쉐도잉 연습: Learning Software Engineering During the Era of AI | Raymond Fu | TEDxCSTU - YouTube로 영어 말하기 배우기

어려움
쉐도잉 컨트롤
0% 완료 (0/128 문장)
Transcriber: Brenda Meza Reviewer: Emilia Soso At the turn of the century, when I started to learn software engineering, one of my professors told us that in the future, every job will be a programming job.
⏸ 일시 정지
재생 속도:
반복 횟수:
대기 모드:
자막 동기:0ms
모든 문장
128 문장
1
Transcriber: Brenda Meza Reviewer: Emilia Soso At the turn of the century, when I started to learn software engineering, one of my professors told us that in the future, every job will be a programming job.
0:00.00 – 0:17.23 (17.2s)
2
That was in 2001.
0:17.46 – 0:18.86 (1.4s)
3
And he said that we’re holding a golden ticket to job security.
0:19.66 – 0:24.06 (4.4s)
4
Just last month, the CEO of GitHub said that the future of programming is natural language.
0:24.86 – 0:32.43 (7.6s)
5
It looks like the prediction of my professor at the turn of the century is going to become true, but probably not in the way that he had imagined.
0:34.00 – 0:42.36 (8.4s)
6
Artificial intelligence is capable of writing code for you through a natural language prompt.
0:43.20 – 0:48.90 (5.7s)
7
GitHub Copilot can complete code for you and fix bugs for you.
0:49.26 – 0:54.23 (5.0s)
8
And ChatGPT can create an entire project for you within seconds.
0:54.66 – 0:59.40 (4.7s)
9
And all these tools are available to anyone.
0:59.93 – 1:02.76 (2.8s)
10
So I find myself wondering, have we lost our golden tickets to job security?
1:03.66 – 1:08.63 (5.0s)
11
And as a CSTU professor and a father to a daughter who studied Computer Science, there's a bigger question for me.
1:09.63 – 1:17.90 (8.3s)
12
If AI is going to do programming, is it still worth it for us to learn software engineering anymore?
1:18.60 – 1:25.10 (6.5s)
13
Today, I would like to explore this question with all of you guys.
1:26.70 – 1:30.26 (3.6s)
14
Let’s talk about what AI can do and more importantly, how our students of software engineering can prepare for the future roles of a real software engineer.
1:30.43 – 1:41.76 (11.3s)
15
So let’s dive in.
1:42.20 – 1:43.33 (1.1s)
16
First, let’s talk about what AI is good at.
1:43.76 – 1:46.43 (2.7s)
17
In terms of programming, AI is really good at generating thousands of lines of code.
1:47.70 – 1:53.40 (5.7s)
18
It translates between programming languages.
1:53.70 – 1:56.20 (2.5s)
19
It can create user interfaces and fix bugs for you.
1:56.40 – 2:01.23 (4.8s)
20
And it excels at repetitive tasks, and, you know, pattern recognition.
2:02.06 – 2:06.96 (4.9s)
21
You know, once I asked ChatGPT to create a project for me, a dating app like Tinder in Python.
2:07.93 – 2:14.70 (6.8s)
22
And within seconds it actually created a complete application with user profiles, the swiping logic, and even a sample database.
2:15.30 – 2:24.16 (8.9s)
23
The only thing it didn't do for me is find me a date.
2:24.83 – 2:27.90 (3.1s)
24
(Laughs) But AI has a lot of limitations. We have to accept that.
2:28.36 – 2:33.56 (5.2s)
25
It still doesn’t understand the why behind all the tasks we ask them to do.
2:33.73 – 2:37.83 (4.1s)
26
It needs your human input for real-world context and scenarios.
2:38.16 – 2:44.23 (6.1s)
27
It may not work well, prioritizing long-term business goals and assessing trade-offs.
2:44.73 – 2:50.66 (5.9s)
28
And last but not least, it's not reliable.
2:51.30 – 2:54.46 (3.2s)
29
It hallucinates and sometimes gives the wrong answer.
2:55.16 – 2:58.46 (3.3s)
30
The statistics say that 55% of the developers today are actually starting to use Copilot, but only 30% of them are accepting the outcome without any changes.
3:00.50 – 3:12.36 (11.9s)
31
So if you are a developer and you are not in the first 55%, that means you’re not using AI, and you’re in trouble.
3:12.36 – 3:20.00 (7.6s)
32
But if you are in the 30%, that means you trust AI too much.
3:20.60 – 3:25.16 (4.6s)
33
You may be in bigger trouble.
3:25.33 – 3:27.26 (1.9s)
34
So all the leading AIs today are built on top of large language models, and it’s trained on the text of human knowledge.
3:29.50 – 3:37.36 (7.9s)
35
It’s impressive.
3:37.70 – 3:38.70 (1.0s)
36
If you give a clear prompt, it’ll give you very good results.
3:38.70 – 3:42.50 (3.8s)
37
But all the strategic thinking are still us. It’s the human.
3:42.86 – 3:47.20 (4.3s)
38
You can think of AI as a brilliant junior developer that you hire to your team, and they can do a lot of jobs very quickly and efficiently.
3:47.46 – 3:57.93 (10.5s)
39
