跟讀練習: What Stanford students actually use ChatGPT for - 通過影片學習英語口語

選擇練習模式

每種模式都會開啟全屏練習室。

282 句

如果句子過短或過長，請點選 Edit 進行調整。

30 seconds.

10 seconds.

Question number

one.

Question number one.

What kind of work do you delegate to ChachiPT, and what kinds of work do you keep for yourself?

Lydia kind of touched on this a little bit.

Gabriel.

I think I'm trying to give it as much grunt work as I can.

And not even grunt work that, you know, I'm like, okay, I have to do this for a class, but I'm gonna pass it off.

If I have to do it for the class, I'm gonna do it.

But, like, I was writing a research paper the other day, And this goes back to the last question.

My teacher specifically in the syllabus says, use Chachupiti to help you with research.

If you find a really long research article, throw it in, get a summary, then read it so you find the good nuggets.

Because it's not worth looking through a 60-page article for one little thing

that you never find that Chachupiti could have found in 30 seconds.

I don't think there's a critical thinking aspect of that process that you're losing out to Chachupiti.

But even when I get stuck in an essay, and I'm like, okay, there's one sentence that I need to refigure.

I don't give it to chat, because I want to be able to have that critical thinking and keep it to myself.

And I don't want to get into a practice of delegating little tasks that require even minimal amounts of critical thinking, because that way I feel like I'm learning.

And back to the last question is, like, are you learning faster than the curriculum?

I don't think I'm learning faster than the curriculum, but I think ChatGBT is.

And so if I give it to ChatGBT, it's learning, and I'm not.

So I still want to be in the shoes where, you know, my brain's doing the thinking.

But when my brain is doing thinking that already 100% knows how to do, then I can pass.

Got it.

I'm curious about where that line is with grunt work.

Like what technically qualifies as grunt work?

Like, does anyone else have anything to add to this idea of like, where do you draw the line between something that, okay, fine, I'll just pass it off to ChatGPT versus something that's like, no, I know it's a small sentence,

but I'm going to push through this one.

I'm going to figure it out on my own.

Yes, Eric.

Yeah, to me, it's like the difference between knowledge and decision.

I keep the decisions to myself, but if I need knowledge about something or information, I would ask ChatChapuKee for it.

But if it's something like I have to make a decision which sentence to change my essay, then I'll probably do it myself.

Actually, I independently came to the same opinion as Eric.

The framework I added under was kind of like teaching, like using AI to teach you versus to solve for you.

So I think that there's a lot of cases where like, even if there's like a problem that like, I didn't know how to do this problem, right?

I don't know if it's like too bad to even just like say, how do you solve this problem?

More problems like this.

That's like very close to the line, I think.

But like, that's another application.

But also I think there's this other like kind of dichotomy that we weren't talking about, which is like meaningfully like consulting chat GPT when you don't necessarily have a problem.

So like kind of like, oh, you know, like the chat GPT reflex, right?

Like, oh, I have a question.

Why didn't I actually use ChatGPT now and solve that?

Instead of, I have a homework problem, now I have to decide, do I want to use ChatGPT or not?

I think there's this whole other use case that we maybe didn't talk about as much.

I think throughout the day, more and more, I've come to realize that there are times that I can use ChatGPT to help me,

even though it's not related to a specific homework assignment or a task like a research paper I need to write, but just a question, like a curiosity.

Or maybe even like, oh, I want to start journaling maybe I can use ChatGPT and I'll just journal now because I can use ChatGPT.

It's a good framework.

Yeah, Aliana.

I would say, I'm just so honest, I don't know what work I keep to myself because I feel like ChatGPT has taken over my life. But...

That's the real...

For example, if we go back to the topic of writing an essay, I would definitely keep the structure and inspiration from myself.

Like, I'm going to be the one providing what style, what thoughts, opinions I'm looking for in this essay.

but like I'm gonna write him a draft, like write a child-gibb-d draft for sure,

but like rephrasing, polishing, and also elevating the style or these type of things, it's definitely something that I won't put that much time in,

but it doesn't mean I'm just gonna give my professor a child-gibb-d generated essay.

