πŸ’­ The End of Front-End Development 정리

AromahyangΒ·2025λ…„ 4μ›” 15일

Frontend

λͺ©λ‘ 보기
8/9
post-thumbnail

2023λ…„ 3월에 μž‘μ„±λœ The End of Front-End DevelopmentλŠ” AI둜 인해 FEκ°€ μœ„ν˜‘λ°›λŠ”λ‹€λŠ” 견해에 λ°˜λŒ€ν•œλ‹€λŠ” λ‚΄μš©μ΄λ‹€.
μ €μžμ˜ 생각이 λΉ„μœ μ™€ ν•¨κ»˜ 잘 μ„€λͺ…λ˜μ–΄ μžˆμ–΄μ„œ μΈμƒκΉŠμ€ λΆ€λΆ„λ§Œ μ˜μ—­ν•΄μ„œ μ •λ¦¬ν–ˆλ‹€.

ν›„μ†νŽΈμΈ The Post-Developer Era도 κ³§ μž‘μ„±ν•  μ˜ˆμ •μ΄λ‹€.

πŸ’‘TL;DR

  • AIκ°€ λ°œμ „ν•˜κ³  μžˆμ§€λ§Œ, ν™˜κ°μ΄ μžˆμ„ 수 있고 λŒ€κ·œλͺ¨ μ„œλΉ„μŠ€ κ°œλ°œμ— λ―Έν‘ν•˜κΈ° λ•Œλ¬Έμ— κ°œλ°œμžκ°€ 사라지지 μ•Šμ„ 것이닀.
  • AI의 닡변을 λ§Ήμ‹ ν•˜μ§€ μ•Šκ³  λΉ„νŒμ μœΌλ‘œ μƒκ°ν•˜λ©° ν˜‘μ—… λ„κ΅¬λ‘œ ν™œμš©ν•œλ‹€λ©΄ 개발 생산성이 ν–₯상될 것이닀.
  • 개발 생산성이 ν–₯μƒλ˜λ©΄ λ‹€μ–‘ν•œ κΈ°λŠ₯을 그만큼 λΉ λ₯΄κ²Œ λ§Œλ“€μ–΄λ‚Ό 수 있고 λ‹€μ–‘ν•œ 버그듀을 λ§Œλ‚  수 있기 λ•Œλ¬Έμ— 개발자의 μˆ˜μš”λŠ” 계속 μžˆμ„ 것이닀.

Introduction

This sentiment is all over social media right now:
μ΄λŸ¬ν•œ μ •μ„œ(AIλŠ” FEκ°œλ°œμ— λŠ₯μˆ™ν•΄μ§ˆκ±°λ‹€)λŠ” ν˜„μž¬ μ†Œμ…œ λ―Έλ””μ–΄ μ „λ°˜μ— 걸쳐 μžˆμŠ΅λ‹ˆλ‹€:

I couldn't disagree more. I don't think front-end developer jobs are going anywhere. And I'm getting pretty sick of the FUD(Fear, Uncertainty, and Doubt) being spread online.
So, in this blog post, I'm going to share my hypothesis for what will happen. Things are going to change, but not in the scary way people are saying.
μ „ λ™μ˜ν•  수 μ—†μŠ΅λ‹ˆλ‹€. ν”„λ‘ νŠΈμ—”λ“œ 개발자의 μΌμžλ¦¬λŠ” μ‚¬λΌμ§ˆ 것 κ°™μ§€ μ•Šμ•„μš”. 그리고 μ˜¨λΌμΈμ—μ„œ 퍼지고 μžˆλŠ” 곡포, λΆˆν™•μ‹€μ„±, μ˜μ‹¬μ—λ„ κ½€ 지쳐가고 μžˆμŠ΅λ‹ˆλ‹€.
κ·Έλž˜μ„œ 이 λΈ”λ‘œκ·Έ κ²Œμ‹œλ¬Όμ—μ„œλŠ” μ•žμœΌλ‘œ 일어날 일에 λŒ€ν•œ 제 가섀을 κ³΅μœ ν•˜κ³ μž ν•©λ‹ˆλ‹€. 상황이 λ°”λ€” κ²ƒμ΄μ§€λ§Œ μ‚¬λžŒλ“€μ΄ λ§ν•˜λŠ” λ¬΄μ„œμš΄ 방식은 아닐 κ²ƒμž…λ‹ˆλ‹€.