But it's up to us human to define the vision, to validate the results and ensure what we're building is good for the society.
3:58.26 – 4:08.10 (9.8s)
40
So there’s another thing that I want to talk about that AI is struggling with.
4:09.76 – 4:14.36 (4.6s)
41
It's struggling to communicate and collaborate with human beings.
4:14.56 – 4:18.80 (4.2s)
42
Well, maybe you will say this is more of a human problem, right?
4:19.63 – 4:23.30 (3.7s)
43
We humans sometimes deal with the same problem too.
4:23.46 – 4:26.26 (2.8s)
44
But this is something we will have to work out.
4:26.80 – 4:29.56 (2.8s)
45
Let AI do what AI is good at, and we humans can take care of the boring jobs such as handling office politics.
4:29.56 – 4:37.80 (8.2s)
46
So talk about the capabilities and limitations of AI.
4:38.63 – 4:42.80 (4.2s)
47
Now we can take a look at the software engineering roles.
4:43.20 – 4:46.33 (3.1s)
48
So software engineering roles is not just about writing code.
4:47.66 – 4:52.10 (4.4s)
49
It actually is about how we need to understand what the user needs.
4:52.10 – 4:57.26 (5.2s)
50
We need to collaborate across roles and also make tough decisions with empathy and responsibility.
4:57.26 – 5:04.70 (7.4s)
51
This is what a software engineer should be doing, right?
5:04.90 – 5:07.90 (3.0s)
52
We're not just text executors.
5:07.90 – 5:09.60 (1.7s)
53
The best engineers are not the ones who code the fastest, but the ones who think the deepest.
5:10.66 – 5:15.80 (5.1s)
54
So a good engineer will take messy problems, ambiguous problems, and guide machines towards structured and meaningful outcomes.
5:17.16 – 5:26.26 (9.1s)
55
So there are system architects who design the best solutions, and they should be the AI collaborators who use AI to implement those solutions.
5:27.00 – 5:38.16 (11.2s)
56
And then they need to be ethical technologists to make sure the solutions that we’re building are truly benefiting human beings.
5:38.76 – 5:46.96 (8.2s)
57
So AI is actually democratizing a lot of complicated technical tasks.
5:47.96 – 5:52.93 (5.0s)
58
Like, today a designer can mock up an application with a prompt.
5:52.96 – 5:59.73 (6.8s)
59
And also marketers, they don’t need data engineers.
6:00.03 – 6:03.13 (3.1s)
60
They can just run data analytics without writing any code.
6:03.13 – 6:07.73 (4.6s)
61
Does that mean software engineers are losing advantages?
6:08.43 – 6:11.73 (3.3s)
62
The answer is no.
6:12.43 – 6:13.56 (1.1s)
63
It still remains essential for software engineers.
6:15.63 – 6:18.83 (3.2s)
64
And the reason is as follows.
6:18.83 – 6:21.20 (2.4s)
65
First, we understand AI better.
6:22.13 – 6:25.30 (3.2s)
66
We not only know how to prompt, and we also know what’s under the hood.
6:25.80 – 6:30.10 (4.3s)
67
The models, the data pipelines, the limitations and risks.
6:30.46 – 6:34.76 (4.3s)
68
And the understanding of these are very important because AI is integrated into every product we’re using and we’re building in the future.
6:35.50 – 6:44.36 (8.9s)
69
Second, we can make better use of AI when building software.
6:44.66 – 6:48.80 (4.1s)
70
So nowadays anybody can prototype a demo or create a simple application of features.
6:49.40 – 6:55.83 (6.4s)
71
But software engineers think of the bigger picture.
6:56.30 – 6:59.06 (2.8s)
72
We are actually using AI to build a production-ready software that’s scalable and reliable with long-term maintainability.
6:59.70 – 7:08.46 (8.8s)
73
Finally, we are making AI better.
7:09.83 – 7:12.56 (2.7s)
74
We fine-tune models.
7:13.00 – 7:14.53 (1.5s)
75
We optimize the performance and improve usability.
7:14.53 – 7:18.66 (4.1s)
76
We make AI available and useful for everybody else.
7:19.10 – 7:22.96 (3.9s)
77
The next generation of AI is still built by software engineers.
7:23.46 – 7:27.16 (3.7s)
78
Do you guys remember this quote from CEO of GitHub?
7:27.96 – 7:31.13 (3.2s)
79
This is not a reality yet.
7:31.66 – 7:33.50 (1.8s)
80
It's still up to the software engineers to improve AI and make this happen.
7:33.53 – 7:39.70 (6.2s)
81
So software engineers were not losing the golden ticket to job security.
7:41.53 – 7:46.26 (4.7s)
82
As a matter of fact, we’re collecting even more because we’re no longer just building software.
7:46.53 – 7:51.16 (4.6s)
83
We're actually building the future intelligence itself.
7:51.36 – 7:55.06 (3.7s)
84
And how we train, direct and supervise AI today will define the kind of systems, technology and society that we’re building tomorrow.
7:57.13 – 8:08.43 (11.3s)
85
AI is raising the floor, but software engineers are raising the ceiling.