I would still look at the essay to learn from its

language with how was my draft different from the one child-gibb-d gave it to me, and how were the language that ChatGD wrote better than the one I wanted.

I learn and adapt to the language it provides to me, but like what Eric says,

I'm still making a decision of which one I want to use.

But I do think I learned a lot from my conversations with Chat

because I realized that now I talk more professionally or use more words that are AI generated,

but in a way it really phrased my goals well or like phrased my purposes better, if that makes sense.

Okay, you're providing the outline, ChatGPT is providing the result.

What if that result is, you look at it and you're like, yeah, sure, that's pretty good.

Pass it off.

Is that the essay?

Yeah, yeah.

It's kind of like learning, like learning the ChatGPT as well.

It's like, it learns my habits and I learn its habit.

And I like improve myself based on what I learned from Chat.

Got it, love that.

Yes, Nick.

So along the similar line, I would say I want to keep as much of academic work to myself as possible,

because these are skills that I would actually like to learn and work through myself, because there might be a moment in an interview or in a job where I might not have access to AI,

and I actually did this myself.

But for personal projects, it's really good for getting the ball rolling.

It can take you from 0 to 25 really easily

if you give it a bullet list of points that you want to do.

And so I feel like that's like your own personal curiosity and like your own personal projects.

That is the kind of work that I would want to delegate first and then I can pick it up by myself.

Totally.

Has AI impacted your decision of what to major in?

So I don't think it's necessarily the major that teaches us what we want to do.

I think it's like all the experiences like outside of classes as well that kind of help us learn more about ourselves.

So I want to do startups personally.

If I look at a lot of people in the startup industry like today, A lot of them don't go to college, a lot of them are high school dropouts.

And I feel like the reason why they're able to succeed is because they know what they want to do.

So I feel like kind of looking at this question, I feel like what's more important is, do you know what you want to do with your major?

Like, do you know what kind of life you envision for yourself?

So I feel like if you know the answer to that question, I think it's more important to know how to do AI.

Yep, love that.

Absolutely.

So I'm an international relations major and, you know, at Stanford, it's like STEM versus humanities oftentimes, I feel like.

100

But I realize I remain committed to my major, but my interests within my field have changed.

101

And so instead of focusing on, you know, geopolitical conflicts, I'm trying to look at tech

102

or how AI can influence a lot of national security issues, et cetera.

103

Okay, so how AI impacts the field that you already wanted to study.

104

Millie, I saw you nodding a lot throughout the last two.

105

Do you want to add to that?

106

Yeah, I agree with Drew.

107

I personally also switched from CS to EE, and the reason is seeing CS being so good at coding.

108

I started having like, exist enterprises, seeing it doing a better job than I do.

109

So I feel like there are maybe other skills that maybe AI cannot do.

110

But now I'm rebutting myself because I realized that, oh, actually it's getting so much better too as I'm getting better like am I like other stuff?

111

So I would say like for me right now, it's like I see it as either you like be really good at CS

112

and like invent like advanced a frontier of AI or like you use AI, like use it as like an application and you don't need to be like

113

that deep into like the core behind AI

114

if you want to like just like pull APIs and like use it on the applications.

115

So I go to a top where I could also like use a lot of AI.

116

Love it, love it.

117

You don't need to know how to build a large language model yourself

118

if you want to take advantage of all the benefits.

119

Okay, cool.

120

Yeah, so I guess building on top of that, obviously your major is important and learning how to learn is also important.

121

So I think having academic rigorous classes is good for you.

122

But I think in a world where we have chat to see our fingertips

123

and where anyone can be technical if they just buy code, I think what sets people apart is the people skills like growing up like I was really shy

124

and like I always had a hard time talking to people

125

and so I think

126

that in this like new like world where we have such

127

a technical like aspect to each of us it's more important

128

to also like be able to express yourself clearly um and so like one of the

129

most interesting classes I'm in right now is like an improv class um actually with Roshan

130

but um I I think in that class, I learned a lot more about expressing myself,

131

thinking of my feet, and other skills that you don't necessarily think of as academic, but still very important for your life.

132

Fascinating.

