1
00:00:00,240 --> 00:00:02,360
Hi, and welcome to the Neil
Ashton Podcast.

2
00:00:03,000 --> 00:00:05,880
In each episode, we explained
some of the fascinating ways

3
00:00:05,880 --> 00:00:09,080
that science and engineering are
changing the world around us.

4
00:00:09,760 --> 00:00:12,720
We talked to leading engineers
from elite level sports like

5
00:00:12,800 --> 00:00:16,840
cycling and Formula One to some
of the world's top academics to

6
00:00:16,840 --> 00:00:20,960
understand how fluid dynamics,
machine learning, supercomputing

7
00:00:21,360 --> 00:00:22,960
are bringing in a new era of
discovery.

8
00:00:23,920 --> 00:00:27,040
We also hear some of their life
stories, their career advice,

9
00:00:27,640 --> 00:00:30,040
the lessons they've learned on
the way that I hope will be

10
00:00:30,040 --> 00:00:33,800
helpful to you too.
So sit back and enjoy this

11
00:00:33,800 --> 00:00:40,840
episode.
Hi, and welcome back to the New

12
00:00:40,840 --> 00:00:43,160
Russian Podcast.
So today I'm delighted to be

13
00:00:43,160 --> 00:00:45,240
joined by Professor Paola
Chinella.

14
00:00:45,440 --> 00:00:48,240
She's a professor of fluid
mechanics at Sorbonne University

15
00:00:48,240 --> 00:00:52,680
in Paris, and she's also the new
director of the Sorbonne Cluster

16
00:00:52,680 --> 00:00:55,680
for Artificial Intelligence,
also in Paris.

17
00:00:56,280 --> 00:00:59,520
And what I've found particularly
interesting about her career is

18
00:00:59,520 --> 00:01:02,360
that she's not come through AI,
through the usual, you know,

19
00:01:02,360 --> 00:01:05,000
computer science route.
She is very much panel

20
00:01:05,000 --> 00:01:07,760
engineering and fluid dynamics
specialist.

21
00:01:08,080 --> 00:01:12,440
We talk about where she started
in Italy, the transition to

22
00:01:12,440 --> 00:01:17,080
France, moving back, the idea
of, you know, academia having to

23
00:01:17,080 --> 00:01:21,840
move between different countries
to, to secure the tenure

24
00:01:21,840 --> 00:01:25,200
positions, how how challenging
it, it, it can be, especially

25
00:01:25,200 --> 00:01:27,360
for, you know, for family and,
and personal life.

26
00:01:28,200 --> 00:01:30,560
But she really, well, I liked
it.

27
00:01:30,560 --> 00:01:34,000
Her passion for, for fluids, you
know, she talked about working

28
00:01:34,000 --> 00:01:38,520
on these exotic dense gases and
real gas effects and, and how

29
00:01:39,080 --> 00:01:43,160
some of those LED her towards
the uncertainty quantification

30
00:01:43,160 --> 00:01:46,880
and Bayesian methods.
And working with statisticians

31
00:01:47,040 --> 00:01:51,160
helped her then realize that she
could do the same around machine

32
00:01:51,160 --> 00:01:54,760
learning and really wanting to
collaborate with different

33
00:01:54,760 --> 00:01:56,480
disciplines and, and different
groups.

34
00:01:56,880 --> 00:02:00,040
And, and in some ways, she's
represents a very important

35
00:02:00,040 --> 00:02:03,720
bridge between what we might
call, you know, classical CFD,

36
00:02:03,720 --> 00:02:06,680
you know, numerical methods,
turbulence modelling and this

37
00:02:06,680 --> 00:02:09,560
new world of, of, of AI for
science.

38
00:02:09,560 --> 00:02:13,720
And, and she's played actually
a, a major role in the wider CFD

39
00:02:13,720 --> 00:02:15,200
community.
You know, she's an editorial and

40
00:02:15,200 --> 00:02:18,520
chief of computers and fluids
associate editor for

41
00:02:18,520 --> 00:02:20,200
International Journal heat and
fluid flow.

42
00:02:20,600 --> 00:02:23,840
And she also helps to coordinate
the Akof Tak special interest

43
00:02:23,840 --> 00:02:27,640
group on machine learning and
fluids fluid dynamics, which has

44
00:02:28,040 --> 00:02:30,640
led to a really successful
conference series.

45
00:02:30,640 --> 00:02:34,800
This ML fluids that I have been
fortunate to to have helped her

46
00:02:35,240 --> 00:02:40,440
a little bit as well.
And, but really we, she is an

47
00:02:40,440 --> 00:02:43,400
educator and she is an academic
and, and we had a really

48
00:02:43,400 --> 00:02:48,440
interesting discussion as well
on what it means in this new era

49
00:02:48,440 --> 00:02:51,880
of AI, what, what's the valuable
thing that we should be teaching

50
00:02:52,080 --> 00:02:54,720
new students at undergraduate
and postgraduate level?

51
00:02:54,880 --> 00:02:57,960
And really what is the potential
for methods?

52
00:02:58,040 --> 00:03:01,160
And one of the things we talk
about is, you know, the

53
00:03:01,160 --> 00:03:03,880
differences between surrogate
modelling and turbans modelling

54
00:03:03,880 --> 00:03:05,520
and how could they be
interconnected?

55
00:03:05,520 --> 00:03:09,920
When is the value of those?
And and we finish off talking a

56
00:03:09,920 --> 00:03:12,480
little bit on the AI for science
and how there is this

57
00:03:12,480 --> 00:03:15,520
opportunity for cross
collaboration across discipline.

