Shadowing Practice: The Future of Data Science - Data Science @ Stanford - Learn English Speaking with Video
בחר מצב תרגול
כל מצב פותח חדר תרגול במסך מלא.
402 משפטים
אם המשפטים קצרים או ארוכים מדי, לחץ על ערוך כדי להתאים אותם.
1
Stanford University let's do a quick round of introductions all right
2
I guess I'll start I'm Ewan Ashley I'm a cardiologist in the School of Medicine I have the honor
3
and privilege of leading the biomedical data science initiative
4
I usually say I was born a geek
5
but trained as a cardiologist I did medical school in Scotland
6
you could probably tell from my accent I did some statistics during a PhD in Oxford
7
and moved to Stanford about 12 years ago, and so from when I first visited, this was clearly the place I wanted to eventually call home,
8
and the combination of bringing the sort of engineering and computer science
9
and theoretical tools
10
that we have developed in a place like this to the application of medicine is the thing that excites me the most, so looking forward to the discussion today.
11
So I'm Vijay Pandey.
12
I'm the leader of the Theory Initiative at the School of Humanities and Sciences.
13
I'm actually in a couple of different departments.
14
I'm in chemistry but also with courtesy appointments in computer science and structural biology.
15
I'm also director of the program in biophysics.
16
My research is at the intersection of a lot of different areas, but especially we use a lot of data science methods, especially for understanding fundamental biophysics and also for coming up with new small molecule therapeutics.
17
My name is Hector Garcia Molina.
18
I'm in the computer science and electrical engineering departments.
19
I was talked into being the director of the Stanford Data Science Initiative.
20
My colleagues twisted my arm, which is, it has a lot of engineering folks, but there are also other people in statistics and other departments,
21
and we have a lot of already existing collaborations with other groups on campus.
22
And I should point out that the executive director, the guy who really does everything is Steve Eglash, who's sitting right here.
23
And if I get a really hard question, I'm just going to ask Steve to answer it.
24
I'm John Hennessy.
25
I'm actually a computer, I have the same appointment Hector has, double E and CS appointment.
26
I'm a computer scientist.
27
I came here to figure out what the difference between a data scientist and a computer scientist was, which I expect to learn today.
28
But anyway, obviously I have an interest in this.
29
I think it's an important initiative and I think it has many of the interdisciplinary roots, I think, that we've tried to hone and improve here in the university.
30
It also, I think one of the other things that's attractive perspective is
31
it has deep roots on everything from theory to application
32
and I think that promises a lot of opportunity for the university where we have faculty across that entire spectrum.
33
Perfect.
34
Okay, great.
35
Thank you.
36
So, the first set of questions is more about understanding what
37
data science is just to make sure we're all on the same page.
38
So, let me ask by just asking maybe each of you to pick one application area for data science, be it, you know, healthcare, care, climate, whatever, one application area,
39
and explain kind of how data science is going to revolutionize that field, and we can take it in turn.
40
All right.
41
Okay, I'll go first.
42
So one of the things
43
that I have been focused on in my own group over the last few years is a tremendous technological advance, which is the change in the cost of sequencing human genome.
44
You may be aware
45
that it cost $3 billion to fund the Human Genome Project about a decade ago
46
and that over the course of the last 10 years
47
that cost has dropped down to $1 ,000
48
so a million fold drop in the cost of sequencing in genome and that has had an already an absolutely transformative
49
role to play in changing medicine to give you an example of
50
that we had a little baby admitted to the neonatal intensive
51
care unit at the children's hospital a few months ago who had not a mystery syndrome
52
but something we understood well long qt syndrome this was associated with cardiac arrest
53
and the little baby had multiple cardiac arrests on day one of life.
54
We know the genetic basis of that condition, but at the moment, with the standard approach, a genetic test would take three months.
55
It would look at five genes, so very small data, and it would give an answer that would help the physicians then do what we're increasingly calling precision medicine.
56
We'll probably come to that phrase later.
57
With the new technology, we can do a whole genome sequence which reduces six billion data points in 24 hours, analyze it using algorithms that used to take days
58
that now take hours and put the result back into the hands of the treating physicians in days.
59
So if I was to think of one example
60
that could encapsulate what we can do and what data science could bring to medicine, we've looked at one gene at a time, we've looked at you go in and you have a lab sample and maybe 10 reagents come back to you.