133

You happen to be getting out of an academic class, but that's also something you can get out of a club, an extracurricular, something like that,

134

which is interesting, because I think it kind of brings up maybe the major isn't the solution to the, how do I learn these other skills thing, necessarily.

135

Do you trust that professors know how to teach in an AI world, or do you feel like you're learning faster than the curriculum update?

136

Emily.

137

So this class called Spy, Fly, and Algorithms, it's a class on American intelligence.

138

And the professor this year just decided to try something new.

139

We have to write a memo on the nuclear conflict in North Korea.

140

And last year, like, the professor addressed the fact that people could just chat their entire memo.

141

And so this year, she literally taught us how to use AI to help us find the right sources to create our memo.

142

And a lot of our assignments are now structured around how are we interacting with AI to create this memo.

143

And then part of our

144

that memo assignment is now talking about like how did you interact with AI to find the sources,

145

what were the gaps in the research or the sources that AI gave you, and you have to like actually talk it out with your TA.

146

And so now there's a lot more like AI proof initiatives in my classes that you notice.

147

Love that.

148

More AI proof initiatives.

149

It's similar to what Emily was talking about, about her professor given the opportunity to do this.

150

In one of my classes, they actually require you to use ChatGPT.

151

Or like not only ChatGPT, but also like, for example, you have to use ChatGPT for specific tasks.

152

You also have to use notebook L.M for like analyzing like large numbers of papers.

153

You have to use, like for example, we have to conduct a lot of interviews

154

and you have to ask people who you take interviews with and transcribe it with AI.

155

So I think also sometimes there are those professors who just give you opportunity to learn with AI

156

and I think some of my professors do it now

157

because we're going to graduate to the world where we're going to have AI

158

and I think that is really crucial to just give us this chance

159

and requirement as well yeah

160

so actually requiring you to use the ai not just uh wondering

161

if it's happening in the background yeah

162

so i feel like i align more with the second point like i worked on a project in high school

163

that like allowed students to like use uh like passing a youtube video

164

and then like learn

165

that material like online uh by like passing it to chat with you through like a transcription

166

but i feel like that just allows people to learn

167

so much faster right

168

but you have an available tutor at all times right it

169

doesn't your professor needs to take some time to respond to all of his emails, right?

170

So this is just immediate, right?

171

And then you can get like quiz questions, all of this material just like on the fly, right?

172

So I think that sort of experience of just learning immediately and then maybe somebody with a greater like, you know, depth of that subject, maybe you go to them later,

173

but I think you can learn much faster through the use of AI.

174

Do you think most of your peers use ChatGPT that way?

175

I unfortunately don't think so.

176

I think the majority of people will just let it like supplement everything they do, But I think if we have a focused environment where you are forced to learn it,

177

along with ShadDZ giving you some information and hints, that's the best way to learn.

178

There needs to be a critical thinking aspect to it.

179

Cool.

180

Who else resonates with that or wants to offer a kind of perspective?

181

Yeah.

182

Theoretically, I think AI can definitely be used as a tool

183

that can help all of us learn a lot quicker than we have been learning.

184

But what I noticed from some of my friends is that it doesn't necessarily supplement it.

185

Sometimes they use it to fully replace it.

186

And so I think that's part of the dangers with using really, really powerful models like ChatGBT,

187

because they're so good now that it can be used to solve a whole lot of questions.

188

And we should be using it as a tool instead of something that fully replaces our thinking.

189

Okay, cool.

190

Yeah.

191

I'd argue that the assignments that professors have been setting haven't been changed like what they've been doing in previous years.

192

And I feel like they've been set in a world that didn't have AI.

193

And if the goal of college is to prepare people for a future workforce where you have AI access to knowledge,

194

your fingertips, then the learning should be like an assessment should be given in a way that limits that.

195

So they could prepare for a workforce where the barrier is in knowledge, you know, inaccessibility.

196

But it's how do you put information together on a pretty big place?

197

Okay.

198

Okay, if the purpose of college is to prepare us for the workforce after college,

199

Stefan, Che, earlier, the purpose of college is to learn about who we are and learn about ourselves.