8:10.23 – 8:15.06 (4.8s)
86
And I want to share this not just with… You can applaud, that’s okay.
8:15.86 – 8:19.50 (3.6s)
87
I want to share this with not just system engineers.
8:19.63 – 8:22.66 (3.0s)
88
This is for everyone, all right?
8:22.66 – 8:24.30 (1.6s)
89
We have AI that’s raising us up from the floor.
8:24.33 – 8:26.76 (2.4s)
90
But it’s us humans that have to reach to the ceiling and raise up the ceiling.
8:26.76 – 8:31.56 (4.8s)
91
All right, so after all this, now we can talk about software engineering education, right.
8:32.96 – 8:37.73 (4.8s)
92
So, in the past, coding was a very important piece of software engineering education.
8:37.73 – 8:42.86 (5.1s)
93
But software engineering education is not just about writing code.
8:42.86 – 8:46.53 (3.7s)
94
It's also about teaching you how to break complex problems into steps, think logically and critically, and harness the digital tools to build solutions that really matters.
8:46.53 – 8:58.30 (11.8s)
95
So in a time when AI is everybody’s assistant, engineers become the orchestrators.
8:59.33 – 9:06.66 (7.3s)
96
We remove barriers and open doors.
9:07.03 – 9:10.30 (3.3s)
97
And in order for us to be a successful software engineer, the students should go beyond learning code as quickly as possible and get into the following things.
9:11.56 – 9:22.40 (10.8s)
98
So in order to become a successful engineer in the future, we should focus on mastering the foundations.
9:24.26 – 9:29.90 (5.6s)
99
The data structure, the algorithm, the programming concepts.
9:30.50 – 9:33.53 (3.0s)
100
They are still very important.
9:33.53 – 9:34.96 (1.4s)
101
Spend enough time to learn all these and become an expert on them because they’re very important basics.
9:34.96 – 9:41.86 (6.9s)
102
Next, think about a system like an architect because, you know, aim higher.
9:42.80 – 9:49.53 (6.7s)
103
Meet the expectation of a senior engineer as soon as possible.
9:49.96 – 9:53.60 (3.6s)
104
And think about designing systems that are reliable and scalable.
9:53.60 – 9:58.03 (4.4s)
105
Go beyond, go full-stack across disciplines.
9:59.83 – 10:03.43 (3.6s)
106
The days when a software engineer could focus on either the front end or the back end or the database are gone.
10:04.00 – 10:11.53 (7.5s)
107
The future software engineers are full-stack engineers.
10:11.86 – 10:14.76 (2.9s)
108
And there’s more.
10:15.13 – 10:16.13 (1.0s)
109
You need to also get into the other disciplines like design, product, data, project management, and be prepared to wear multiple hats.
10:16.23 – 10:23.76 (7.5s)
110
Practice communication and collaborations.
10:25.33 – 10:28.40 (3.1s)
111
Learn to work with people through team projects.
10:28.80 – 10:33.03 (4.2s)
112
Because in the future, if you can explain and connect, it will become increasingly important, and it will set you apart.
10:33.43 – 10:42.83 (9.4s)
113
Use AI as a creative partner.
10:44.56 – 10:46.56 (2.0s)
114
Embrace AI, don’t hate it.
10:47.16 – 10:49.13 (2.0s)
115
And learn LLM, generative AI, model fine-tuning and RAG, etc.
10:49.53 – 10:56.06 (6.5s)
116
You discuss your project with AI, and delegate your work to AI as if it’s one of your teammates.
10:56.76 – 11:02.46 (5.7s)
117
Last but not least, stay adaptable.
11:03.86 – 11:06.33 (2.5s)
118
Tools change, principles last.
11:06.96 – 11:09.80 (2.8s)
119
So you should always focus on learning how to learn.
11:10.46 – 11:13.83 (3.4s)
120
So in the future, when everyone can code a little, the ones who can master the craft, will build the path for everyone and become the leader.
11:15.20 – 11:25.26 (10.1s)
121
So in the era of AI, software engineering is becoming the foundation of leadership.
11:26.00 – 11:32.06 (6.1s)
122
I've talked a lot about programming, but perhaps programmer is no longer the right term we should be using to refer to software engineers.
11:34.60 – 11:41.60 (7.0s)
123
The software engineers of the AI era should be visionaries who can define meaningful problems.
11:42.20 – 11:50.33 (8.1s)
124
A bridge builder who can connect tools, teams and disciplines, and leaders who not only lead human beings, but also lead AI.
11:50.70 – 12:01.00 (10.3s)
125
So the future doesn't belong to those who code the fastest, it should belong to the ones who think deeply, adapt quickly, and collaborate efficiently.
12:01.96 – 12:12.70 (10.7s)
126
They are the ones who don't just predict the future.
12:13.30 – 12:16.16 (2.9s)
127
We build the future.
12:16.66 – 12:17.90 (1.2s)
128
Thank you.
12:19.26 – 12:20.26 (1.0s)