61
But now we have technology that can deliver 6 billion results back to you.
62
And there's a huge signal to noise issue there.
63
And so the sorts of things
64
that colleagues in engineering have been dealing with for many years suddenly become extremely relevant to things
65
that are happening in the medical school.
66
I think the example I would give is also in the healthcare area, but from a different perspective.
67
What I think a lot about is how we design small molecule drugs.
68
Right now, the procedure is very empirical.
69
If you think about it, when we do other things, like we design a bridge, like the bridge, the Bay Bridge, we don't do that the way we design drugs.
70
If we do it that way, we would have some idea for the bridge.
71
We'd think it works, and then we'd first send mice over the bridge to see if they survive.
72
And then if the mice survive, we would take people who really want to go to San Francisco and we'd put them over the bridge first.
73
So we would never do that, right?
74
And so it's interesting to think, why can't we engineer small molecules in an analogous way?
75
And there's many different answers to this.
76
This is an extremely difficult problem.
77
But I think the opportunity now is that we have a huge data sets, and so huge that you can't have a single person wrap their head around them.
78
And I think that the promise of data science is
79
that we'll be able to algorithmically gain insights and insights that are really non -obvious.
80
Victor?
81
Okay, well, just to be contrarian, I'm gonna mention old applications like finance, commerce, manufacturing, where they're not as trendy,
82
but they're still very important.
83
And there's a lot of great things that can happen to produce products and services cheaper, more efficiently, and safer again,
84
because we have a lot more data about the process of manufacturing or about the customers and their needs and their desires.
85
So it's a combination of more data and better ways to do things.
86
So more data and better algorithms.
87
And better algorithms.
88
Yeah, yeah.
89
And better services that we can provide.
90
Yeah, makes sense.
91
So I think one thing that is going to be possible now is the work that's going on on deep learning.
92
And really, the funny thing about deep learning is deep learning will give computers what we would call common sense.
93
It's really the ability to devise relationship.
94
And that's happening now because we've learned enough about machine learning and we have enough data.
95
You want to learn about the world?
96
Go read the internet.
97
Read the internet, put down all the relationship you get about the internet, and then build a system that can reason about them.
98
So you need both large amounts of data and large amounts of compute.
99
And we're going to spend a lot of money doing doing things
100
which people could do easily right now for smaller amounts of money, but in ten years they'll be a lot cheaper and that will be liberating.
101
For me, for my area, the application of data science that's really exciting is the application actually to computer security.
102
So as you might realize, attackers are constantly doing new things and for us it's not new data, it's actually, it's not large amounts of data and new algorithms,
103
it's actually the fact that there's new types of data that are being collected in order to detect attacks.
104
Just as examples, LinkedIn, for example, is constantly being bombarded with fake accounts.
105
So how do you detect fake accounts in the massive sea of accounts that they have?
106
That's all based on data science techniques.
107
But is this really a new field or is it just like John says?
108
Is this just more computer science and more statistics?
109
Is there something fundamentally new happening here?
110
Are there kind of new techniques that are being developed, new ideas that are being developed? or is it just more of, it's just a new name to an old concept?
111
Yeah, please, Hector.
112
Can I take a crack at that?
113
Definitely, data science is new in the same sense that Columbus discovered a new continent 500 years ago,
114
where there happened to be a lot of people living there already.
115
So, it's sort of similar.
116
There's a lot of data -related work that's been going on for the past years, But, what's really new is the discovery
117
that a group of influential people have discovered
118
that there's a lot of value and power in data and mining it and analyzing it, and that's what's exciting.
119
Now, I'm not implying that everything has been done, right?
120
Even after Columbus discovered America, there were still a lot of good things that happened in America.
121
There was South America. Did you?
122
There was South America, for example.
123
Well, I meant America in the broad sense.
124
Broad sense, America.
125
The Americas.
126
Now, Hector, I want to point out that Columbus' discovery did not go so well for the people
127
I hope you're not going to do the same thing in data science.
128
Why not?
129
Why not?
130
Yeah, it's a good question.
131
I think this is a good question because Jeff Ullman, my colleague in computer science, once said, anything that has to call itself a science ain't one.
132
And I think there's some truth to that.
133
You know, the question will be, take computer science's discipline, right?
134
Over time, yeah, surely its foundations are in lots of other disciplines, but over time there became a core body of knowledge that distinguished what computer scientists know, right?