200

Just pinning these on the board as we talk

201

because we're going to come back to this idea of the purpose of college here in one of our later questions.

202

When AI can assist in essays, projects, code, how should this university measure your success, your intelligence?

203

I think there's two parts of this.

204

I think one is like the high school level one where they're like, okay, AI, it can assist, so we're not going to allow it to and do your other essays in class.

205

You know, that was like a huge switch.

206

And now I think that's being reassessed as in, okay, this tool is like actually going to exist.

207

So we have to figure out how, you know, we incorporate that.

208

But then I think at the university level, you know, again, you can't ignore it.

209

It's going to be a part of things.

210

But then also, like everything, AI is still a product, you know, and there's still a lot of profitability within it.

211

And companies like OpenAI have different price models.

212

So I think part of it is also exploring inequities within students' ability to use AI.

213

If everyone has access to base, but then someone else is premium, and they're all being graded the same, and they're allowed to use Chats of ET,

214

but one person can go through 400 articles of five minutes, and other people can only use three, is that fair?

215

So should Stanford, I think like, I have an excellent, should Stanford be paying for everyone to have at least some premium version

216

so that everyone has an equal base playing field.

217

And then when you're measuring success, it's not based on can you afford 20 extra dollars a month to get a premium or not.

218

I mean, it depends on how you measure success.

219

Like if you're trying to evaluate how well a person could ask JAT-GPT to implement an algorithm, then you should be able to use AI in your class, right?

220

But if you're taking a class like 106b, where you're learning about all sorts of algorithms and the objective

221

and kind of expectation is that you're going to be able to finish that class and be able to implement them yourself.

222

I do believe that it still makes sense to kind of like evaluate

223

and test students on their ability to come up with this in an environment devoid of AI.

224

Yeah, I think it's just a reality that we have to deal with.

225

We're going to have to use AI.

226

We need to get good at it.

227

And I think the way to measure

228

that is to say like how well do you understand what you just built in two seconds, right?

229

If you've If you've got code of something, how much do you understand what it's good at, what it's bad at?

230

Can you debug if there's an issue?

231

I think that's what they should be measuring.

232

If you build something quickly, we should be asked questions about what we built, maybe find something really quick.

233

That's what they should be testing.

234

They should be testing how well you understand the thing that you just built with AI.

235

Your generation has the most powerful tools and yet perhaps the most uncertain future.

236

Very plainly, how does this make you feel?

237

Yeah, I feel both excited and scared.

238

And it could be like full of excitement or full of fear, depending on the day and depending on the time.

239

It's so weird because I feel like I'm so excited in the sense

240

that the first time that I used AI to generate a website

241

or generate an image or I was so annoyed that I have to go and design this poster.

242

But now with ChachiPT, I could design a flyer instantly.

243

I was just so relieved and so excited that I could spread this message.

244

But then sometimes I also feel very scared

245

because like I just don't know like AI is able to do all these things

246

that I've been learning about like web development for, for, you know, like spending three years like really honing my skills.

247

What am I going to do in the future?

248

Like is AI going to like continuously improve and take over some of this kind of cognitive work?

249

And so what's very interesting is that like, what I know from psychology is that like, when we are both like scared and when we're both excited,

250

we're actually engaging the same kind of autonomic nervous system.

251

And it's really a matter of perspective sometimes that, what kind of emotion that we're feeling.

252

And so I think it's just interesting.

253

It's like very fluctuating and it's very, very interesting times.

254

Cool.

255

I love that spectrum, that kind of binary, excited, excited and afraid.

256

Obviously there's so much nuance here, but I'm on purpose gonna make it really black and white.

257

I'm gonna ask, raise your hand if you feel more excited than afraid.

258

And then I'm gonna say, raise your hand if you feel more fearful than excited.

259

Okay, so think to yourself which one you're gonna raise.

260

Okay, raise your hand if you feel more excited than afraid.

261

Okay, raise your hand if you feel more afraid than excited.

262

Okay.

263

Let's hear two voices from each.

264

why I feel like I'm scared of it, but also at the same time, like I know that like being scared of it like does me no good.

265

Like all I need to do is like to look at the solution to see like, like so yeah, that is like the fact now.

