跟讀練習: AI Agents, Clearly Explained - 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。
137 句
如果句子過短或過長，請點選 Edit 進行調整。
1
AI AI AI AI AI AI - You know, more agentic.
AI AI AI AI AI AI——你知道的，就是更具代理能力。
2
- Agentic capabilities.
3
- AI agent.
4
- Agents.
5
- Agentic workflows.
6
- Agents.
7
- Agents.
8
- Agents.
9
- Agents.
10
- Agents.
11
- Agents.
12
- All right, most explanations of AI agents is either too technical or too basic.
13
This video is meant for people like myself.
14
You have zero technical background, but you use AI tools regularly,
15
and you wanna learn just enough about AI agents In this video, we'll follow a simple one,
16
two, three, learning path by building on concepts you already understand, like ChatGPT, and then moving on to AI workflows,
17
and then finally, AI agents, all the while using examples you will actually encounter in real life.
18
And believe me, I wanna tell you, those intimidating terms you see everywhere, like RAG, RAG, or React, they're a lot simpler than you think.
19
Let's get started.
20
Kicking things off at level one, large language models.
21
Popular AI chatbots like ChatGPT, Google Gemini, and Cloud are applications built on top of large language models, LLMs.
22
And they're fantastic at generating and editing text.
23
Here's a simple visualization.
24
"You, the human, provides an input, and the LLM produces an output based on its training data." For example,
25
if I were to ask ChachiPT to draft an email requesting a coffee chat, my prompt is the input,
26
and the resulting email that's way more polite than I would ever be in real life is the output.
27
So far, so good, right?
28
Simple stuff.
29
But what if I asked ChatGPT when?
30
"my next coffee chat is." Even without seeing the response, both you and I know ChatGPT is gonna fail because it doesn't know that information.
31
It doesn't have access to my calendar.
32
This highlights two key traits of large language models.
33
First, despite being trained on vast amounts of data, they have limited knowledge of proprietary information, like our personal information or internal company data.
34
Second, LLMs are passive.
35
Keep these two traits in mind moving forward.
36
Moving to level two, AI workflows.
37
Let's build on our example.
38
What if I, a human, told the LLM, every time I ask about a personal event,
39
perform a search query and fetch data from my Google Calendar before providing a response.
40
With this logic implemented, the next time I ask, when is my coffee chat with Elon Husky,
41
I'll get the correct answer because the LLM will now first go into my Google Calendar to find that information.
42
Here's where it gets tricky.
43
What if my next follow -up question is, what will the weather be like that day?
44
The LLM will now fail at answering the query
45
because the path we told the LLM to follow is to always search my Google Calendar, which does not have information about the weather.
46
This is a fundamental trait of AI workflows.
47
They can only follow predefined paths set by humans.
48
And if you want to get further.
49
What if I added more steps into the workflow by allowing the LM to access the weather via an API, and then just for fun, use a text to audio model to speak the answer.
50
The weather forecast for seeing Elon Husky is sunny with a chance of being a good boy.
51
Here's the thing.
52
No matter how many steps we add, this is still just an AI workflow.
53
Even if there were hundreds or thousands of steps, if a human is the decision maker, there is no AI agent involvement.
54
Pro tip, retrieval augmented generation or RAG, is a fancy term that's thrown around a lot.
55
In simple terms, RAG is a process that helps AI models look things up before they answer, like accessing my calendar or the weather service.
56
Essentially, RAG is just a type of AI workflow.
57
By the way, I have a free AI toolkit
58
that cuts through the noise and helps you master essential AI tools and workflows.
59
I'll leave a link to that down below.
60
Here's a real world example.
61
Following Helena Liu's amazing tutorial, I created a simple AI workflow using make .com.
62
Here you can see that first, I'm using Google Sheets to do something.
63
Specifically, I'm compiling links to news articles in a Google Sheet, and this is that Google Sheet.
64
Second, I'm using perplexity to summarize those news articles.
65
Then, using Claude and using a prompt that I wrote, I'm asking Claude to draft a LinkedIn and Instagram post.
66
Finally, I can schedule this to run automatically every day at 8 :00 AM.
67
As you can see, this is an AI workflow because it follows a predefined path set by me.
68
Step one, you do this.
69
Step two, you do this.
70
Step three, you do this.
71
And finally, remember to run daily at 8 :00 AM.