135
And it's about algorithms and complexity and computability and autonoma, and there's a core body of knowledge.
136
My guess is the same thing will happen here.
137
There'll be a core body of knowledge.
138
It will build on statistics and mathematics and other disciplines,
139
computer science, but it will be increasingly distinct as our applications frontier moves forward over time.
140
Also, a lot of times we define these disciplines by the curriculum that we teach, Because, you know, you can be in a computer science department
141
or a biology department or a math department and do whatever research.
142
But often it's that discipline that we want to teach the younger generation, which is a key aspect of it.
143
I think the absolute scale is also larger.
144
I mean, I think there's no debating that.. compared to 10 years ago, the amount of data in the world, the amount of data that is being passed through the pipes on an hourly basis is greater.
145
And so that can, one, lead to, I think, learning that wasn't possible with smaller data.
146
But I think it can lead to opportunity in extending current disciplines.
147
So it's certainly not replacing anything, I think, as John said.
148
It really builds on those, brings them together.
149
And I think, for me, it's almost like I don't really care at some level whether it's new or not.
150
What is clearly new is the availability of data and the problem as to what we're going to do with it.
151
What are the techniques that make the field its own unique distinctive field from, you know, just machine learning or just databases or things we've had in the past?
152
Are there specific things that folks here can learn about that make them data scientists?
153
Sometimes it's not necessarily going to be something truly unique.
154
In making an analogy to other things like biophysics, biophysics is a mixture of physics and biology in other areas.
155
And it's maybe more about sort of how we do that mixture, what the cocktail is like.
156
Because, you know, different cocktails can have similar ingredients, but different quantities of each.
157
I think lots of people will want to own certain techniques and will want to have them under their wing.
158
And I think that that's okay.
159
I think that, again, for me, it's the application that counts, rather than exactly which box we put it in.
160
And so, for sure, I think the most common answer to your question is machine learning, deep learning, and applying that to extremely large data.
161
That's probably the closest to the thing that's new that's sort of associated with data science.
162
But really, isn't that just an extension of statistics?
163
I think probably, you know, many levels it is.
164
I mean, I would say there's a very broad range of tools that come from multiple toolboxes, but I would.
165
Yeah, I think that's right.
166
Yeah, we're using a lot of existing tools, but the process of applying these tools to new domains, new applications will improve the tools over time.
167
So I think Dan's point's well taken.
168
I'd say right now, saying you're a data scientist
169
or you're working on data science is a way of defining a collection of disciplines
170
and knowledge that you have from different disciplines.
171
It's not yet distinctive in that way, and it may not be, because I think you can see statistics
172
and computer science kind of embracing these techniques and developing them further as a natural evolution of their own disciplines.
173
So basically, we don't quite know yet what makes a data scientist a data scientist, but We know what we see.
174
But we intend to discover.
175
We know what we see.
176
We're going to learn.
177
That's part of the problem.
178
That's part of the challenge, I guess.
179
But it does sound like if you want to become a data scientist, you need to learn, you need to know machine learning, you need to know statistics.
180
Those are the general areas that you must know in order to call yourself a data scientist.
181
Is there agreement on that?
182
Yeah.
183
Yes?
184
Yeah.
185
And something about computation.
186
Always helpful.
187
So the question is, what is the role of universities in this new world?
188
I mean, how do we get the data that is proprietary, has privacy regulations around it, The companies are just not going to share the data with us.
189
So how are we going to do data science in academia if we don't have the core data?
190
I mean, I actually would challenge the assumption that the company is the only ones with data.
191
I think Ewan can speak very much to genomic data, which there's huge of it.
192
And actually, we have a whole hospital there full of data.
193
I think astronomy is very data rich.
194
I think not all fields are data rich, but I think there's many, many fields just within academia that are themselves data rich.
195
Yeah, I mean, I think if you take the hospital for a moment, I mean the 500 beds, many of them are telemetry beds which mean that we have real -time signals coming from the heart rate,
196
from the ECG tracing, electrical tracings of the heart.
197
We have every day on probably too many patients
198
but we have lab tests that are coming in streams
199
and for the most part that data is reviewed on a daily basis, kind of reviewed somewhat in a time series way manually on a screen at least,
200
no longer on paper and then kind of dismissed.