λ‚˜λ„ μž‘λ…„μ— "FE개발만 ν•˜λ©΄ AI둜 λŒ€μ²΄λ  κ±°λ‹€"λΌλŠ” 말을 μ‚¬λžŒλ“€ μ‚¬μ΄μ—μ„œ, SNSμ—μ„œ κ½€ λ΄€μ—ˆλ‹€.
그런 κΈ€ ν˜Ήμ€ 말을 λ³Ό λ•Œλ§ˆλ‹€ 은근 슀트레슀 λ°›μ•˜μ—ˆλŠ”λ°, 이 λΆ„μ˜ 글을 쑰금 더 일찍 봀으면 μ’‹μ•˜μ„ κ±°λΌλŠ” 생각이 λ“ λ‹€.

Here we go again

Pretty much since the beginning, there's been a concern that web developers would be made redundant by some new technology. In the 2000s, it was WordPress. In the 2010s, it was Webflow. In the early 2020s, it was "no code" tools.
(μ€‘λž΅)
And yet, web developers continue to exist.
μ²˜μŒλΆ€ν„° μ›Ή κ°œλ°œμžλ“€μ΄ μƒˆλ‘œμš΄ 기술둜 인해 일자리λ₯Ό μžƒμ„ κ²ƒμ΄λΌλŠ” μš°λ €κ°€ μžˆμ—ˆμŠ΅λ‹ˆλ‹€. 2000λ…„λŒ€μ—λŠ” μ›Œλ“œν”„λ ˆμŠ€κ°€ κ·Έλž¬κ³ μš”. 2010λ…„λŒ€μ—λŠ” Webflowμ˜€μŠ΅λ‹ˆλ‹€. 2020λ…„λŒ€ μ΄ˆλ°˜μ—λŠ” 'λ…Έμ½”λ“œ' λ„κ΅¬μ˜€μŠ΅λ‹ˆλ‹€.
(μ€‘λž΅)
κ·ΈλŸΌμ—λ„ λΆˆκ΅¬ν•˜κ³  μ›Ή κ°œλ°œμžλŠ” 계속 μ‘΄μž¬ν•©λ‹ˆλ‹€.

There is an enormous difference between this HTML document and the sorts of code front-end developers write today.
이 HTML λ¬Έμ„œμ™€ μ˜€λŠ˜λ‚  ν”„λ‘ νŠΈμ—”λ“œ κ°œλ°œμžκ°€ μž‘μ„±ν•˜λŠ” μ½”λ“œ μ‚¬μ΄μ—λŠ” μ—„μ²­λ‚œ 차이가 μžˆμŠ΅λ‹ˆλ‹€.

즉, κ°„λ‹¨ν•œ μ›Ήμ‚¬μ΄νŠΈλ₯Ό λ§Œλ“œλŠ” 것과 μ „λ¬Έκ°€(μ—¬κΈ°μ—μ„œλŠ” ν”„λ‘ νŠΈμ—”λ“œ 개발자)κ°€ λ§Œλ“œλŠ” 것에 λΆ„λͺ…ν•œ 차이가 있기 λ•Œλ¬Έμ— κ°œλ°œμžκ°€ AI둜 λŒ€μ²΄λ˜κ±°λ‚˜ 사라지지 μ•Šμ„ 거라고 μ–˜κΈ°ν•œλ‹€.