72
One last thing, if I test this workflow and I don't like the final outputs, of the LinkedIn post for example,
73
as you can see right here, it's not funny enough.
74
And I'm naturally hilarious, right?
75
I'd have to manually go back and rewrite the prompt for Claude.
76
Okay.
77
And this trial and error iteration is currently being done by me, a human.
78
So keep that in mind moving forward.
79
All right, level three AI agents.
80
Continuing the make .com example, let's break down what I've been doing so far as the human decision maker.
81
With the goal of creating social media posts based off of news articles, I need to do two things.
82
First, reason or think about the best approach.
83
I need to first compile the news articles, then summarize them, then write the final posts.
84
Second, take action using tools.
85
I need to find and link to those "use perplexity" for real -time summarization, and then "clawed" for copywriting.
86
So, and this is the most important sentence in this entire video, the one massive change that has to happen in order for this AI workflow to become an AI agent,
87
is for me, the human decision maker, to be replaced by an LLM.
88
In other words, the AI agent must reason.
89
What's the most efficient way to compile these news articles?
90
Should I copy and paste each article into a Word document?
91
to compile links to those articles, and then use another tool to fetch the data.
92
Yes, that makes more sense.
93
The AI agent must act, aka do things via tools.
94
Should I use Microsoft Word to compile links?
95
No, inserting links directly into rows is way more efficient.
96
What about Excel?
97
No, the user has already connected their Google account with make .com, so Google Sheets is a better option.
98
Pro tip: because of this,
99
the most common configuration Agents must reason and act, so re -act.
100
Sounds simple once we break it down, right?
101
A third key trait of AI agents is their ability to iterate.
102
Remember when I had to manually rewrite the prompt to make the LinkedIn post funnier?
103
I, the human, probably need to repeat this iterative process a few times to get something I'm happy with, right?
104
An AI agent will be able to do the same thing autonomously.
105
In our example, the AI agent would autonomously add in another LLM to critique its own output.
106
Okay, I've drafted V1 of a LinkedIn post.
107
How do I make sure it's good?
108
Oh, I know.
109
I'll add another step where an LLM will critique the post based on LinkedIn post practices.
110
And let's repeat this until the best practices criteria are all met.
111
And after a few cycles of that, we have the final output.
112
That was a hypothetical example.
113
So let's move on to a real world AI agent example.
114
Andrew Ng is a preeminent figure in AI and he created this demo website that illustrates how an AI agent works.
115
I'll link the full video down below, but when I search for a keyword like skier, Enter.
116
The AI vision agent in the background is first reasoning
117
What a skier looks like a person on skis going really fast in snow for example, right?
118
I'm not sure and then it's acting by I looking at clips in video footage,
119
trying to identify what it thinks a skier is,
120
indexing that clip and then returning that clip to us.
121
Although this might not feel impressive, remember that an AI agent did all that instead of a human reviewing the footage beforehand,
122
manually identifying the skier and adding tags like skier, mountain, ski, snow.
123
The programming is obviously a lot more technical and complicated than what we see in the front end, but That's the point of this demo, right?
124
The average user like myself, wants a simple app that just works without me having to understand what's going on in the backend.
125
Speaking of examples, I'm also building my very own basic AI agent using NAN.
126
So let me know in the comments what type of AI agent you'd like me to make a tutorial on next.
127
To wrap up, here's a simplified visualization of the three levels we covered today.
128
Level one, we provide an input and the LM responds with an output easy.
129
Level two, for AI workflows, we provide an input and tell the LLM to follow a predefined path that may involve in retrieving information from external tools.
130
The key trait here is that the human programs a path for the LLMs to follow.
131
Level three, the AI agent receives a goal, and the LLM performs reasoning to determine how best to achieve the goal,
132
takes action using tools to produce an interim result, observes that interim result, and decides whether iterations are required,
133
and produces a final output that achieves the initial goal.
134
The key trait here is that the LLM is the decision maker in the workflow.
135
If you found this helpful, you might want to learn how to build a prompts database in Notion.
136
- Bye.
137
Have a great one.
同頻道
B21
Write a Killer Résumé in the AI Era: 5 New Rules!
Interview EnglishWorking Adults
B212
Why Are You a Good Fit for this Role - How to Answer
American AccentBusiness EnglishInterview EnglishUniversity Students
B28
What is Your Biggest Weakness? (Ace this Interview Question!)
