WEBVTT

1
00:00:00.480 --> 00:00:03.270
Very happy to be here
in Boston this morning,

2
00:00:03.780 --> 00:00:08.130
and today is a special day in Boston
because I'm with you, Arun. Arun,

3
00:00:08.130 --> 00:00:12.900
you are our chief data scientist
and it's today your ninth

4
00:00:12.960 --> 00:00:17.760
anniversary at BCG X. So time to
step back and, maybe to start with,

5
00:00:18.060 --> 00:00:22.170
sharing a bit how you started. You
decided to do a PhD in academia.

6
00:00:22.470 --> 00:00:24.450
Why on earth did you decide that?

7
00:00:25.380 --> 00:00:26.460
Not just a PhD,

8
00:00:26.460 --> 00:00:30.390
it was also a PhD because as an engineer
I was building probability theory-based

9
00:00:30.390 --> 00:00:31.290
machine learning models,

10
00:00:31.500 --> 00:00:35.820
and I got more interested in more
differential equation-based AI machine

11
00:00:35.820 --> 00:00:39.840
learning models. So my PhD is more
in the math behind deep learning,

12
00:00:40.140 --> 00:00:43.470
and I was driven purely by curiosity.
I was like, oh, I knew this family.

13
00:00:43.650 --> 00:00:48.060
I wanted to learn more about
another family. And ever
since I got out of my PhD,

14
00:00:48.420 --> 00:00:51.690
the whole deep learning wave took off
and the reinforcement learning wave took

15
00:00:51.690 --> 00:00:54.960
off. And part of my PhD, I
learned the math behind all that.

16
00:00:55.230 --> 00:00:57.900
So that's the reason I enjoy what I do,

17
00:00:57.900 --> 00:01:00.120
because I understand the
math and being at BCG,

18
00:01:00.120 --> 00:01:02.010
it's also the doted line to business.

19
00:01:02.370 --> 00:01:06.840
And we see tons of innovation today-faster
and faster pace of innovation.

20
00:01:07.260 --> 00:01:12.120
How do you see the evolution of the
intersect between business on one side and

21
00:01:12.150 --> 00:01:13.740
science on the other side: the whole?

22
00:01:14.400 --> 00:01:18.120
In the last decade, Silvain,
that divide has gone.

23
00:01:18.120 --> 00:01:21.150
So it's diluted so much,

24
00:01:21.630 --> 00:01:23.580
because when I started my career,

25
00:01:24.660 --> 00:01:27.690
we didn't have academic faculty
working so much in industry.

26
00:01:27.900 --> 00:01:32.400
And now there's so much cross play
and with BCG as part of the scientific

27
00:01:32.400 --> 00:01:34.980
network that I manage, we have
faculty working with us closely,

28
00:01:34.980 --> 00:01:38.250
that is science working closely
with business. In a world where

29
00:01:39.930 --> 00:01:43.170
we want to have a cutting edge
solution, we have quick access.

30
00:01:43.170 --> 00:01:46.200
That access 10 years ago,
15 years ago did not exist.

31
00:01:46.440 --> 00:01:51.180
So that divide is going to blur out
more and more in the coming years.

32
00:01:51.420 --> 00:01:56.370
And what's the scientific domain
that you think will impact us most in

33
00:01:56.370 --> 00:01:57.660
the next two, three years?

34
00:01:57.810 --> 00:02:02.580
Emerging trend for me is like something
which is more in the manufacturing

35
00:02:02.580 --> 00:02:06.180
side. Okay, let's take an example,
right? A large manufacturing plant,

36
00:02:06.210 --> 00:02:09.510
you have like a $200 million machinery
that's doing. And guess what?

37
00:02:09.510 --> 00:02:13.680
All those parts are defined by classic
human intelligence historically.

38
00:02:14.040 --> 00:02:16.860
But now what's happening is
in order to remove waste,

39
00:02:16.920 --> 00:02:21.060
in order to get more efficiency out
of this $200, $300 million machines,

40
00:02:21.360 --> 00:02:26.160
people want to use generative models
to identify the best shape for the

41
00:02:26.160 --> 00:02:29.970
missionary parts. So this
is territory that people,

42
00:02:30.030 --> 00:02:33.960
we did not think about it because two,
three years ago, again, pre-GenAI,

43
00:02:33.960 --> 00:02:37.800
we didn't have large language models,
but this is actually foundation model,

44
00:02:38.070 --> 00:02:41.790
generative model, but not speaking English
words that you and I can understand,

45
00:02:41.790 --> 00:02:46.020
but more about creating shapes which
are going to have specific impact in the

46
00:02:46.020 --> 00:02:49.830
manufacturing process. So that's
where the next two, three years,

47
00:02:50.130 --> 00:02:52.680
I would predict a big push in that space.

48
00:02:52.860 --> 00:02:54.750
Thank you, Arun. This
was super interesting.

49
00:02:55.140 --> 00:02:55.680
Thanks Silvain.