WEBVTT

1
00:00:00.240 --> 00:00:03.540
Happy to be in sunny Boston this
morning with you, Youssef. Youssef,

2
00:00:03.630 --> 00:00:07.080
you're a data scientist
at BCG X. Prior to BCG X,

3
00:00:07.440 --> 00:00:10.890
you did a PhD in
econometrics and statistics.

4
00:00:10.890 --> 00:00:15.510
And I'll start with an easy question.
Why on earth did you pursue a PhD?

5
00:00:15.990 --> 00:00:20.730
I've been fascinated with how
people make choices and how we can

6
00:00:21.330 --> 00:00:25.800
model the full complexity of how people
make choices and be able to predict

7
00:00:25.800 --> 00:00:26.480
that.

8
00:00:26.480 --> 00:00:31.380
The PhD was really to
focus specifically on

9
00:00:31.410 --> 00:00:33.450
behavioral modeling.
So exactly that topic.

10
00:00:34.410 --> 00:00:39.150
And how do we capture the
full complexity of that in a

11
00:00:39.150 --> 00:00:41.790
model and be able to
reliably make predictions?

12
00:00:42.780 --> 00:00:47.190
So capturing that complexity
for the time of your PhD,

13
00:00:47.640 --> 00:00:51.300
anything striking from that experience?
You could share with us a concept,

14
00:00:51.300 --> 00:00:53.460
an experience, something.

15
00:00:53.910 --> 00:00:58.350
A lot of the choices
that people make are not

16
00:00:58.560 --> 00:01:03.270
based on rationality. It may
or may not be a surprise,

17
00:01:04.350 --> 00:01:09.210
but one key takeaway that
I've noticed was that a lot

18
00:01:09.210 --> 00:01:14.160
of models make more
assumptions than necessary,

19
00:01:14.850 --> 00:01:18.900
and I think it's better
to embrace the uncertainty

20
00:01:19.590 --> 00:01:22.500
and for the model to simply say,

21
00:01:22.500 --> 00:01:27.480
"I don't know," or have a big
kind of confidence interval than

22
00:01:28.110 --> 00:01:29.100
be precisely wrong.

23
00:01:29.430 --> 00:01:33.570
So Youssef, I know you have
one passion around education.

24
00:01:34.110 --> 00:01:37.590
You've been doing lots
of Tutoring here at PCG,

25
00:01:37.590 --> 00:01:42.150
but also when you were a
student, and your mission,

26
00:01:42.210 --> 00:01:46.710
I understand, is to make the most
out of AI or GenAI for education.

27
00:01:46.770 --> 00:01:48.900
How can AI help there?

28
00:01:49.290 --> 00:01:53.850
It has a potential of making
available, to everyone,

29
00:01:53.850 --> 00:01:55.200
their own personal tutor.

30
00:01:55.530 --> 00:02:00.240
The tutor will be able to adapt
to everyone's knowledge gaps.

31
00:02:00.660 --> 00:02:02.730
And I mean, essentially,

32
00:02:02.760 --> 00:02:07.680
the one-on-one training with feedback
has the potential of making education a

33
00:02:07.680 --> 00:02:11.070
lot more accessible than in
the regular classroom setting?

34
00:02:11.280 --> 00:02:12.270
Fascinating,

35
00:02:12.270 --> 00:02:15.720
because we know that it's a huge
barrier in different parts of the world,

36
00:02:15.720 --> 00:02:20.250
the lack of resources to be
able to fund proper education,

37
00:02:20.430 --> 00:02:24.840
and I agree that this is one of the most
beautiful outcomes we can expect from

38
00:02:25.020 --> 00:02:27.570
GenAI. Thank you for sharing, Youssef.
Thank you.