58
00:03:15,520 --> 00:03:20,000
So she is somebody that I am
inspired by and always enjoy

59
00:03:20,000 --> 00:03:23,040
working with.
And I hope you enjoy learning

60
00:03:23,040 --> 00:03:25,640
more about her and her work in
this conversation.

61
00:03:25,640 --> 00:03:28,640
So sit back and enjoy this
conversation with Paula.

62
00:03:28,800 --> 00:03:31,440
So yeah, thanks.
Thanks for coming on this,

63
00:03:31,560 --> 00:03:34,080
really appreciate it.
I've loved working with you on

64
00:03:34,080 --> 00:03:36,720
various projects over the years,
but this is a good opportunity

65
00:03:36,720 --> 00:03:40,320
to hear more about, you know,
you and, and what you do.

66
00:03:41,080 --> 00:03:44,640
And I guess maybe it's a
starting question.

67
00:03:45,520 --> 00:03:48,160
You know, you started in
mechanical engineering in, in

68
00:03:48,160 --> 00:03:51,160
classical fluid mechanics, I
guess, and then you move through

69
00:03:51,160 --> 00:03:53,760
now into, you know, machine
learning and, and AI.

70
00:03:54,840 --> 00:03:57,120
But how did all that happen?
You know, rewind where?

71
00:03:57,120 --> 00:03:59,600
Where did this love for fluid
mechanics start?

72
00:03:59,960 --> 00:04:04,080
Well, when I started, so I had
to choose the university.

73
00:04:04,080 --> 00:04:06,520
I wanted to do fundamental
physics.

74
00:04:06,640 --> 00:04:09,480
I, I wanted to go to, to
fundamental physics,

75
00:04:09,480 --> 00:04:12,440
astrophysics or something or to
mathematics.

76
00:04:12,960 --> 00:04:18,360
And well, my mother said, no,
this is not a good job because

77
00:04:18,360 --> 00:04:21,279
the only opportunity for you is
to become a researcher that's

78
00:04:21,279 --> 00:04:23,720
not well paid.
So be an engineer.

79
00:04:23,920 --> 00:04:27,320
So good for engineering.
And then I started to do serial

80
00:04:27,440 --> 00:04:31,720
engineering actually.
And after one year, I discovered

81
00:04:31,720 --> 00:04:35,760
that they didn't like, and I
started to look for things which

82
00:04:35,760 --> 00:04:37,960
were closer to physics and to
mathematics.

83
00:04:38,400 --> 00:04:44,080
And well, a friend of mine said,
you know, I'm doing free

84
00:04:44,080 --> 00:04:45,680
mechanics.
There's quite a lot of

85
00:04:45,680 --> 00:04:47,720
mathematics in that, and also
some physics.

86
00:04:47,720 --> 00:04:51,600
So, right.
And that's how I went to free

87
00:04:51,600 --> 00:04:55,960
mechanics and I didn't like it.
Actually, my my professor in

88
00:04:55,960 --> 00:05:00,360
Italy is Michela Napolitano.
He was, he worked with NASA.

89
00:05:00,360 --> 00:05:04,680
He was, he had a PhD with Sally
Rubin, who was actually the one

90
00:05:04,680 --> 00:05:07,320
of the first editors in chief of
computers and suites.

91
00:05:08,000 --> 00:05:11,280
And he gave us a book by
Shapiro.

92
00:05:12,000 --> 00:05:14,720
It's fast profiles.
I think it's a book.

93
00:05:14,720 --> 00:05:16,680
You know, Shapiro is a, is a
professor.

94
00:05:16,680 --> 00:05:21,000
He was professor at MIT I think.
And he said he he gave us the

95
00:05:21,000 --> 00:05:24,960
book the first day of the
classroom and he said if you

96
00:05:25,200 --> 00:05:28,840
don't like this book, you can
change, go to another take

97
00:05:28,840 --> 00:05:31,640
another course to just abandoned
the fluid mechanics option.

98
00:05:32,040 --> 00:05:35,840
And I read it and I loved it so.
And that's how I meant for the

99
00:05:35,840 --> 00:05:40,400
mechanics.
So that's interesting.

100
00:05:40,400 --> 00:05:45,640
And then did you always want to
go and do a PhD or were you sort

101
00:05:45,640 --> 00:05:48,240
of debating of going into
industry or something?

102
00:05:48,560 --> 00:05:51,240
Yes, because you know my family,
everybody.

103
00:05:51,680 --> 00:05:53,760
So my family is a family of
professors.

104
00:05:53,760 --> 00:05:57,480
Well, they are professors more
in the secondary school, but

105
00:05:57,480 --> 00:06:02,120
they all did high studies, let's
say university and so on.

106
00:06:02,680 --> 00:06:05,720
So I wanted to be a professor
and I wanted to be a researcher

107
00:06:05,720 --> 00:06:08,440
most first of all, in the 1st
place, I wanted to be a

108
00:06:08,440 --> 00:06:14,200
researcher And, and So what I
knew that was necessary to have

109
00:06:14,200 --> 00:06:19,320
a PhD for being a researcher.
And so well, since the beginning

110
00:06:19,320 --> 00:06:23,360
I was looking for a PhD.
And at that time it was very

111
00:06:23,360 --> 00:06:27,280
hard to have a PhD in Italy
because there was no traditional

112
00:06:27,280 --> 00:06:33,960
PhD, I would say 30 years ago.
And the people went abroad for

113
00:06:33,960 --> 00:06:37,840
the PhD actually, because in
Italy it was not very valued.