맥락 및 배경

2021년, 레이몬드 푸는 TEDx 강연에서 인공지능(AI)의 시대에 소프트웨어 공학을 배우는 문제에 대해 이야기했습니다. 그는 공학을 배우기 시작했을 때, 모든 직업이 프로그래밍 직업이 될 것이라는 예측을 듣고 그 가능성에 대해 고찰했습니다. 오늘날 AI의 발전으로, 프로그래밍 작업이 자동화되고 있지만, 여전히 인간의 역할이 중요하다는 점을 강조했습니다.

일상 대화를 위한 5가지 주요 구문

  • AI는 코드 작성을 쉽게 만들어줍니다. - AI의 도움을 받으면 많은 작업을 신속하게 처리할 수 있습니다.
  • 인간의 입력이 필요합니다. - AI는 전체적인 비전과 전략적 사고를 이해하지 못합니다.
  • 소프트웨어 공학은 단순히 코드를 작성하는 것이 아닙니다. - 사용자 요구를 이해하고 협업하는 것이 중요합니다.
  • AI는 제한이 있습니다. - AI의 결과를 맹신하기보다는 비판적으로 평가해야 합니다.
  • 우리는 인간으로서의 책임을 져야 합니다. - AI가 처리할 수 없는 직무는 여전히 오랫동안 인간이 맡아야 합니다.

단계별 섀도잉 가이드

이 동영상의 난이도를 효과적으로 극복하기 위해, 다음과 같은 섀도 스피치(shadow speech) 방법을 사용할 수 있습니다:

  1. 동영상 선택: 어려움을 느끼는 섹션을 선택합니다. 이 연설의 경우, AI의 기능과 한계 부분이 적합합니다.
  2. 듣기 연습: 전체 내용을 처음부터 끝까지 들어보세요. 어떤 주제인지 이해합니다.
  3. 문장 따라 말하기: 한 문장씩 반복하여 말해보세요. 이 과정에서 유튜브 영어 공부를 통해 발음을 교정할 수 있습니다.
  4. 비교 및 수정: 자신의 발음과 레이몬드 푸의 발음을 비교해 보세요. 필요한 수정 사항을 적고 다시 연습합니다.
  5. 자신감 있게 말하기: 최종적으로 배운 내용을 바탕으로 이를 활용해 스스로도 말을 만들어보세요. 이 과정은 shadowing site에 게시하거나 친구와 함께 연습해보세요.