201
there is no algorithm sitting in the background
202
that understands like your fraud alert on your credit card you know what the normal boundaries for you are
203
and probably there should be but actually they've got resources
204
that we don't have at Google you could run a 10
205
million core job it wouldn't be a big deal that's impossible to do on campus so I
206
but I grew Dan I think our our data is a
207
drop in the bucket we don't need data on 500 patients or even a year's worth of 500 patients.
208
We need data on 10 million patients, or 20 million patients, if you really want to start working on the problem.
209
And that's got to be a full set of data, not only genomic data.
210
We need lots of medical data.
211
And I think we're going to have to find a way, and I thought the secret was going to be, I thought you were on this panel, Dan, as the one who was going to help us break into their systems and send all the data out.
212
I see.
213
Data theft is a different panel.
214
That's a whole other department.
215
The issue is related to privacy, one of the topics new to you, because I don't think it's an issue of Stanford data versus other industry data.
216
It's an issue of people are very paranoid these days for various reasons, and even the government.
217
public data doesn't want to share it.
218
So getting any type of data is hard.
219
I mean, we've had some trouble in the past, although I'm sure that that's no longer going to be true, with our hospital trying to get any data out of there was hard.
220
And there's a good reason for that.
221
There's HIPAA and all sorts of constraints.
222
One of the things we learned about the NSA Snowden thing is
223
that just collecting the metadata in fact gave them a fairly good look into some people's lives.
224
So how you anonymize it while maintaining the structure of the data in such a way
225
that you can reason from it is going to be harder, I think.
226
We have a lot of admissions data on campus.
227
Can we use that for research?
228
I think it is the intention that we have the education data and we're going to use that as a way to, and this I think is one of the big issues about moving to more online methods is
229
that you can build the analytics and get more information about what students learn and have a shorter feedback loop.
230
I think we definitely want to do that.
231
So the university would be willing to share education data that is being collected with the researchers in the data science initiative?
232
Sure, sure, sure.
233
And even admissions data at some level, obviously.
234
There are certain things you've got to keep anonymous, but at some level.
235
That's great.
236
That's good to hear.
237
It's a fascinating data set.
238
They wanted our help to study this data and tell them what to do, what they could do with this data.
239
But at the end of the discussion, I asked, well, okay, so we're very interested.
240
How do we get the data?
241
Oh, I don't think we can give you the data.
242
Anyone who's here who would like to come and work on it, please come and find me afterwards
243
and I can point you to any number of data sets that need people to work on them.
244
So I think absence of data is not our challenge, but there is still a role, I think, for synthetic data sets.
245
I could just see, when you go to your doctor, I guess you are a doctor, but you must have a doctor.
246
I can sort of tell him.
247
Here, I have this pill for you.
248
We've tested it on Tron, this simulated human whose simulation was written by an undergrad I have here at Stanford, and it works fine.
249
Tron survived.
250
And the students graduate and we don't understand how the simulation works, but we think it's okay.
251
I just, I think the synthetic, I worry about models that are developed around synthetic data
252
and the fact that you fit the algorithm to match the synthetic data.
253
I worry about that.
254
I also worry about, we've kind of moved away from this name big data.
255
And I think one of the reasons we did is there's lots of big data out there, but lots of it isn't particularly complicated.
256
It's simple data, it's not complex relationships.
257
And when you think about synthetic data, how are you going to get synthetic data that has deep, complex relationships that are not artificial but, in fact, look like the real thing?
258
I think, ultimately, we're all talking about the same thing.
259
We're not going to simply put data sets together and then sit back and say, well, we feel really good now.
260
We've got a great data set.
261
The point of putting the data set together is to ask a question.
262
So I suspect we don't disagree as much as maybe we sounded like we might have done.
263
Yeah, I think essentially we led off with some examples of questions.
264
But I think questions are very domain -specific.
265
and the part that is common is maybe more methods.
266
I could give another quick example just from the idea that it builds upon federated queries but moves towards federated statistics because, again, as we mentioned earlier,
267
probably one hospital, I'm going to keep away from the domain, but one unit, it may not have all the data in one place
268
that you want to analyze and there may be privacy
269
and we'll get to this privacy or regulatory reasons that you can't actually bring that data together in one place.
270
There's a lot of really challenging problems and just getting data that's usable.
271
It's not just you get some file from somebody and you're all set to go.