114
00:06:38,320 --> 00:06:42,200
It was at the very beginning.
I was maybe the 11th cycle of

115
00:06:42,200 --> 00:06:46,680
PhD, which means that the, the
PhD degree in Italy was created

116
00:06:46,680 --> 00:06:50,120
11 years before I start, you
know, so it was relatively

117
00:06:50,320 --> 00:06:57,480
recent and, and there were very
few fellowships for a PhD, so it

118
00:06:57,480 --> 00:07:02,120
was very difficult to have them.
And that's how at some point I,

119
00:07:02,200 --> 00:07:05,160
I decided to move to France.
So, yeah.

120
00:07:05,480 --> 00:07:08,360
And and then, well, I loved
France and remained here.

121
00:07:09,800 --> 00:07:14,400
Yeah, OK.
And your topic was on more

122
00:07:14,400 --> 00:07:18,400
numerical schemes, more like
compressible to how did you get

123
00:07:18,400 --> 00:07:22,280
to that, you know, decision I I
guess on the.

124
00:07:22,280 --> 00:07:24,520
Topic Actually the very
beginning.

125
00:07:24,520 --> 00:07:28,200
At the very beginning I was
doing schemes for incompressible

126
00:07:28,200 --> 00:07:30,600
flow.
It was I started with the lid

127
00:07:30,600 --> 00:07:35,080
driven cavity and the velocity
vorticity formulation of the

128
00:07:35,080 --> 00:07:38,840
nagastric equation.
So that was my master tests, but

129
00:07:38,840 --> 00:07:43,800
then when I was looking for a pH
for a PhD in France, actually

130
00:07:43,800 --> 00:07:47,480
the beginning should have been
just a second master degree.

131
00:07:47,560 --> 00:07:50,440
And after I decided to remain
for the PhD in France.

132
00:07:50,920 --> 00:07:56,200
But anyway, so my, my supervisor
in Italy knew Professor Alano,

133
00:07:56,720 --> 00:08:00,280
who was one of the founders of
the ICCFD conference too.

134
00:08:01,000 --> 00:08:03,960
And he said, well, I know this
guy in France is doing good job

135
00:08:03,960 --> 00:08:06,520
with, but it's compressible.
So it's another thing.

136
00:08:06,600 --> 00:08:09,680
And then if you don't, you're
not scared about compressible

137
00:08:09,680 --> 00:08:12,120
flows, you can go to him and
he's very good.

138
00:08:12,120 --> 00:08:15,320
She's doing, she's doing
numerical schemes and so on.

139
00:08:15,760 --> 00:08:18,800
And so that's how I decided, oh,
Paris is not bad.

140
00:08:18,800 --> 00:08:21,720
Actually it was because of Paris
more than because of the

141
00:08:21,760 --> 00:08:24,840
compressible schemes, because my
options were to go to the

142
00:08:24,840 --> 00:08:29,560
Fonkerman Institute, to go to
the US or to go to Paris.

143
00:08:29,560 --> 00:08:33,240
And what I said about Paris, I
have to go to Paris.

144
00:08:34,240 --> 00:08:39,000
And so I arrived here and
another proposed to me to work

145
00:08:39,000 --> 00:08:43,440
on high order schemes which were
a quite recent topic at that

146
00:08:43,440 --> 00:08:46,240
time.
You know, at that time high

147
00:08:46,240 --> 00:08:49,600
order was second order schemes
actually because everybody was

148
00:08:49,600 --> 00:08:52,400
doing 1st order.
So 2nd order scheme were already

149
00:08:52,400 --> 00:08:54,760
high order.
And he said to me, but we are

150
00:08:54,760 --> 00:08:57,280
going to move to 3rd order.
OK, yeah.

151
00:08:57,720 --> 00:09:00,600
And that's, you do unsteady
flows because at the, you know,

152
00:09:00,600 --> 00:09:04,560
at that time people were mostly
doing steady oiler or steady

153
00:09:04,560 --> 00:09:08,120
Navy spokes.
And so, well, I started a PhD

154
00:09:08,120 --> 00:09:12,080
for high order scheme to capture
steady phenomena, even if it was

155
00:09:12,080 --> 00:09:16,760
only runs at the time.
And, and, and yes, it was a high

156
00:09:16,760 --> 00:09:19,680
order finding volume schemes.
And then I spent quite a lot of

157
00:09:19,680 --> 00:09:23,120
my career on on high order
scheme because of that.

158
00:09:24,360 --> 00:09:28,240
And, and was it therefore a
natural progression to continue

159
00:09:28,240 --> 00:09:32,480
going down the route to sort of
post doc and, and, and you know,

160
00:09:32,480 --> 00:09:36,560
faculty position with was it
just an evolution or, or was it

161
00:09:37,480 --> 00:09:43,160
a, a challenge to, to, to
progress down that route?

162
00:09:43,880 --> 00:09:48,000
It was a challenge because as
you know, there are not many

163
00:09:48,000 --> 00:09:51,440
positions in academia.
They're very challenging.

164
00:09:53,320 --> 00:09:59,320
Also, when I, when I was in
France, I spent quite one year,

165
00:09:59,800 --> 00:10:03,040
the second master degree and
then three years of PhD in

166
00:10:03,040 --> 00:10:06,040
France.
And at some point I wanted to go

167
00:10:06,040 --> 00:10:09,400
back to Italy.
And in Italy there were zero

168
00:10:09,400 --> 00:10:16,680
positions basically.
So I passed the competitions in

169
00:10:16,680 --> 00:10:19,520
France because in France, you
know, the positions are open,

170
00:10:19,520 --> 00:10:22,840
you have to to get a
qualification, which is a

171
00:10:22,840 --> 00:10:26,160
national qualification.
So this depends on how many

172
00:10:26,160 --> 00:10:29,840
papers if you have taught
courses not and so on.

