跟读练习: Satya Nadella demos an app he built | Microsoft AI Tour Bengaluru - 通过YouTube学习英语口语
C2
You know, we had a Thanksgiving weekend in the US a few weeks ago.
106 句
如果句子过短或过长，请点击 Edit 进行调整。
1
You know, we had a Thanksgiving weekend in the US a few weeks ago.
2
And so I had a chance to say like, "what else can one do over Thanksgiving other than build?" And so I built an app of my own, using all of the stuff that Karan was showing.
3
So this is my Azure environment.
4
And by the way, this is my regular PC that I travel with, so hopefully nothing happens.
5
But, and so this is, it's in fact, I have this app deployed, I think, in, south central Canada.
6
This is my GitHub repo.
7
And, it's fun, right?
8
So I kind of have my typical setup, in fact, is, Windows 365, which travels with me essentially everywhere.
9
And then in there, obviously I have my GitHub, and then it's Codespaces.
10
So it's kind of like turtles all the way.
11
So you have your code spaces running on, GitHub in Windows 365 instance.
12
And then, the idea that now you can go in.
13
And my favorite thing of course, to do is your, you know, come in, in the morning and just start issuing, whatever coding tasks.
14
And so this is where I go and just, I usually fire off five or six things.
15
It creates, five or six draft branches.
16
And then ultimately, at the end of the day, I go back and mostly delete the branches, but there is a PR or two, I'll accept and go on and, work with it.
17
So it's fun.
18
And then the fact that you have all these models, in fact, I think now, I'm mostly using, a lot more codex-max, it's fantastic.
19
It's fast. I'm using obviously Claude.
20
So, Claude Opus 4.5 as well.
21
But, the thing that I've now gotten used to is I have enough trust to just say auto and it picks.
22
And so, if I'm just, really, I don't I want to really be efficient with my token limits.
23
And so therefore picking auto seems to be really, really a good way to go about it.
24
So anyway, so great I did all this, so what the heck did I build?
25
What I said is, okay, what's my dream?
26
My dream is to figure out how to get a job, in this Copilot team.
27
So I said, Matt, I like your deep research stuff, but I want to add a lot more to deep research, so I build my own deep research.
28
And so, with all these models available, I said, okay, what if I could start putting new decision frameworks?
29
So one decision framework, Ondrej Karpathy recently talked about this LLM council, which I love a lot.
30
So I implemented this idea that you can now have all the models available to you.
31
So GPT, Opus, Gemini, Kimi K2, Grok, what have you, all of these models.
32
And then you can select a chairman.
33
So you have council members or the selection committee.
34
And then you have a chair.
35
And then you can go issue any query and have it come back and tell you what it thinks.
36
Then another decision framework I implemented was this thing called DxO.
37
We did this actually in healthcare first.
38
So this was done. There you go.
39
Now let's see.
40
Let's see how all my oh, it's so fantastic.
41
So, DxO is a thing, as I said, we implemented, for healthcare and you have specific roles.
42
So you have a lead researcher.
43
And the lead researcher in this case is Opus.
44
It sort of does the breadth first research.
45
Then you have another role, which is a critical reviewer.
46
In this case, I am selected 5.1 GPT 5.1 And their role is to find any method errors.
47
Especially bias and recency bias, what have you.
48
Then we have a data analyst.
49
So I picked or I picked a domain expert Gemini I then said data analyst, I picked Kimi K2.
50
So, this when we published the DxO paper, it performed better than any one frontier model.
51
So this is in the context of very high stakes health outcomes.
52
And I said, hey, I want same thing for any decision.
53
I want to make. So I implemented that.
54
So I implemented another one as well called Ensemble.
55
So this is basically use all models and as just essentially a set of MCP servers, anonymize the responses.
56
So take out even who is responding with what.
57
Give them, alpha, beta, gamma and then synthesize into one, response.
58
So these are three decision frameworks.
59
In fact, I even extended it, by the way, I built a shopping thing.
60
I built even a finance thing, but basically decision frameworks.
61
And then of course, like a good sort of crazy South Asian, cricket fanatic.
62
What do you use it for?
63
To select the all time best Indian test cricket team, especially in a time like this.
64
After what happened in the last test series, I think it's time to get to work.
65
So what I did is I'll show you the history side if I go ahead.
66
And in fact, you will see, that mostly that's what I've been using it for, which is the MLB lineup was also just crazy.
67
It is fantastic.
68
But the test team, so I let me go show you some of the stuff.
69
So this is the, what happened.
70
This is the chairman synthesis.
71
So this is the AI Council, so it came back and it says, you know what?
72
I figured out all of the, Sunil opens Sehwag opens with him.
73
Dravid makes it obviously and what have you.
74
But look at this areas of complete consensus.
75
is Gavaskar, Sehwag, Dravid, Tendulkar, Kohli, Kapil, Dev, Ashwin, and Bumrah.
76
Key debates.