Looking into the future

LLMs aren't able to validate their assumptions, or test their hypotheses. They can't confirm whether what they're saying is true or not. (μ€‘λž΅) Sometimes, parts of that response are nonsensical. The OpenAI team refers to these as β€œhallucinations”.
LLM은 μžμ‹ μ˜ 가정을 κ²€μ¦ν•˜κ±°λ‚˜ 가섀을 ν…ŒμŠ€νŠΈν•  수 μ—†μŠ΅λ‹ˆλ‹€. 그듀은 μžμ‹ μ΄ λ§ν•˜λŠ” 것이 사싀인지 μ•„λ‹Œμ§€ 확인할 수 μ—†μŠ΅λ‹ˆλ‹€. (μ€‘λž΅) OpenAI νŒ€μ€ 이λ₯Ό "ν™˜κ°"이라고 λΆ€λ¦…λ‹ˆλ‹€.

And so, the accuracy will improve, but it'll never be perfect. (μ€‘λž΅) If you're not a programmer, you won't be able to tell which parts are accurate or not. You won't be able to spot the hallucinations.
λ”°λΌμ„œ μ •ν™•λ„λŠ” ν–₯μƒλ˜κ² μ§€λ§Œ μ™„λ²½ν•˜μ§€λŠ” μ•Šμ„ κ²ƒμž…λ‹ˆλ‹€. (μ€‘λž΅) ν”„λ‘œκ·Έλž˜λ¨Έκ°€ μ•„λ‹ˆλΌλ©΄ μ–΄λ–€ 뢀뢄이 μ •ν™•ν•œμ§€ μ•„λ‹Œμ§€ ꡬ뢄할 수 μ—†μŠ΅λ‹ˆλ‹€. ν™˜κ°μ„ λ°œκ²¬ν•  수 없을 κ²ƒμž…λ‹ˆλ‹€.

LLM의 ν™˜κ° ν˜„μƒμ„ μ™„λ²½ν•˜κ²Œ ν•΄κ²°ν•  수 없을 κ²ƒμ΄λ―€λ‘œ 전문가인 κ°œλ°œμžκ°€ 직접 사싀여뢀λ₯Ό 확인해야 ν•œλ‹€.
AIκ°€ λ§Œλ“€μ–΄λ‚Έ 잘λͺ»λœ 정보(ν™˜κ°)λ₯Ό κ²€μ¦ν•˜μ§€ μ•Šκ³  κ·ΈλŒ€λ‘œ 썼을 λ•Œ λˆ„κ°€ μ±…μž„μ§ˆ 것인가?
AIκ°€ μ±…μž„μ§€μ§€ μ•Šμ„ κ²ƒμ΄λ―€λ‘œ κ°œλ°œμžκ°€ 항상 ν•„μš”ν•  κ±°κ³  사라지지 μ•Šμ„ 것이닀.

AI isn't magic. It's only as good as its training data. Code snippets are all over the internet, and are often generic. By contrast, every codebase is unique. There are very few large open-source codebases. How's the AI supposed to learn how to build big real-world projects?
AIλŠ” λ§ˆλ²•μ΄ μ•„λ‹™λ‹ˆλ‹€. AIλŠ” ν•™μŠ΅ λ°μ΄ν„°λ§ŒνΌλ§Œ μš°μˆ˜ν•  λΏμž…λ‹ˆλ‹€. μ½”λ“œ μŠ€λ‹ˆνŽ«μ€ 인터넷에 널렀 있으며 일반적인 κ²½μš°κ°€ λ§ŽμŠ΅λ‹ˆλ‹€. λ°˜λ©΄μ— λͺ¨λ“  μ½”λ“œλ² μ΄μŠ€λŠ” κ³ μœ ν•©λ‹ˆλ‹€. λŒ€κ·œλͺ¨ μ˜€ν”ˆ μ†ŒμŠ€ μ½”λ“œλ² μ΄μŠ€λŠ” 거의 μ—†μŠ΅λ‹ˆλ‹€. AIλŠ” μ–΄λ–»κ²Œ μ‹€μ œ λŒ€κ·œλͺ¨ ν”„λ‘œμ νŠΈλ₯Ό κ΅¬μΆ•ν•˜λŠ” 방법을 ν•™μŠ΅ν•  수 μžˆμ„κΉŒμš”?