173
00:10:30,200 --> 00:10:34,840
And you it's basically sort of
minimum certification, which

174
00:10:34,840 --> 00:10:39,520
says, OK, you are fit to be a
professor.

175
00:10:39,760 --> 00:10:42,600
OK, not, not a professor, but an
assistant professor.

176
00:10:43,560 --> 00:10:47,800
And then we once you have got
got, once you got the, the, the

177
00:10:48,040 --> 00:10:52,120
qualification, you have to apply
at the universities which have

178
00:10:52,120 --> 00:10:54,800
positions open and you have to
compete with other guys.

179
00:10:54,800 --> 00:11:00,760
So I eventually got a position
here in Paris, but I wanted to

180
00:11:00,760 --> 00:11:04,680
go back to Italy.
So I renounced the position,

181
00:11:04,680 --> 00:11:09,000
even it was rank at first.
And I decided to go back to

182
00:11:09,000 --> 00:11:11,760
Italy.
And there I started with a post

183
00:11:11,760 --> 00:11:15,040
doc because there were no
permanent positions open.

184
00:11:15,880 --> 00:11:21,520
And after some time after my
post doc, I moved to a close by

185
00:11:21,520 --> 00:11:25,320
university because, well, my
home university is Bari.

186
00:11:25,360 --> 00:11:27,800
It's in the South of Italy, it's
Apulia.

187
00:11:28,200 --> 00:11:32,240
And I went to Leche, which is
even even more S you know, it's

188
00:11:32,760 --> 00:11:39,000
the very tip of the hill.
And, and there I got the

189
00:11:39,000 --> 00:11:43,320
eventually a faculty position
and I worked there for eight

190
00:11:43,320 --> 00:11:48,200
years as an assistant professor.
And I was the only assistant

191
00:11:48,200 --> 00:11:51,120
professor in fluid dynamics of
the whole university.

192
00:11:51,240 --> 00:11:58,440
I was the only one.
And, and what was what was your

193
00:11:58,440 --> 00:12:02,640
main focus back was back then,
you know, when did this move to

194
00:12:02,640 --> 00:12:06,600
the the Rands modelling and
certainty quantification and

195
00:12:06,600 --> 00:12:08,840
things like that?
Was that during that period or

196
00:12:08,840 --> 00:12:11,000
was that?
Actually when I arrived in

197
00:12:11,000 --> 00:12:16,480
lecture, I was, as I said, I was
the only free mechanicist and I

198
00:12:16,480 --> 00:12:21,400
wasn't part of a group, a larger
group in energetics.

199
00:12:22,280 --> 00:12:26,560
And I started to think about
what could I do which is related

200
00:12:26,560 --> 00:12:31,240
to energetics.
And I discovered almost by by by

201
00:12:31,240 --> 00:12:35,800
chance dense gases.
And so dense gases are

202
00:12:35,800 --> 00:12:39,760
compressible.
So it's it's gases.

203
00:12:39,760 --> 00:12:43,560
So it's compressible.
Compressible flows of organic

204
00:12:43,560 --> 00:12:47,040
fluids, which are governed by
complex equations of states

205
00:12:47,040 --> 00:12:50,640
could be super critical CO2, but
it's it's light gas.

206
00:12:51,040 --> 00:12:55,000
But you can have denser gases or
more heavier gases like

207
00:12:55,400 --> 00:13:01,080
refrigerants or hydrocarbons.
And these gases are used in

208
00:13:01,080 --> 00:13:06,160
processes, industrial processes,
or also in some thermodynamic

209
00:13:06,160 --> 00:13:09,920
cycles like, well, of course the
refrigeration cycles, but also

210
00:13:11,040 --> 00:13:14,880
direct cycles, which means, for
instance, organic Franklin

211
00:13:14,880 --> 00:13:17,920
cycles, which are like the
ranking cycle, which works with

212
00:13:17,920 --> 00:13:20,400
these strange gases instead of
water.

213
00:13:21,360 --> 00:13:26,080
And the problem with these gases
is that since it is industrial,

214
00:13:26,880 --> 00:13:29,920
industrial fluids they are very
ill characterized.

215
00:13:30,240 --> 00:13:33,880
Actually you don't know exactly
about the properties about

216
00:13:33,920 --> 00:13:36,680
equational state you should use
there are few data.

217
00:13:37,600 --> 00:13:43,440
The material properties are
given in only technical sheets

218
00:13:43,880 --> 00:13:46,440
with a lot of uncertainties.
And that's how I came to

219
00:13:46,600 --> 00:13:49,560
uncertainty quantification,
because at some point I

220
00:13:49,560 --> 00:13:52,440
discovered polynomial care stuff
and so on.

221
00:13:52,680 --> 00:13:56,560
And I said, oh, this is perfect
because I have a fluid where

222
00:13:56,600 --> 00:13:58,960
actually don't know exactly how
it behaves.

223
00:13:59,280 --> 00:14:03,600
So I would like to characterize
the impact on the CFD solution

224
00:14:04,000 --> 00:14:07,840
of the, the, the, the bad
knowledge of the fluid

225
00:14:07,840 --> 00:14:10,600
properties, not only the
thermodynamics, but also the

226
00:14:10,600 --> 00:14:13,160
transport properties are not
very well known.

227
00:14:13,880 --> 00:14:17,120
And so, yeah, that's how I moved
to uncertainty quantification.

