수도 있습니다.

This can be viewed more quickly and efficiently.

채찌피티를 통해 돌아오는 결과는 개발자가 필요로 하는 것과 완전히 일치하지 않을 수도 있습니다.

The results returned by ChatGPT may not completely align with what the developer needs.

하지만 보통의 경우 작동하는 솔루션으로서 형태를 갖추고 있고 조정해 나갈 수 있을 정도로 충분한 수준입니다.

However, in most cases, it has a form that functions as a workable solution and is sufficient enough to be adjusted.

채찌피티 같은 AI를 사용한 코드는 딱 결과로 나오는 그 코드만 얻게 되고 어떤 토론의 과정을 거치지 않는다 라는 이야기인 거죠.

The point is that using AI like ChatGPT only gives you the code as a result, without going through any discussion process.

또한 완벽하게 딱 들어맞는 코드를 제공하는 건 아니지만 웬만큼은 동작해주는 경우가 많다라는 건데요.

It doesn't provide a perfectly fitting code, but it often works well enough.

결국은 그 정도선에 그치게 된다 라는 뉘앙스로도 볼 수 있을 것 같습니다.

In the end, it seems that it can also be seen in the nuance that it will only reach that level.

토론을 거쳐서 더 좋은 결과를 얻어낼 수 있는 기회가 없어진다 라는 느낌이에요.

I feel that the opportunity to achieve better results through discussion is lost.

이건 마치 그냥 제 생각으로는 여러 종류의 개발자들이 모인 곳에서 왁자지껄하게 코드 리뷰를 해가며 더 나은 혹은 최상의 코드를 뽑아내는 경우와

This is like a situation where various types of developers gather together, noisily conducting code reviews to produce better or optimal code.

완벽하진 않지만 빠르게 시작할 수 있는 그럭저럭 쓸만한 코드를 바로 가져다 쓸 수 있는 경우

It's not perfect, but there are cases where you can quickly start using pretty decent code.

이 두 가지 경우 정도로 비교할 수 있을 것 같아요.

I think it can be compared to these two cases.

그런데 저는 이쪽 아티클의 내용을 보면서 그런 생각이 들긴 했습니다.

However, while looking at the content of this article, I did have such thoughts.

왜 굳이 두 개 중에 하나를 선택해야하나요?

Why do we have to choose one out of the two?

그 외에도 이러한 AI 사용 금지에 대해 반발해서 스택 오플로우 게시나 중재 및 검색에 대해 반발해서 코드 리뷰하는 것이 중요한 부분에 대해서는

In addition, regarding the backlash against the ban on using such AI, it's important to emphasize the significance of code reviews in response to the posts and moderation of Stack Overflow as well as the search functionalities.

사람들이랑 또 머리를 맞대고 코드 리뷰하면 되는 거 아닌가? 이런 생각이 살짝 들기도 했습니다.

Isn't it possible to just brainstorm together with people and do the code review? I had a slight thought like that.

개인적인 생각이긴 하고 비겁하게 중립을 유지하는 것 같긴 하지만 말이죠.

It's just my personal opinion, and it may seem cowardly to maintain neutrality, but still.

그 외에도 이러한 AI 사용 금지에 대해 반발해서 스택 오플로우 게시나 중재 및 검색에 대해 반발해서 코드 리뷰하는 것이 중요한 부분에 대해서는

In addition, regarding the backlash against the ban on using such AI, it is important to focus on code reviews related to Stack Overflow posts, moderation, and searches.

검증과 같은 자원봉사 활동을 멈추는 일도 일어나고 있었던 것 같습니다.

It seems that there were also instances of halting volunteer activities, such as verification.

더 편리한 도구 사용 자체를 막는다는 것에 반발하고 있는 건데요.

It's a backlash against the act of preventing the use of more convenient tools.

스택 오플로우 쪽은 6월쯤 아마도 이러한 반발의 여파로 금지 정책 자체를 좀 완화한 모양입니다.

It seems that Stack Overflow somewhat relaxed its ban policy around June, probably due to the backlash.

아직도 한참 시끄러운 이야기인 거죠.

It's still quite a noisy story, isn't it?

자원봉사자 개발자들의 큐레이팅에 많은 부분을 의존할 수밖에 없는 구조를 가진 상태에서

In a situation where it is inevitable to rely heavily on the curation of volunteer developers.

그들의 일을 좀 더 효율적으로 처리해주기 위한 툴로서의 AI 사용까지 막는 것은

Preventing the use of AI as a tool to handle their work more efficiently.

반발하고 있는 것 같습니다.

It seems like they are resisting.

이런 느낌으로 접근한 겁니다.