266

So where do we go from there?

267

So I feel like I'm just like more cautious, but also at the same time, I understand that this is like just reality.

268

Totally.

269

Well, I appreciate you guys being willing to chat that side because I know I forced you into an impossible binary.

270

Final thoughts, Sid.

271

To add on to what Neha said, I think it's interesting because I think our end goal has now, like, we're thinking too small.

272

Like, we need to think of, like, really broad end goals.

273

And, like, sure, building a website right now, that might be an end goal.

274

But tomorrow might be to build, like, you know, hundreds of websites every day.

275

And, like, you know, I think we think too small.

276

Like, this is more efficiency, so therefore we should be able to push that end goal further and do so much more and tackle those big problems, like, exploring other planets, you know, stuff like that.

277

So think more broadly.

278

I love it.

279

Thank you so much you guys.

280

That is our last question which brings us to the end of our presentation.

281

Give yourselves a round of applause.

282

A hand or a round of applause.

下載應用

流利口語所需的一切

AI 發音評分逐句打分

音標練習掌握每個音

詞彙學習建立你的詞庫

詞彙遊戲邊玩邊學

掃碼

Download on theApp Store Get it onGoogle Play

FOR YOU

本片段的口語練習目標

這段對話聚焦於日常交流中的邏輯表達與觀點闡述，非常適合提升雅思口語練習中的“觀點組織”能力。通過模仿對話者清晰劃分任務邊界（如區分“ grunt work”與“ critical thinking”）的表達方式，學習者能掌握如何有條理地說明覆雜想法，同時鍛鍊即時回應的流暢度，這正是影子跟讀（shadow speak）訓練中核心的“語流同步”與“邏輯復刻”能力。

實用短語庫

delegate...to...：將……委託給……
grunt work：枯燥的基礎工作
critical thinking aspect：批判性思維層面
draw the line between...and...：劃分……與……的界限
come to realize：逐漸意識到

這些短語在雅思口語和日常交流中高頻出現，通過影子跟讀（shadow speech）反覆模仿其發音與語調，能快速提升表達的地道性。

突破口語薄弱點

對話中大量使用“轉折連詞”（如but、however）和“插入語”（如I think、you know），形成自然的口語節奏。許多學習者在影子跟讀時容易忽略這些細節，導致表達生硬。建議通過“shadowspeaks”訓練法，逐句模仿對話者的停頓、重音和語調變化，尤其注意“but”後的語氣轉折。例如，當說到“but I want to be able to have that critical thinking”時，“but”需稍作停頓，“critical thinking”加重語氣，這樣才能準確傳遞語義重點。長期練習可有效改善“中式英語”的生硬感，提升口語的流暢度與自然度。

影片中的語法

說話人最常用的結構，並附上影片中的原話：

結構	影片中的用法
現在完成進行時 have/has been + -ing — 以前開始、現在仍在繼續的動作	have been learning · have been setting · I've been learning
被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的	are now structured · are forced · been changed
現在完成時 have/has + 過去分詞 — 過去發生但與現在仍有關聯的事	I've come · has taken · have changed
定語從句 who / which + 從句 — 補充說明人或事物	about, which is · people who you

什麼是跟讀法？

跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧，最初開發用於專業口譯員的培訓，並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大：您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同，跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性，語調，節奏，連讀，聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。

影子跟讀法: 閱讀完整分步指南 →

跟讀練習: What Stanford students actually use ChatGPT for - 通過影片學習英語口語

選擇練習模式

下載應用

推薦影片

5 Things You Should Never Ask ChatGPT | Vantage with Palki Sharma

Fight brain-rot with Worked Examples

Translating Claude’s thoughts into language

13k

Go’s Big Mistake

Prompt Engineering - Guidelines

Are we heading to an era of stagnation or innovation?

Introducing GPT-6 Astra for developers

How to Lock In and Actually Get Things Done | Andrew Huberman

AI can do your homework. Now what?

Is it time to dismantle the puzzle?

Angela Lee Duckworth The key to success Grit

本片段的口語練習目標

實用短語庫

突破口語薄弱點

影片中的語法

什麼是跟讀法？