77
Oh man. VVS, do you have him or you don't?
78
And look at the way it's sort of, made the decision.
79
Phi-1 basically said and Claude, inclusion or Laxman was heavily weighted because of the crisis management.
80
And as a good Hyderabadi, I love GPT 5.1 and Claude.
81
And of course Kumble would say, "this is pretty cool".
82
Do you really need left arm swing or you need, whatever you call Kumble bowling.
83
Which is obviously the stats one out, 619 wickets and, and so they selected and oh, and then captaincy debate.
84
Kohli vs Dhoni and they selected Kohli.
85
And so it goes off and then annotates.
86
What I love by the way is I implemented it as even a streaming thing.
87
it is not deployed in south central Canada.
88
I would sort of show it to you, but, it's just nice to see.
89
It's essentially like a chain of debate, not a chain of thought.
90
So I can see the morals debate and then synthesize.
91
So anyway, so that's one example.
92
DxO is another one.
93
This is very neat because, what it does is you can see by raw, right.
94
So I see the first the exhaustive search of the lead researcher.
95
It gives me again what did, what are all the things the critical reviewer will find like, okay, what are the method problems and gaps and weaknesses.
96
So for example, era bias.
97
Classic thing that happens, when you sort of compare across generations, you make all kinds of mistakes because you don't adjust for, any of the stats and the difficulty of the wickets, I mean, walking out in a West Indies or an English wicket there is not covered.
98
I mean, how does one play even in any anyway?
99
So to be able to take all of that into account and then see the debate between the various models to resolve.
100
So anyway, so I've had and then the same thing with ensemble as well.
101
So the point I wanted to make was I built this over maybe a couple of hours, and now I'm constantly refining it.
102
In fact, one of my hopes, as I said, is to be able to I by the way, this is all going to come to Copilot, and I'm convincing my friends and they're saying, yeah, you can apply as a junior product manager if you are competent enough.
103
So I'm still in the process of interviewing, but not that said, I think all of this will come because this is to me the next generation of metacognition.
104
So if you think about these decision frameworks, you have all these agents, you're working with agents, but the metacognition is still us.
105
And this is tools for metacognition is how I think about it.
106
And so, to me, building out these type of frameworks, building out these type of agents is interesting to select the Indian cricket team, but think about whether it's a supply chain decision, whether it's a healthcare decision, whether it is a finance decision, these type of chain of debates with multiple agents participating, is going to be a lot of what we are going to all, build in our systems, in our agentic system.
下载应用
AI 为你说出的每个句子打分
TRENDING
热门
背景与上下文
在近日的一场活动中，微软首席执行官萨提亚·纳德拉展示了他在感恩节假期期间所开发的一款应用程序。他在视频中分享了自己的开发过程，探索了云计算和代码管理的最佳实践。这段对话展示了他如何通过使用不同的人工智能模型来增强决策框架，并提到了一些他在开发中所使用的工具与方法，对于希望提升英语口语技能的学习者来说，提供了很好的上下文和灵感。
日常交流的五个常用短语
- What else can one do（还可以做些什么）
- start issuing coding tasks（开始发布编码任务）
- trust to just say auto（相信自动选择）
- what the heck did I build（我到底建了什么）
- perform better than any one frontier model（表现优于任何前沿模型）
通过这些短语，学习者能够在日常生活中和职场环境中更流畅地交流，并提升他们的英语口语练习能力。
逐步跟读指南
在视频中，萨提亚使用了许多专业术语和复杂的句子结构，可能对某些学习者构成挑战。以下是一些建议，以助于克服这些困难：
- 收看视频时启用字幕：这可以帮助你在听的同时看到文字，理解内容。
- 暂停并重复听：对于较难的段落，暂停视频并重复听几次，可以更好地消化信息。
- 跟读练习：尝试和萨提亚同步跟读，模仿他的语音语调，这对于提升发音和流利度非常有效。
- 记录新词汇和短语：在观看过程中，把不懂的词汇和短语记录下来，并进行查阅和背诵。
- 定期复习：反复观看视频中的某些片段，巩固记忆，提高你的英语口语水平。
通过看YouTube学英语，学习者可以有效地利用现代技术，提升自己的英语能力，尤其是通过shadowspeaks和shadow speak的练习，使他们在雅思口语练习中获得更大的成功。
什么是跟读法？
跟读法 (Shadowing) 是一种有科学依据的语言学习技巧，最初开发用于专业口译员的培训，并由多语言者Alexander Arguelles博士普及。这个方法简单而强大：您在听英语母语原声的同时立即大声重复——就像是一个延迟1-2秒紧跟说话者的影子。与被动听力或语法练习不同，跟读法强迫您的大脑和口腔肌肉同时处理并模仿真实的讲话模式。研究表明它能显着提高发音准确性，语调，节奏，连读，听力理解和口语流利度——使其成为雅思口语备考和真实英语交流最有效的方法之一。