272
You have to organize the data.
273
You have to understand what does it mean when you have a field there, how an instrument was used to measure this, and how do I combine this reading in this field with another reading that's related
274
but was done with a different instrument or measured in a different way, for instance.
275
So there's a lot of challenges.
276
How do I represent the accuracy in the data itself?
277
And generally, how do we adapt the curriculum to deal with teaching data science and data science skills.
278
I mean, I have one idea towards this.
279
One of the things
280
that I have always been really impressed about about the new CS curriculum is
281
that CS can be done in a very project -oriented way.
282
If you think about like 106A, students can see the value of computer science very quickly because they're actually doing it.
283
And you can imagine having data science -like projects in other disciplines where the project has to do with analyzing the data.
284
And maybe they don't take the data themselves, and that would be very onerous and almost may be impossible in some cases, but I can imagine that there could be applications like that.
285
And I think my personal prediction is that education will start to look more
286
and more project -based as I think it has a greater impact on the students
287
and also becomes just more exciting for them.
288
Yeah, we've had successes here with some classes where at the beginning of the class the instructor has put together 20, 30 data sets and then the students can pick one that interests them.
289
Often they can even go and talk to the person who generated
290
that data set and then they can do an interesting project with it.
291
So I think that same idea could be used in many disciplines.
292
So you see this as, I mean, there are many faculty here are teaching classes
293
that could adapt to that model where there would be more data sets available
294
and the courses would actually focus on how to analyze those data sets?
295
Possibly.
296
I mean, I'm not sure, since I'm not an expert, I don't know how far we can go to data mine Shakespeare or something like that, but it's...
297
It can be done.
298
There's actually a professor here in the English department who's been analyzing old manuscripts.
299
I don't know, I think Shakespeare.
300
Yeah, right.
301
Well, I think then in the medical school, I think we have a slightly different challenge in that we're generally training people not to be data scientists, we're training them to be doctors,
302
but yet data science will be such an important part of medicine going forward
303
that we need them to be much more quantitatively aware than they currently are.
304
And so there's been a number of movements towards thinking about how to improve
305
or increase the amount of exposure to those topics
306
that our medical students have because they're going to need to understand the theory behind it
307
if not necessarily have to be able to actually apply the command line and all the things that they might...
308
And understand what they can trust and what they can't.
309
If you think about core data scientists, as opposed to...
310
We're going to have lots of students who learn about data science techniques in the context of whatever they're doing, biology, chemistry, engineering.
311
But I think we also probably want to think about a core educational program
312
that trains people who are deeper in the fundamental techniques
313
and applying those techniques and maybe not so focused on the applications in any single area.
314
And that's a good model for how that might be done. So, here I have to remind everybody of this example of a father who went
315
and complained to Target about his teenage daughter getting bombarded with ads for cribs and for diapers,
316
only to find out that Target figured out that his teenage daughter was pregnant before she told him.
317
So, yeah, that was quite interesting.
318
True story.
319
True story.
320
True story.
321
Absolutely true story.
322
So, the question is, how did Target actually do this?
323
And it turns out what they did is actually, well, they had data scientists who were actively working on their data,
324
where they had mothers sign up for their baby registry, and then they started analyzing what do pregnant mothers,
325
soon -to -be mothers, what do they buy?
326
and it turns out like in their second trimester they tend to buy unscented lotion
327
and in their first 20 weeks they load up on supplements like calcium
328
and magnesium and quickly they correlated that to you know the fact
329
that someone is expecting just based on their purchase patterns
330
so that's an example of you know kind of crossing the border of ethics and data
331
and I guess that's what you're getting at is it is
332
there a limit should there be limits to what data is being collected
333
and generally what do we do about the ethical question.
334
And should we teach them the difference between correlation and causation?
335
There you go.
336
There you go.
337
Very important.
338
Very important.
339
If you have the data, you can study it.
340
And if you discover patterns like this, as long as it's not referring to a specific teenage girl, I don't see an issue.
341
It's the specifics of when it leaks that this specific girl is pregnant, that's when you have problems.
342
So then, Hector, let's put it in.
343
So now Now you discover this connection,
344
you publish it, and the next thing that happens is Target then deploys that algorithm using your discovery,
345
or somebody else deploys it, and they start then bombarding the advertisements.
346
Now they've used your insight to create a problem.