228
00:14:17,200 --> 00:14:21,080
But at the beginning was not
runs, it was even Euler, you

229
00:14:21,080 --> 00:14:26,440
know, but you don't know exactly
the thermodynamic, you cannot

230
00:14:26,440 --> 00:14:29,200
characterize completely the
thermodynamic behavior of your

231
00:14:29,200 --> 00:14:33,600
gas because you, you, you don't
know which equation of state you

232
00:14:33,600 --> 00:14:36,560
should use.
And the parameters of the

233
00:14:36,560 --> 00:14:39,800
equation of state are, you know,
not very accurate.

234
00:14:42,600 --> 00:14:44,720
That's interesting.
So it's more driven from that.

235
00:14:44,960 --> 00:14:51,440
Yeah, I guess did that help you
working in a more complex area

236
00:14:51,440 --> 00:14:52,640
in a way?
Did it give you more

237
00:14:52,640 --> 00:14:55,600
understanding of the
fundamentals of fluid mechanics?

238
00:14:55,600 --> 00:14:58,600
I guess when people just study
single phase simple

239
00:14:58,600 --> 00:15:01,120
incompressible, maybe they don't
appreciate the.

240
00:15:01,200 --> 00:15:05,560
Complexity, yeah.
Actually I I adore those gases

241
00:15:05,560 --> 00:15:09,880
because, well at least
theoretically they could exhibit

242
00:15:09,880 --> 00:15:13,400
very exotic behaviors.
In particular they are expected

243
00:15:13,400 --> 00:15:16,240
to exhibit under some
thermodynamic conditions.

244
00:15:16,240 --> 00:15:20,800
If if you take a heavy enough
gas, it is expected to exhibit

245
00:15:20,800 --> 00:15:23,680
expansion shock waves.
And actually there is a

246
00:15:23,680 --> 00:15:28,280
community around that which is
called the non ideal

247
00:15:28,680 --> 00:15:31,800
compressible pseudo dynamics
community, which has spent

248
00:15:31,800 --> 00:15:36,800
several years trying to expect
to to have an experimental proof

249
00:15:36,920 --> 00:15:40,520
of the existence of these
expansion shock waves in single

250
00:15:40,520 --> 00:15:44,960
phase compressible flows.
So the first one to postulate

251
00:15:44,960 --> 00:15:47,440
that is Hans Bitte.
So the the physician, the

252
00:15:47,520 --> 00:15:53,800
physicist Hans Bitte, Nobel
Prize who showed that for Vander

253
00:15:53,800 --> 00:16:00,720
Waals gases with some values of
the of the polytropic exponent,

254
00:16:01,440 --> 00:16:05,520
you can get a region where the
second principle of

255
00:16:05,520 --> 00:16:10,680
thermodynamics forbids the the
classical shock waves.

256
00:16:10,680 --> 00:16:12,560
So the the the compression shock
waves.

257
00:16:12,840 --> 00:16:17,520
So instead you have a
compression fan and conversely,

258
00:16:17,520 --> 00:16:21,440
you can have expansion shock
waves and instead of expansion

259
00:16:21,440 --> 00:16:23,400
fan.
So everything is reversed.

260
00:16:24,040 --> 00:16:27,120
And then several researchers
work on that.

261
00:16:27,120 --> 00:16:32,920
Thompson, Michael Kramer at at
Virginia Tech and so on.

262
00:16:33,400 --> 00:16:36,760
And people were fascinated by
these gases because everybody

263
00:16:36,760 --> 00:16:40,720
wanted to prove experimentally
that that was possible.

264
00:16:40,720 --> 00:16:44,200
And eventually there was a group
in Russia at some point we did

265
00:16:44,200 --> 00:16:49,280
an experiment with FC-70, which
is a, you know, fluorocarbon,

266
00:16:49,280 --> 00:16:51,920
those which are very, very bad
for the ozone layer.

267
00:16:52,480 --> 00:16:55,440
So they did an experiment with
this and they said, oh, here,

268
00:16:55,640 --> 00:16:57,440
here you have the expansion
shockwaves.

269
00:16:57,800 --> 00:17:01,680
But then the people started to
challenge the experiment and

270
00:17:01,680 --> 00:17:05,240
said no, this is impossible,
Probably it was to face and so

271
00:17:05,240 --> 00:17:07,680
on.
And then in depth there was a

272
00:17:07,680 --> 00:17:09,960
group.
So the group of Piero Corona,

273
00:17:10,200 --> 00:17:14,920
there is a big group in TU Delft
who tried to reproduce an

274
00:17:14,920 --> 00:17:18,680
experiment with different gases.
Also you have Alberto Guardo,

275
00:17:18,680 --> 00:17:22,000
name Milan.
He, he got an ERC actually on

276
00:17:22,000 --> 00:17:24,560
that.
And what it's very, very

277
00:17:24,560 --> 00:17:28,240
difficult because this shock
expansion shock waves exist in a

278
00:17:28,240 --> 00:17:31,640
very tiny thermodynamic region
and all the uncertainties

279
00:17:31,640 --> 00:17:35,440
associated with the shock tube
with the membrane breaking and

280
00:17:35,440 --> 00:17:39,320
so on can perturb the actual
development of the shockwave.

281
00:17:39,320 --> 00:17:43,440
And so we are never sure it's
really a pure expansion

282
00:17:43,440 --> 00:17:47,960
shockwave.
So, well, and the interest of

283
00:17:47,960 --> 00:17:50,800
that thing besides the, let's
say the, the, the, the

284
00:17:50,800 --> 00:17:55,800
scientific curiosity was that
some people were expecting that

285
00:17:55,920 --> 00:18:01,400
you could exploit this behaviour
to get better energy conversion

286
00:18:01,400 --> 00:18:04,280
cycle with the reduced losses
and so on.