I approached it with this kind of feeling.

사실 이 정도 내용까지 살펴보다 보면 무슨 생각이 드냐면

In fact, when you look into content to this extent, what comes to mind?

결국 AI와 사람의 싸움이 아닌 겁니다.

In the end, it is not a fight between AI and humans.

결국은 사람과의 싸움으로 이미 번지고 있다는 거예요.

In the end, it's already spreading as a fight with people.

그리고 이러한 트래픽의 변화는 스택 오플로우뿐만 아니라

And this change in traffic is not only for Stack Overflow.

학술 플랫폼인 체그라는 서비스에서도 유사하게 일어나고 있습니다.

Similar events are happening on the academic platform Chegg as well.

원래 이 플랫폼은 어려운 숙제나 스템 과정의 과제를 위해서

This platform was originally designed for difficult homework or STEM course assignments.

전문가 혹은 튜터 도움을 연결해주는 형태로 서비스를 제공해서

By providing a service that connects you with experts or tutors.

대학생들에게 인기가 많은 서비스라고 해요.

It's said to be a popular service among university students.

역시나 채찌피티 같은 즉각적인 대답이 아니라

As expected, it's not an immediate response like ChatGPT.

사람이 대답해주기까지 기다려야 하는 부분이 필연적으로 발생할 수밖에 없기 때문에

Because there will inevitably be parts where one has to wait for a person to respond.

스택 오플로우와 유사한 분위기가 만들어지는 거죠.

It creates an atmosphere similar to Stack Overflow.

실제 이 서비스는 트래픽 뿐만 아니라 주식과 매출과

In reality, this service is not only about traffic but also about stocks and sales.

수익도 어느 정도씩 떨어지고 있다고 합니다.

It is said that profits are also declining to some extent.

그리고 그 이유로 채찌피티를 이야기하고 있고요.

And that's why we are talking about ChatGPT.

올해 초에 아마존이 내부코드에 채찌피티 코드를 사용하지 말라고 한 것도

Earlier this year, Amazon instructed not to use ChatGPT code in their internal code.

꽤 이슈였던 것도 기억납니다.

I also remember that it was quite an issue.

AI 혹은 데이터 사이언스의 근간을 구성하는 통계학에서는

In statistics, which forms the basis of AI or data science,

늘 인과관계와 상관관계를 혼동하지 않도록 주의를 기울여야 한다고 강조합니다.

It emphasizes the importance of being careful not to confuse cause-and-effect relationships with correlations.

하지만 최근의 여러 통계자료를 보아서는

However, looking at various recent statistical data...

생성형 AI 서비스의 출현과 높은 대중적 인기를 얻게 된 것이

The emergence of generative AI services and their high popularity among the public.

스택 오플로우와 같은 인간 대 인간 간의 토론 및

discussions between humans like Stack Overflow and

다소 느려 보이는 전통적인 지식 전달 방식의 인기 저하와

The decline in popularity of traditional knowledge transmission methods, which appear somewhat slow, and

인과관계를 가지는 것 처럼도 보입니다.

It seems to have a causal relationship.

명확하게 알 수는 없지만 상관관계는 있다고 볼 수 있겠죠.

It's not clear, but we can say there is a correlation.

AI 분야에선 비교적 최근에

In the field of AI, it is relatively recent.

AI 할루시네이션에 대한 이야기와는 다르게는

Unlike the discussion about AI hallucinations,

많이 나오고 있습니다.

It is coming out a lot.

간단하게 표현하자면 정확하지 않은, 맞지 않은 정보를

To put it simply, inaccurate or incorrect information.

AI가 생성한다라는 건데요.

It means that AI generates it.

이렇게 맞지 않은 정보에 대한 대응은

Response to such inaccurate information

결국 AI를 더 정확하게 만들거나

In the end, making AI more accurate or

오류를 줄이는 방법 쪽으로 생각을 할 수도 있고요.

You could also think about ways to reduce errors.

또 하나는 사람이 어느 정도 관여해서

Another thing is that a person is somewhat involved.

틀린 정보를 걸러낼 수 있게 하는 대응도 생각해볼 수 있다고 합니다.

It is said that we can consider measures that allow us to filter out incorrect information.

그런데 꼭 AI 할루시네이션과 같은

However, it is definitely similar to AI hallucination.

거창한 인기 전통적인 정보를

Grand popular traditional information.

AI 할루시네이션에 대한 이야기를 꺼내지 않아도

Even without bringing up the topic of AI hallucination

비슷한 경우는 이미 많이 일어나고 있다고 저는 생각합니다.

I believe that similar cases are already happening a lot.