347
Yeah, but that's going to happen.
348
Whatever algorithm I develop, it can be used by bad guys to do bad things, so I shouldn't be doing my research?
349
This is a common issue in technology, from atom bombs to anything else.
350
I mean, I think the challenge especially is that if Hector didn't do it, maybe 10 or 20 years later, someone else would do it.
351
Of course, somebody else would do it.
352
In the digital world, increasingly, you have to do something really contrived to,
353
if you really want ultimate privacy.
354
It's the honest.
355
I mean, Dan's right.
356
There's a scale issue here and a pervasiveness issue that's very different.
357
I mean, well, look at credit card data.
358
Everybody, you know, does everybody use credit cards?
359
Yes.
360
And imagine that data being crunched not just by the credit card form, but then sold, because, of course,
361
it is sold, and used in lots of different ways to determine buying techniques and things, people, I mean, it's, and this is true in all the stores now.
362
You know, the new way to try a new product out is you pick a few stores, and you go watch who buys it, and you watch information about who buys it, and they use a credit card.
363
Now you deduce from that what kind of people might buy
364
that product and you decide what stores should have it and how much space it should get.
365
That's why when I go shopping, I wear a brown paper bag on my head.
366
You do.
367
And you use cash?
368
I use cash.
369
He uses Bitcoin.
370
Some kind of regulation is going to be necessary.
371
I think trying to shut down the research side of the agenda doesn't work.
372
I mean, it basically doesn't work.
373
So you're going to have to figure out what companies' regulations are about security and privacy.
374
And the problem with that is that that involves government.
375
And government moves about 10 times slower than technology moves.
376
So it's gonna be very hard to develop that.
377
And you just have to look at what's happened in copyright law and how far out of date it is.
378
Not, you know, 50 years out of date, probably something like that, right?
379
And so we're gonna have that problem as a society.
380
This is a societal issue, it's gonna have to get.
381
Okay, I guess we have to leave some topics for next time.
382
I did wanna talk about the role of libraries and can libraries help with data collection and data retention?
383
But I guess those are things we'll have to talk about next time.
384
I think they'll have to is the answer.
385
They'll have to.
386
They'll have to adapt, including the Stanford Library.
387
Yes?
388
It would be really nice if the Stanford Library...
389
We don't have the most old -fashioned library in the world.
390
Exactly.
391
We're ahead of that curve in terms of some.
392
Oh, yeah.
393
Actually, I did want to advertise.
394
There's something called...
395
Maybe you guys want to talk about the Stanford Digital Repository.
396
I think they don't think libraries about books.
397
Libraries about data.
398
Okay.
399
Well, thank you, everybody.
400
Thank you.
401
Thank you.
402
For more, please visit us.
הורד אפליקציה
כל מה שאתה צריך כדי לדבר בשטף
הגיית AIציון לכל משפט
תרגול IPAשלוט בכל צליל
אוצר מיליםבנה את מאגר המילים שלך
משחק אוצר מיליםלמד תוך כדי משחק
אותו ערוץ
C12
Stanford's Sapolsky On Depression in U.S. (Full Lecture)
Daily ConversationUniversity Students
C18
Martin Andrews – Definitive Stem Cell and Gene Therapy for Child Health: Stanford Childx Conference
Business EnglishUniversity StudentsWorking Adults
C275