287
00:18:05,600 --> 00:18:10,640
It has been abandoned right now
a little bit because actually

288
00:18:11,600 --> 00:18:14,440
while there is a very small
industry supporting that because

289
00:18:14,560 --> 00:18:19,000
it's very specific machines and
it's very small companies

290
00:18:19,000 --> 00:18:22,880
working on that with few funds
for to invest in research.

291
00:18:23,560 --> 00:18:28,080
But it was a great period.
So it was actually, and also it

292
00:18:28,080 --> 00:18:30,800
has the impact, even if you
don't find the expansion shock

293
00:18:30,800 --> 00:18:33,320
waves, at some point, you don't
care because there are plenty of

294
00:18:33,320 --> 00:18:37,120
systems with real gas effects,
not expansion shockwave, but

295
00:18:37,120 --> 00:18:40,040
still real gas effects.
And that you need to

296
00:18:40,040 --> 00:18:44,720
characterize to have better
conversion cycles or better heat

297
00:18:44,720 --> 00:18:47,120
pumps or, you know, all these
stuff.

298
00:18:47,200 --> 00:18:50,840
And so we're still working.
So we are right now a project

299
00:18:50,840 --> 00:18:53,960
ongoing with the, with, with the
German team.

300
00:18:54,360 --> 00:18:57,520
We are doing the, the
simulations and the machine

301
00:18:57,520 --> 00:19:01,120
learning and they are doing the
experiments and it's a very nice

302
00:19:01,120 --> 00:19:03,160
team.
And yeah, we, yeah.

303
00:19:04,120 --> 00:19:08,760
Very cool.
And you, you mentioned at the

304
00:19:08,760 --> 00:19:12,240
beginning that, you know, you
were in Italy, you went to

305
00:19:12,240 --> 00:19:16,680
Paris, you went back to Italy.
What then was the what?

306
00:19:16,680 --> 00:19:20,240
What led to you, you know, going
back to Paris again?

307
00:19:21,680 --> 00:19:26,400
Yeah, well, again, I mean, it's
not easy to stay in academia in

308
00:19:26,400 --> 00:19:32,160
Italy because the, you know, the
funding system, well, the

309
00:19:32,160 --> 00:19:36,520
academic system in Italy is
chronically underfunded, but

310
00:19:36,520 --> 00:19:38,880
severely underfunded.
That's why you find Italians

311
00:19:38,880 --> 00:19:41,080
everywhere.
Actually, I've noticed maybe

312
00:19:41,080 --> 00:19:43,360
that there are lots of Italians
everywhere.

313
00:19:45,280 --> 00:19:48,760
So, so the, the academic
position are very few.

314
00:19:48,760 --> 00:19:52,200
I had one actually.
But then you are alone.

315
00:19:52,280 --> 00:19:54,720
It's very difficult to get
funding.

316
00:19:55,840 --> 00:20:00,280
The possibility of having a
career are very, So you, you

317
00:20:00,400 --> 00:20:02,760
have to be very patient, you
know.

318
00:20:03,400 --> 00:20:10,840
And also, my husband wasn't
Italian and he never could find

319
00:20:10,840 --> 00:20:14,400
a satisfactory job in Italy.
So at some point he said, well,

320
00:20:14,400 --> 00:20:18,680
let's go back to Paris.
And there I applied because I

321
00:20:18,680 --> 00:20:20,560
wanted to move to a
professorship.

322
00:20:20,560 --> 00:20:24,560
I was assistant professor, I
wanted to move a professor and I

323
00:20:24,960 --> 00:20:28,000
got a position in Paris.
And that's how we we moved back

324
00:20:28,000 --> 00:20:31,000
to Paris because he also had a
job here in Paris.

325
00:20:32,760 --> 00:20:37,800
Ah, OK, I understand.
So you well, as you say, Paris

326
00:20:37,800 --> 00:20:41,040
is not such a bad place to to to
be.

327
00:20:41,480 --> 00:20:45,600
Easy for two person It's easier
to find both a job in Paris than

328
00:20:45,600 --> 00:20:48,520
find both a job in the in
southern Italy.

329
00:20:49,560 --> 00:20:55,680
Yeah, but that does seem to be,
I guess, an overriding thing of

330
00:20:55,680 --> 00:21:00,880
academia that it's almost this
necessity to move, right.

331
00:21:00,880 --> 00:21:02,440
It's it's kind of a weird
profession.

332
00:21:02,440 --> 00:21:06,160
Even in the US or the UK, very
few people will have their

333
00:21:06,160 --> 00:21:09,400
entire career at one place,
particularly earlier on.

334
00:21:09,640 --> 00:21:15,120
They will have to have you found
that it's almost, it's a fairly

335
00:21:15,120 --> 00:21:18,840
unfair thing in a way that you
have to move your family and do

336
00:21:18,840 --> 00:21:22,520
things just to progress.
Yeah, yeah, that that that's

337
00:21:22,520 --> 00:21:26,760
true.
And that's because, well, for

338
00:21:26,760 --> 00:21:29,600
instance, in France, there is
even a rule in some, in some

339
00:21:29,600 --> 00:21:32,120
disciplines, like in
mathematics, that if you have

340
00:21:32,120 --> 00:21:36,520
been an assistant professor in a
department, you cannot be a

341
00:21:37,280 --> 00:21:39,000
professor in the same
department.