But I think we're still a very long way from major tech companies letting go of their developer staff and replacing them with prompt engineers. It seems to me like there are several potentially-unsolveable problems that stand in the way of this becoming a reality.
ν•˜μ§€λ§Œ μ£Όμš” 기술 기업듀이 개발자λ₯Ό ν•΄κ³ ν•˜κ³  이듀을 즉각적인 μ—”μ§€λ‹ˆμ–΄λ‘œ λŒ€μ²΄ν•˜κΈ°μ—λŠ” 아직 갈 길이 λ©€λ‹€κ³  μƒκ°ν•©λ‹ˆλ‹€. 이λ₯Ό μ‹€ν˜„ν•˜λŠ” λ°λŠ” λͺ‡ κ°€μ§€ ν•΄κ²°ν•˜μ§€ λͺ»ν•  잠재적 λ¬Έμ œκ°€ μžˆλŠ” 것 κ°™μŠ΅λ‹ˆλ‹€.

μ½”λ“œ μŠ€λ‹ˆνŽ«μ„ 잘 μ œμ‹œν•˜λŠ” κ²ƒλ§Œ 보고 κ°œλ°œμžλŠ” AI둜 λŒ€μ²΄ν•˜λŠ” κ±°λŠ” 2023년에도, μ§€κΈˆλ„ μ„£λΆ€λ₯Έ νŒλ‹¨μΈ κ±° κ°™λ‹€.

Augmenting, not replacing

You might not know it from this article, but I'm actually pretty optimistic about AI. πŸ˜…
I think the most likely scenario is that tools like GPT-4 become integrated into developer tooling, and used to amplify what a skilled developer can do.
이 κΈ€μ—μ„œλŠ” 잘 λͺ¨λ₯΄μ‹œκ² μ§€λ§Œ μ €λŠ” 사싀 AI에 λŒ€ν•΄ κ½€ λ‚™κ΄€μ μœΌλ‘œ μƒκ°ν•©λ‹ˆλ‹€. κ°€μž₯ κ°€λŠ₯성이 높은 μ‹œλ‚˜λ¦¬μ˜€λŠ” GPT-4와 같은 도ꡬ가 개발자 도ꡬ에 ν†΅ν•©λ˜μ–΄ μˆ™λ ¨λœ κ°œλ°œμžκ°€ ν•  수 μžˆλŠ” 일을 μ¦ν­μ‹œν‚€λŠ” 데 μ‚¬μš©λ˜λŠ” 것이라고 μƒκ°ν•©λ‹ˆλ‹€.

Carpenters weren't replaced by power tools, accountants weren't replaced by spreadsheets, photographers weren't replaced by digital cameras / smartphones, and I don't think that developers will be replaced by LLMs.
λͺ©μˆ˜λŠ” 전동 곡ꡬ둜, νšŒκ³„μ‚¬λŠ” μŠ€ν”„λ ˆλ“œμ‹œνŠΈλ‘œ, μ‚¬μ§„μž‘κ°€λŠ” λ””μ§€ν„Έ 카메라/슀마트폰으둜 λŒ€μ²΄λ˜μ§€ μ•Šμ•˜λ“―μ΄
, κ°œλ°œμžκ°€ LLM으둜 λŒ€μ²΄λ˜μ§€λŠ” μ•Šμ„ 것이라고 μƒκ°ν•©λ‹ˆλ‹€.

κ°œλ°œμ— λŒ€ν•œ μˆ˜μš”λŠ” μ¦κ°€ν•˜κ³  μžˆλŠ” μƒνƒœμ΄λ©°, 개발의 생산성이 LLM 덕에 ν–₯μƒλœλ‹€λ©΄ 그만큼 λ§Žμ€ κΈ°λŠ₯을 λ§Œλ“€κ³  λ§Žμ€ 버그듀이 λ°œμƒν•  것이기 λ•Œλ¬Έμ— κ°œλ°œμžκ°€ 사라지지 μ•Šμ„ 것이닀.