Steve Jobs' 2005 Stanford Commencement Address
University Students
C17
2019 Stanford Commencement address by Tim Cook
Business EnglishUniversity StudentsWorking Adults
C219
2016 Stanford Commencement address by Ken Burns
Daily ConversationUniversity StudentsWorking Adults
C16
Steve Jobs' 2005 Stanford Commencement Address (with intro by President John Hennessy)
Business EnglishUniversity StudentsWorking Adults
B23
Sundar Pichai gives Stanford’s 2026 Commencement address
Business EnglishInterview EnglishUniversity Students
FOR YOU
המלץ על סרטון
C12
478 משפטים
Stanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Economics of Generative AI
Daily ConversationUniversity Students
C14
23 משפטים
3-D Brain Scans | Pathways 2
Daily ConversationUniversity Students
B21
220 משפטים
How Modern Data Teams Work (Engineer, Analyst, Scientist, Architect, ML)
Daily ConversationUniversity Students
C11
398 משפטים
Sec Growth DataScience staff meeting Sep 14 2022
Daily ConversationUniversity StudentsWorking Adults
C12
30 משפטים
Would a society based on genetics be… good? | Kathryn Paige Harden
Daily ConversationUniversity Students
C11
708 משפטים
實戰練英文 | NVIDIA 黃仁勳榮獲2025霍金獎，分享AI革命與創業心路；AI將改變所有的工作，讓你重回工作重點；想成功，一定要告訴自已這句話 | 劍橋大學演講 | 長文、註解、大字版、中英對照
Daily ConversationUniversity Students
C11
346 משפטים
Trust the Process: How to Design Teams That Actually Work
Daily ConversationUniversity StudentsWorking Adults
C11
244 משפטים
1. Introduction to statistics - Lecture 1
Daily ConversationUniversity Students
C11
719 משפטים
View From The Top with Roelof Botha, Managing Partner of Sequoia Capital
Daily ConversationUniversity StudentsWorking Adults
C11
332 משפטים
Science Writing: Practice Makes It (Almost) Perfect
Daily ConversationUniversity Students
C11
206 משפטים
Anxiety, Stress, Loneliness & Depression Cured In 9 Weeks? | Dr. Sahar Yousef | Raj Shamani Clips
Daily ConversationUniversity Students
C11
108 משפטים
LTEC 782: Design-Based Research in Education (Session 01)
Daily ConversationUniversity Students
אוצר מילים והערות הגייה לשיעור הזה
שיעור הדיבור הזה ברמה C1 מבוסס על הסרטון “The Future of Data Science - Data Science @ Stanford”. בסרטון הזה יש 402 משפטים ו-5232 מילים לתרגול Shadowing. הדיבור נמשך 25:43. הדובר מדבר מהר, כ-203 מילים בדקה, אז צפו לצלילים מחוברים ומקוצרים. 86% מהמילים נמצאות בין 3,000 המילים הנפוצות ביותר באנגלית; כדאי ללמוד את השאר לפני שמתחילים.
אוצר מילים מרכזי בסרטון הזה
15 המילים המתקדמות ביותר בסרטון, עם ההגייה והמשמעות שלהן:
| מילה | הגייה | משמעות |
|---|---|---|
| algorithm שם עצם | /ˈælɡəɹɪðm̩/ | אַלגוֹרִיתְם |
| biophysics שם עצם | ביופיזיקה | |
| molecule שם עצם | /ˈmɑ.lɪ.kjul/ | מוֹלֵקוּלָה |
| colleague שם עצם | /ˈkɒliːɡ/ | עָמִית, עמיתה |
| simulation שם עצם | /ˌsɪm.jəˈleɪ.ʃn̩/ | הדמיה, סימולציה |
| detect פועל | /dɪˈtɛkt/ | גילה, זיהה |
| ingredient שם עצם | /ɪnˈɡɹiːdi.ənt/ | מרכיב, מצרך |
| fundamentally תואר הפועל | עֶקְרוֹנִית | |
| calcium שם עצם | /ˈkælsi.əm/ | סידן, קלציום |
| advertisement שם עצם | /ˈædvɚˌtaɪzmənt/ | פִּרְסוֹמֶת |
| dismiss פועל | /dɪsˈmɪs/ | לפטר |
| distinguish פועל | /dɪˈstɪŋ.ɡwɪʃ/ | הבחין |
| intersection שם עצם | /ˈɪntəɹˌsɛkʃən/ | צֹמֶת |
| cocktail שם עצם | /ˈkɑk.teɪl/ | קוֹקְטֵיל \ קוקטייל |
| prediction שם עצם | /pɹɪˈdɪkʃən/ | תַּחֲזִית |
Phrasal verbs שתשמעו
| מילה | משמעות |
|---|---|
| figure out פועל | הֵבִין, מצא תשובה |
משפטים שכדאי לחזור עליהם
משפטים קצרים ושלמים מהסרטון שאפשר להשתמש בהם בשיחה יומיומית:
- You want to learn about the world?
- Can I take a crack at that?
- Can we use that for research?
- How do we get the data?
- Everybody, you know, does everybody use credit cards?