이 방법들을 통해 일상에서 더 자연스럽게 영어를 사용할 수 있게 되고, 영어 발음 교정에도 큰 도움이 될 것입니다.

쉐도잉이란? 영어 실력을 빠르게 키우는 과학적 방법

쉐도잉(Shadowing)은 원래 전문 통역사 훈련을 위해 개발된 언어 학습 기법으로, 다언어 학자인 Dr. Alexander Arguelles에 의해 대중화된 방법입니다. 핵심 원리는 간단하지만 매우 강력합니다: 원어민의 영어를 들으면서 1~2초의 짧은 지연으로 즉시 소리 내어 따라 말하는 것——마치 '그림자(shadow)'처럼 화자를 따라가는 것입니다. 문법 공부나 수동적인 청취와 달리, 쉐도잉은 뇌와 입 근육이 동시에 실시간으로 영어를 처리하고 재현하도록 훈련합니다. 연구에 따르면 이 방법은 발음 정확도, 억양, 리듬, 연음, 청취력, 말하기 유창성을 크게 향상시킵니다. IELTS 스피킹 준비와 자연스러운 영어 소통을 원하는 분들에게 특히 효과적입니다.

ShadowingEnglish에서 효과적으로 학습하는 방법

  1. 영상 선택: 자연스럽고 명확한 영어가 사용된 YouTube 영상을 선택하세요. TED Talks, BBC 뉴스, 영화 장면, 팟캐스트, IELTS 모범 답변 영상이 좋습니다. URL을 복사해서 검색창에 붙여넣으세요. 짧은 영상(5분 이내)과 실제로 관심 있는 주제부터 시작하는 것이 동기 유지에 효과적입니다.
  2. 먼저 듣고 내용 이해하기: 처음에는 1배속으로 그냥 듣기만 하세요. 아직 따라 말할 필요는 없습니다. 문장의 의미를 파악하고, 화자가 어떻게 단어를 강조하고, 소리를 연결하고, 쉬어 가는지 주목하세요. 내용을 이해한 후 쉐도잉 연습을 하면 효과가 훨씬 좋아집니다.
  3. 쉐도잉 모드 설정:
    • Wait Mode (대기 모드): +3s 또는 +5s를 선택하면 한 문장이 재생된 후 자동으로 잠시 멈춰서 따라 말할 시간을 줍니다. 직접 컨트롤하고 싶다면 Manual을 선택해서 Next를 눌러 진행하세요.
    • Sub Sync (자막 동기화): YouTube 자막이 오디오와 맞지 않을 수 있습니다. ±100ms로 조정해서 정확한 타이밍에 따라갈 수 있도록 맞추세요.
  4. 소리 내어 쉐도잉하기 (핵심 연습): 이것이 연습의 핵심입니다. 문장이 재생되는 순간——또는 일시정지 중에——크고 자신감 있게 소리 내어 따라 하세요. 단순히 단어를 읽는 것이 아니라, 화자의 리듬, 강세, 음의 높낮이, 연음 방식을 그대로 흉내 내는 것이 중요합니다. 목표는 화자의 '그림자'처럼 들리는 것입니다. Repeat 기능으로 같은 문장을 여러 번 반복해서 자연스럽게 입에 붙을 때까지 연습하세요.
  5. 난이도 높이며 꾸준히 연습: 한 구절이 편해지면 더 도전적인 수준으로 올리세요. 속도를 <code>1.25x</code> 또는 <code>1.5x</code>로 높여 빠른 언어 반사 신경을 훈련하세요. Wait Mode를 <code>Off</code>로 설정해서 연속 쉐도잉을 하는 것이 가장 고급스럽고 효과적인 모드입니다. 매일 15~30분씩 꾸준히 연습하면 몇 주 안에 눈에 띄는 변화를 느낄 수 있습니다.

☕ 커피 한 잔 사주기

PayPal로 기부하기