We're not the only ones having this conversation

λ² ν…Œλž‘ μ• λ‹ˆλ©”μ΄ν„° κ²Έ μΌλŸ¬μŠ€νŠΈλ ˆμ΄ν„°μΈ Aaron Blaise의 사둀닀.

he doesn't see these tools as a threat, he sees them as something that will increase animator productivity, and lead to more animator jobs.
Artists and knowledge workers across dozens of industries are having the same conversation right now. People are worried that their job is about to be gobbled up by AIs like GPT-4, DALL-E 2, and Midjourney.
κ·ΈλŠ” μ΄λŸ¬ν•œ νˆ΄μ„ μœ„ν˜‘μœΌλ‘œ 보지 μ•Šκ³  μ• λ‹ˆλ©”μ΄ν„°μ˜ 생산성을 높이고 더 λ§Žμ€ μ• λ‹ˆλ©”μ΄ν„° 일자리λ₯Ό μ°½μΆœν•  수 μžˆλŠ” 기회둜 보고 μžˆμŠ΅λ‹ˆλ‹€.
μˆ˜μ‹­ 개의 μ‚°μ—… λΆ„μ•Όμ˜ μ•„ν‹°μŠ€νŠΈμ™€ 지식 κ·Όλ‘œμžλ“€μ΄ μ§€κΈˆ 같은 λŒ€ν™”λ₯Ό λ‚˜λˆ„κ³  μžˆμŠ΅λ‹ˆλ‹€. μ‚¬λžŒλ“€μ€ GPT-4, DALL-E 2, Midjourney와 같은 인곡지λŠ₯이 μžμ‹ λ“€μ˜ 일자리λ₯Ό μž μ‹ν• κΉŒ 봐 κ±±μ •ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€.

개발 뿐만 μ•„λ‹ˆλΌ μ—¬λŸ¬ μ‚°μ—… λΆ„μ•Όμ—μ„œλ„ AI λ„κ΅¬μ˜ μœ„ν˜‘μ„ 느끼고 μžˆλ‹€.
ν•˜μ§€λ§Œ AI 도ꡬλ₯Ό 일자리λ₯Ό μœ„ν˜‘ν•˜λŠ” 관점 λŒ€μ‹  ν˜‘μ—…μ„ ν†΅ν•œ 일자리 창좜의 κ΄€μ μœΌλ‘œ λ°”λΌλ³΄λŠ” 것도 ν•„μš”ν•œ κ±° κ°™λ‹€.

Using LLMs to help you learn

The wrong way would be to treat it like GPS navigation. When I have to drive somewhere, I pop the address into my GPS, and follow its instructions indiscriminately. I do generally wind up where I need to go, but it requires 0 mental effort on my part. As a result, my sense of direction has totally atrophied. I can't go anywhere now without a synthesized voice telling me where to turn. 😬
잘λͺ»λœ 방법은 GPS λ‚΄λΉ„κ²Œμ΄μ…˜μ²˜λŸΌ μ·¨κΈ‰ν•˜λŠ” κ²ƒμž…λ‹ˆλ‹€. μ €λŠ” μ–΄λ”˜κ°€λ‘œ μš΄μ „ν•΄μ•Ό ν•  λ•Œ GPS에 μ£Όμ†Œλ₯Ό μž…λ ₯ν•˜κ³  λ¬΄μ°¨λ³„μ μœΌλ‘œ μ§€μ‹œλ₯Ό λ”°λ¦…λ‹ˆλ‹€. 보톡은 μ œκ°€ κ°€μ•Ό ν•  곳에 λ„μ°©ν•˜κΈ΄ ν•˜μ§€λ§Œ 정신적인 λ…Έλ ₯이 μ „ν˜€ ν•„μš”ν•˜μ§€ μ•ŠμŠ΅λ‹ˆλ‹€. κ·Έ κ²°κ³Ό λ°©ν–₯ 감각이 μ™„μ „νžˆ μœ„μΆ•λ˜μ—ˆμŠ΅λ‹ˆλ‹€. μ΄μ œλŠ” μ–΄λ””λ‘œ κ°€μ•Ό ν•˜λŠ”μ§€ μ•Œλ €μ£ΌλŠ” ν•©μ„±λœ μŒμ„± μ—†μ΄λŠ” 아무 데도 갈 수 μ—†μŠ΅λ‹ˆλ‹€.