American AccentBusiness EnglishInterview EnglishUniversity Students
B23
How I'd Learn AI From Scratch in 2026 (skip the useless 80%)
American AccentBusiness EnglishUniversity StudentsWorking Adults
B26
Tell Me About Yourself (The BEST Way to Answer this Interview Question)
American AccentBusiness EnglishInterview EnglishUniversity Students
B29
Claude Design is Insanely Easy (even for beginners)
American AccentDaily ConversationWorking Adults
B22
Learn 95% of ChatGPT Work in Under 20 Minutes
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
Tell Me About Yourself - Structure a Strong Answer
American AccentBusiness EnglishInterview EnglishUniversity Students
B12
How to Get Started with Notion (without losing your mind)
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
Google’s AI Course for Beginners (in 10 minutes)!
American AccentBusiness EnglishUniversity StudentsWorking Adults
B27 Job Interview Questions to Prepare For!
American AccentBusiness EnglishInterview EnglishUniversity Students
B2Tell me a time You Handled a Difficult Situation (Interview Question)
American AccentBusiness EnglishInterview EnglishUniversity Students
FOR YOU
推薦影片
C112
25 句
Introducing the Agents API
Business EnglishUniversity StudentsWorking Adults
C246
81 句
Large Language Models explained briefly
American AccentBusiness EnglishUniversity StudentsWorking Adults
![What is AI and ML? [Pt 2] | Generative AI with .NET for Beginners](https://img.youtube.com/vi/Len8KRBFvGo/mqdefault.jpg)
B22
52 句
What is AI and ML? [Pt 2] | Generative AI with .NET for Beginners
Business EnglishUniversity StudentsWorking Adults
B28
59 句
An ex-OpenAI researcher just deleted language from the LLM...
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
207 句
NEW GoHighLevel Voice AI Agent Tutorial 2026 (BEST Features & Upgrades)
Business EnglishWorking Adults
C11
309 句
Claude Code: The Advanced Guide (99% of Devs Skip These Features)
Business EnglishUniversity StudentsWorking Adults
C111
98 句
What are AI Agents?
Business EnglishUniversity StudentsWorking Adults
B21
307 句
The New AI prompting Method Everyone Is Talking About: Loops (And How To Use Them)
Business EnglishWorking Adults
C15
270 句
Transformers, the tech behind LLMs | Deep Learning Chapter 5
American AccentBusiness EnglishUniversity StudentsWorking Adults
B25
333 句
The New Way of Making Content In The Age of AI
American AccentBusiness EnglishUniversity StudentsWorking Adults
背景與上下文
在當今快速發展的科技時代，人工智慧（AI）成為我們日常生活中不可或缺的一部分。影片中，講者向觀眾解釋了“AI代理”的概念，特別是對於那些沒有技術背景，但希望理解如何在生活中應用AI工具的人。這段對話涉及了大型語言模型（LLMs），其如何處理人類指令並生成相應的文本。當我們面對複雜的AI術語時，講者設定了一個逐步學習的結構，使觀眾能夠輕鬆理解並跟上。
日常交流的五個短語
- AI代理（AI Agent）
- 大型語言模型（Large Language Model）
- 工作流程（Workflows）
- 輸入與輸出（Input and Output）
- 個人事件（Personal Event）
逐步跟讀指南
如果你希望提高自己的英語發音並掌握影片中的內容，可以採用以下“影子跟讀（shadow speak）”的方法：
- 首先，選擇一小段影片內容：從這段影片中，找出講者解釋的一個句子，比如“人工智慧代理如何依賴大型語言模型進行互動”。
- 其次，反覆聽這段內容：認真聆聽講者的發音和語調，注意他們如何強調關鍵字。
- 接著進行模仿：在聽的過程中，試著跟著講者的節奏重複這一句。這是提高英語發音的有效方法，也是“英語影子跟讀”的一部分。
- 然後記錄自己的聲音：可以使用手機或電腦錄音，聽一聽自己的發音，注意與講者的差距。
- 最後進行反饋修正：根據錄音進行自我調整，重複這一過程，直到能夠流利地表達。
通過以上步驟，你不僅能夠掌握日常交流中的重要短語，還有助於提升你的英語口語能力。記得保持耐心，持之以恆，配合文章中提到的“影子說話（shadowspeak）”方式，將會更快提高你的發音水平。
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 |
|---|---|
| 被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的 | is meant · being trained · being done |
| 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 | has already connected · I've drafted |
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。