유튜브에서 이야기하는 대로 투자를 했다가 돈을 날렸다라는 이야기

The story of losing money after investing as mentioned on YouTube.

아무 정보 없이 가치 없는 코인에 투자했다가 망한 이야기

A story about failing after investing in a worthless coin without any information.

사기당한 이야기, 피싱, 스미싱

A story about being scammed, phishing, smishing.

다 뭔가 잘못된 정보로 사람에게 피해를 준 건데

Everything has caused harm to people due to some incorrect information.

이건 사람이 노리고 있다는 거고요.

This means someone is targeting you.

AI의 경우는 학습한 내용에서 예측할 수 없는 내용이거나

In the case of AI, it is content that cannot be predicted from what has been learned, or...

맞지 않는 내용이 나온다는 건데

It means that the content doesn't match.

결국 그런 AI들이 학습한 내용이

Eventually, the content that such AIs have learned is

사람들이 만들어낸 여러 데이터를 학습한 거니까

It's because I learned from various data created by people.

사람들이 만들어낸 걸 학습해서 결과를 뱉어내는 거니까

It's because it's learning from what people have created and producing results.

그럴 수도 있겠다 싶긴 합니다.

I can see how that could be the case.

청취자 여러분들께서는 어떻게 생각하시나요?

What do you listeners think?

이렇게 자동화돼서 쉽게 코드나 답을 내주는 생성형 AI가

Generative AI that automates and easily provides code or answers like this.

사람 대 사람 간의 전통적인 지식 전달을

Traditional knowledge transfer between people.

완성시키는 것.

Completing something.

완전히 대체할 수 있다고 생각하시나요?

Do you think it can be completely replaced?

저는 개인적으로 오늘 에피소드를 준비하면서

Personally, while preparing for today's episode,

이런 생각들이 좀 들었습니다.

I had some thoughts like this.

좀 뜬구름 잡는 이야기일 수도 있는데요.

It might be somewhat of a vague or unrealistic story.

자율주행 자동차가 완성되면

When self-driving cars are completed

개인 운전기사, 즉 수행기사가 없어질까요?

Will personal chauffeurs, that is, dedicated drivers, disappear?

완강히 반대하고 손도 안 대려는 사람도 많지만

There are many people who strongly oppose and refuse to get involved.

저렴한 비용으로 대량 생산이 가능한

Capable of mass production at a low cost.

유전자 조작 식량은

Genetically modified food is

쓸모가 없었던가요?

Was it useless?

고급 브랜드의 수제 패션 자파들과

Handmade fashion items from luxury brands.

공산품으로 나오는 저렴한 제품들이

Affordable products that come as mass-produced items.

절대적인 품질상으로 드라마틱하게 차이가 나지 않는

There is not a dramatic difference in absolute quality.

시대에 이르렀지만

Even though we have reached this era

그 가격은 몇십배, 몇백배 이상 차이가 나는 건 또 어떨까요?

What if the price difference is dozens or even hundreds of times?

있는 집에서는 수많은 엘리트 인재를 붙여서

In the house, numerous elite talents are gathered.

과외를 시킬 수도 있겠지만

I could arrange for tutoring.

그럴 여력이 없거나

If there is no capacity for that or...

가정 형편이 매우 어렵고

The family's financial situation is very difficult.

어려운 상황인 경우라도

Even in difficult situations

스스로가 의지를 가지면

If you have willpower on your own

EBS가 됐건

Whether it's EBS

MOOC가 됐건

Whether it became a MOOC

하다못해 유튜브가 됐던 간에

At the very least, it could have been YouTube.

뭐든 접근성 좋은 뭔가를 활용해서

Using something that is easily accessible, no matter what it is.

공부를 해나갈 수 있는 것도

The ability to continue studying is also

그 나름의 의미가 있지 않을까요?

Could it have its own meaning?

그게 아주 고품질에

That's very high quality.

유일무이한 수준에

To an unparalleled level

오류 하나 없는 장인의 손길로 제작되는

Crafted with the flawless touch of a master artisan.

궁극의 뭔가가 아니라도 말이죠

Even if it's not something ultimate.

코드가 됐건 뭐가 됐건

Whether it's code or something else.

말입니다

It is a matter.

성냥의 불친절한 코딩 이야기였습니다

It was a story about the unkind coding of matches.

자막 제공 및 자막 제공 및 광고는 kakaotalk 플러스친구의 홈페이지에서 확인하실 수 있습니다.

You can check the subtitle provision and advertisements on the homepage of KakaoTalk Plus Friends.

Continue listening and achieve fluency faster with podcasts and the latest language learning research.