AI의 닡변을 μ ˆλŒ€μ μœΌλ‘œ λ―ΏλŠ” μ‚¬λžŒμ΄ μ’…μ’… μžˆλŠ” 것 κ°™λ˜λ°, μ μ ˆν•œ λΉ„μœ μΈ κ±° κ°™λ‹€.

Instead of treating it like a GPS, I'd suggest treating it like you're a member of a jury, and the LLM is the defendant, taking the stand.
You'll listen to what they have to say, but you won't accept it as fact. You'll be skeptical, and think critically about every word.
Instead of blindly copy/pasting the code that ChatGPT generates, go through it line-by-line, and make sure you understand. Ask it for clarification. And double-check things that seem suspicious with an authoritative source (eg. the official documentation). Keep in mind that LLMs are 100% confident, but not 100% accurate.
GPS처럼 μ·¨κΈ‰ν•˜λŠ” λŒ€μ‹  λ°°μ‹¬μ›λ‹¨μ˜ μΌμ›μœΌλ‘œ, LLM은 ν”Όκ³ μΈμœΌλ‘œμ„œ 증인석에 μ„œ μžˆλ‹€κ³  μƒκ°ν•˜λ©΄ 쒋을 것 κ°™μŠ΅λ‹ˆλ‹€.
그듀이 ν•˜λŠ” 말을 λ“£κΈ°λŠ” ν•˜κ² μ§€λ§Œ 그것을 μ‚¬μ‹€λ‘œ λ°›μ•„λ“€μ΄μ§€λŠ” μ•Šμ„ κ²ƒμž…λ‹ˆλ‹€. 회의적인 νƒœλ„λ₯Ό μ·¨ν•˜κ³  λͺ¨λ“  말에 λŒ€ν•΄ λΉ„νŒμ μœΌλ‘œ μƒκ°ν•˜κ²Œ 될 κ²ƒμž…λ‹ˆλ‹€.
ChatGPTκ°€ μƒμ„±ν•˜λŠ” μ½”λ“œλ₯Ό λ¬΄ν„±λŒ€κ³  볡사/λΆ™μ—¬λ„£κΈ°ν•˜μ§€ 말고 ν•œ 쀄 ν•œ 쀄 μ‚΄νŽ΄λ³΄κ³  μ΄ν•΄ν–ˆλŠ”μ§€ ν™•μΈν•˜μ„Έμš”. μ„€λͺ…을 μš”μ²­ν•˜μ„Έμš”. 그리고 μ˜μ‹¬μŠ€λŸ¬μš΄ 뢀뢄은 곡신λ ₯ μžˆλŠ” 좜처(예: 곡식 λ¬Έμ„œ)λ₯Ό 톡해 λ‹€μ‹œ ν™•μΈν•˜μ„Έμš”. LLM은 100% ν™•μ‹ ν•  μˆ˜λŠ” μžˆμ§€λ§Œ 100% μ •ν™•ν•˜μ§€λŠ” μ•Šλ‹€λŠ” 점을 λͺ…μ‹¬ν•˜μ„Έμš”.

If you follow this strategy, I think LLMs can offer a lot of value.
이 μ „λž΅μ„ λ”°λ₯Έλ‹€λ©΄ LLM은 λ§Žμ€ κ°€μΉ˜λ₯Ό μ œκ³΅ν•  수 μžˆλ‹€κ³  μƒκ°ν•©λ‹ˆλ‹€.

profile
보여주기 식이 μ•„λ‹Œ κ²½ν—˜ν•˜κ³  배운 것을 μ†Œμ†Œν•˜κ²Œ κΈ°λ‘ν•˜μ˜€μŠ΅λ‹ˆλ‹€.

0개의 λŒ“κΈ€