342
00:21:39,000 --> 00:21:43,040
You have to move another one.
And these small cities where you

343
00:21:43,040 --> 00:21:46,920
have just one university, it's
basically this means that you

344
00:21:46,920 --> 00:21:49,600
have to move.
So this pushes a lot of people

345
00:21:49,600 --> 00:21:53,720
to abandon the idea of moving to
a professorship and they remain

346
00:21:53,720 --> 00:21:56,240
assistant professors for
forever.

347
00:21:56,600 --> 00:21:58,600
Also because the French system
allows it.

348
00:21:58,600 --> 00:22:00,880
It's not like the tenure track
in the US.

349
00:22:00,880 --> 00:22:04,160
If you don't, if you are not
promoted to associate, then you

350
00:22:04,160 --> 00:22:08,280
have to leave academia.
You can remain as associate

351
00:22:08,280 --> 00:22:10,880
professor with a permanent
position forever.

352
00:22:11,240 --> 00:22:16,280
So if you can't move with your
family, basically you cannot

353
00:22:17,280 --> 00:22:22,120
progress in your career.
And and so it's a choice.

354
00:22:23,720 --> 00:22:27,080
Some people are happy with the,
with an assistant professorship

355
00:22:27,080 --> 00:22:29,400
for their life, they do teaching
and so on.

356
00:22:30,240 --> 00:22:34,240
But if you, if you want to move,
so if you want to progress with

357
00:22:34,240 --> 00:22:37,080
the career, yes.
The, the, I don't know, the

358
00:22:37,080 --> 00:22:41,400
academic, the academic world is
built like that.

359
00:22:41,400 --> 00:22:45,680
Why don't know exactly it's also
part of science to move.

360
00:22:45,680 --> 00:22:49,240
We should look in the past at
the the former scientists where

361
00:22:49,240 --> 00:22:53,280
they were moving quite a lot.
You know, even in the

362
00:22:53,280 --> 00:22:56,720
Renaissance, you know, you, you,
you had this, I don't know

363
00:22:56,720 --> 00:23:00,680
Galileo or or Leonardo da Vinci,
they were moving around.

364
00:23:01,320 --> 00:23:06,560
And that's because you need to
to spread knowledge to find

365
00:23:06,560 --> 00:23:09,440
better environment.
So I think it's nice the

366
00:23:09,440 --> 00:23:14,640
difficulties when you also have
a family life and you have to,

367
00:23:14,880 --> 00:23:17,040
so you have two person to, to,
to move.

368
00:23:17,920 --> 00:23:22,400
In some cases it's easy because
1 is a flexible job or, or

369
00:23:22,600 --> 00:23:25,960
things at home.
But otherwise it can be, yeah,

370
00:23:26,720 --> 00:23:29,280
quite challenging.
And I think it's something that

371
00:23:29,280 --> 00:23:33,360
the new generation don't
appreciate that much because,

372
00:23:33,400 --> 00:23:38,400
you know, I, I think I am
what's, what's my generation, I

373
00:23:38,400 --> 00:23:43,240
think generation X and we are,
you know, I, I know that my

374
00:23:43,240 --> 00:23:46,520
children say your generation X
is the generation of suffering.

375
00:23:46,600 --> 00:23:51,120
You know you were you were
raised to suffer, but it's no

376
00:23:51,120 --> 00:23:55,880
longer the case.
But it, but it is a serious

377
00:23:55,880 --> 00:23:59,440
point though, that I've, I
almost feel like sometimes

378
00:23:59,440 --> 00:24:06,800
academia has maybe not to be
careful, but I always, I'm just

379
00:24:06,800 --> 00:24:11,840
amazed that the dedication it
requires to get through it and

380
00:24:11,840 --> 00:24:14,520
to to become Someone Like You,
you know, a top professor.

381
00:24:14,520 --> 00:24:21,880
Like it feels, it's, I feel like
probably many people aren't able

382
00:24:21,880 --> 00:24:27,480
to get to that .1 Obviously
they're not intellectual enough.

383
00:24:27,480 --> 00:24:28,840
They're not, you know, capable
enough.

384
00:24:29,320 --> 00:24:32,240
But but also you have to have a
quiet determination.

385
00:24:32,320 --> 00:24:36,080
Yeah, to, to get there, which is
probably partly a good thing

386
00:24:36,080 --> 00:24:38,960
because it, you know, it's, it's
like a, you know, survival of

387
00:24:38,960 --> 00:24:42,720
the fittest, I guess.
But as you've said to some

388
00:24:42,720 --> 00:24:46,520
people where they just really
would love to stay where they're

389
00:24:46,520 --> 00:24:49,800
living because they have friends
and family and they can't, it's,

390
00:24:50,920 --> 00:24:53,160
it's a shame.
Well, what I used to say to my

391
00:24:53,160 --> 00:24:57,800
PhD students is that the
academic career is kind of

392
00:24:57,800 --> 00:25:04,120
similar to a career in music or
in theatre or in high level

393
00:25:04,160 --> 00:25:07,480
level sport.
OK, if you are a musician, if

394
00:25:07,480 --> 00:25:11,080
you're a pianist or if you are
an opera singer, you have to

395
00:25:11,080 --> 00:25:14,280
move around and it's a very
competitive field.

396
00:25:14,280 --> 00:25:18,360
And you do that not because you
want no quiet life, a quiet

397
00:25:18,360 --> 00:25:22,200
life, but because you are
passioned and you want to be on

398
00:25:22,200 --> 00:25:26,520
the stage and see, you know,
people around you clapping

399
00:25:26,520 --> 00:25:29,400
hands.
And, you know, it's a sort of,

400
00:25:29,760 --> 00:25:34,760
you know, I think that well,
there is a sort of narcissistic

401
00:25:34,760 --> 00:25:39,160
side, maybe like people
remaining in academia, but also,

402
00:25:39,240 --> 00:25:42,240
you know, it's it's an
intellectual work.