דקדוק בסרטון הזה
המבנים שהדובר משתמש בהם הכי הרבה, עם המילים המדויקות מהסרטון:
| מבנה | בסרטון |
|---|---|
| הווה מושלם ממושך (Present perfect continuous) have/has been + -ing — פעולה שהתחילה בעבר ועדיין נמשכת עכשיו | have been dealing · that's been going · who's been analyzing |
| הווה מושלם (Present perfect) have/has + past participle — פעולה שקרתה בעבר ועדיין יש לה חשיבות או השפעה עכשיו | have developed · we've tried · have been |
| סביל (Passive voice) be + past participle — הדגש הוא על מה שקורה, לא על מי שעושה את הפעולה | was born · was talked · was associated |
| משפטי זיקה (Relative clauses) who / which + clause — מידע נוסף על אדם או דבר | arm, which is · guy who really · advance, which is |
הגייה ששווה לשים לב אליה
הדובר משתמש ב-108 קיצורים וצורות מקוצרות, כמו I'm, don't, we're. אמרו אותם בצורה הקצרה, כפי שאתם שומעים אותם.
- הצלילים של “th”: algorithm /ˈælɡəɹɪðm̩/, synthetic /sɪnˈθɛtɪk/, theoretical /ˌθi.əˈɹɛt.ɪ.kəl/, ethical /ˈɛθɪkəl/
- הצלילים של “sh” ו-“zh”: simulation /ˌsɪm.jəˈleɪ.ʃn̩/, retention /ɹɪˈtɛnʃən/, correlation /ˌkɔɹ.əˈleɪ.ʃən/, distinguish /dɪˈstɪŋ.ɡwɪʃ/, intersection /ˈɪntəɹˌsɛkʃən/
- מילים ארוכות — שימו לב להטעמה: cardiologist /ˌkɑɹdiˈɑlədʒɪst/, academia /ˌæk.əˈdi.mi.ə/, ingredient /ɪnˈɡɹiːdi.ənt/, analytics /ˌæn.əˈlɪt.ɪks/, advertisement /ˈædvɚˌtaɪzmənt/
צלילים שדוברי עברית מתקשים בהם:
- /w/ — שפתיים מעוגלות, לא /v/: distinguish /dɪˈstɪŋ.ɡwɪʃ/, quantity /ˈkwɑn(t)ɪti/
- /æ/ — פתוח יותר מה-“אֶ” העברית, לא כמו “a” ב-“cup”: algorithm /ˈælɡəɹɪðm̩/, analyze /ˈæn.əˌlaɪz/, adapt /əˈdæpt/, academia /ˌæk.əˈdi.mi.ə/, cardiac /ˈkɑɹdiˌæk/
- /r/ — הלשון לא נוגעת בשום מקום, לא ה-“ר” הגרונית: algorithm /ˈælɡəɹɪðm̩/, bombard /ˌbɑmˈbɑɹd/, cardiologist /ˌkɑɹdiˈɑlədʒɪst/, federate /ˈfɛdəɹeɪt/, cardiac /ˈkɑɹdiˌæk/
איך לתרגל עם הסרטון הזה
- הקשיבו לכל הסרטון פעם אחת בלי לדבר, ורשמו את המילים שאתם לא מכירים.
- התחילו במהירות 0.75×, תרגלו Shadowing משפט אחר משפט, וחזרו למהירות רגילה כשזה מרגיש קל.
- הקליטו את עצמכם והשוו למקור, תוך תשומת לב למילים כמו algorithm, biophysics, molecule.
מהי טכניקת השאדואינג?
שאדואינג היא טכניקת למידת שפה מבוססת מדע שפותחה במקור להכשרת מתורגמנים מקצועיים ופופולרית על ידי הפוליגלוט ד"ר אלכסנדר ארגואלס. השיטה פשוטה אך עוצמתית: אתה מקשיב לאודיו באנגלית ילידית וחוזר עליו בקול רם מיד — כמו צל העוקב אחרי הדובר עם עיכוב של 1–2 שניות בלבד. בניגוד להאזנה פסיבית או תרגילי דקדוק, שאדואינג מכריח את המוח והשרירים בפה לעבד ולשחזר דפוסי דיבור אמיתיים בו זמנית. מחקרים מראים שזה משפר משמעותית את דיוק ההגייה, האינטונציה, הקצב, הדיבור המחובר, הבנת השמיעה ושטף הדיבור — מה שהופך אותו לאחת השיטות היעילות ביותר להכנה לIELTS Speaking ולתקשורת באנגלית בעולם האמיתי.