403
00:25:42,360 --> 00:25:48,680
And I guess that what if you,
you should you, you do that

404
00:25:48,680 --> 00:25:52,200
because because of passion,
first of all, it's not, it's not

405
00:25:52,200 --> 00:25:58,200
an ordinary job actually.
And you want to push the, the,

406
00:25:58,200 --> 00:26:01,960
the, the, the, the, the, the,
the frontiers of knowledge.

407
00:26:02,280 --> 00:26:07,640
You want your, your work to have
an impact and you know, you want

408
00:26:07,640 --> 00:26:10,040
to do something to move things
forward.

409
00:26:10,440 --> 00:26:12,960
And, and that's a, that's a
passion.

410
00:26:13,080 --> 00:26:19,480
So it's maybe you don't, you
don't make it, but you still try

411
00:26:19,480 --> 00:26:22,240
because you know, it's, it's
like a football player.

412
00:26:22,240 --> 00:26:26,000
Not not all football players are
Lionel Messi or, I don't know,

413
00:26:26,360 --> 00:26:30,880
David Beckham.
That's a very good.

414
00:26:30,880 --> 00:26:35,400
I've not heard it described that
way about a musician or sports

415
00:26:35,400 --> 00:26:37,000
players.
It's kind of true that they also

416
00:26:37,000 --> 00:26:41,200
move around a lot.
And I, I get that's the, that's

417
00:26:41,200 --> 00:26:44,920
the, the old thing though of
universities, isn't it, that you

418
00:26:44,920 --> 00:26:47,400
almost have to wait for the
person above you to, you know,

419
00:26:47,800 --> 00:26:49,680
retire or die to get their
position.

420
00:26:49,680 --> 00:26:53,040
So people end up also moving
around because there's sort of

421
00:26:53,120 --> 00:26:56,960
not enough where I guess in a
big corporate world, people are

422
00:26:56,960 --> 00:27:00,920
more able to, maybe that's
changing, But you know, in a, in

423
00:27:00,920 --> 00:27:03,840
like an Airbus or something,
I'm, I'm sure there's jobs for

424
00:27:03,840 --> 00:27:08,120
life almost just in one company
where academia.

425
00:27:08,120 --> 00:27:11,800
So I, I, I say that because I
think it, people should

426
00:27:11,800 --> 00:27:16,600
appreciate just how hard it is
to get to being a professor,

427
00:27:16,600 --> 00:27:18,560
that it, it's not just an
intellectual thing.

428
00:27:18,560 --> 00:27:23,320
It's like a determination and
passion thing, as you say, to,

429
00:27:23,520 --> 00:27:28,320
to, to do it.
And so all, you know, talking

430
00:27:28,320 --> 00:27:30,880
about, you know, wanting to be
at the forefront.

431
00:27:31,840 --> 00:27:35,640
I mean, you, you were quite
early on looking at learning and

432
00:27:35,640 --> 00:27:38,320
data-driven turbines models and
and machine learning.

433
00:27:38,680 --> 00:27:41,080
You know, when did you start to
get into that?

434
00:27:41,080 --> 00:27:44,960
When did you get a sense that
these methods were an

435
00:27:44,960 --> 00:27:49,120
alternative or an enhancement on
the more traditional numerical

436
00:27:49,120 --> 00:27:51,000
methods sort of line of
research?

437
00:27:51,800 --> 00:27:55,120
Well, actually I.
Stepped into Bayesian methods

438
00:27:55,240 --> 00:27:58,720
when I was doing uncertainty
quantification for dense gases.

439
00:27:59,480 --> 00:28:03,880
And at some point I was
discussing with some colleagues

440
00:28:03,880 --> 00:28:08,480
of mine from from the University
of Trieste, and they had a sort

441
00:28:08,480 --> 00:28:12,560
of startup, I don't know if you
know this startup called Esteco.

442
00:28:12,640 --> 00:28:17,480
They produce software called
Mode Frontier.

443
00:28:18,080 --> 00:28:20,440
Oh yes.
Yes, yes, of course, yeah.

444
00:28:21,400 --> 00:28:25,160
And so, yeah, they were actually
promoting this software in

445
00:28:25,160 --> 00:28:31,000
universities and, and they sold
also academic licenses and so

446
00:28:31,000 --> 00:28:33,760
on.
And so I started to discuss with

447
00:28:33,760 --> 00:28:36,960
them about my problems with
uncertainty quantification in

448
00:28:36,960 --> 00:28:40,480
dense gases.
And they said, OK, but you could

449
00:28:40,480 --> 00:28:42,840
try to solve a numerous problem
and so on.

450
00:28:42,840 --> 00:28:47,440
And, and, and so that I, so I
started to look into the

451
00:28:47,440 --> 00:28:52,320
literature and I started to be
interested into these Bayesian

452
00:28:52,320 --> 00:28:58,840
methods and, and I said, OK,
this could work for dense gases,

453
00:28:58,840 --> 00:29:01,680
but we, well, the thermodynamics
is not the only source of

454
00:29:01,680 --> 00:29:04,360
uncertainty because you also
have the uncertainties

455
00:29:04,360 --> 00:29:05,960
associated with the turbulence
models.

456
00:29:05,960 --> 00:29:09,800
And these are very old problems.
So there are so many turbulence

457
00:29:09,800 --> 00:29:12,720
models, you don't know which one
you have to choose, you don't

458
00:29:12,720 --> 00:29:15,720
know which parameters you should
put into them and so on.

459
00:29:15,720 --> 00:29:20,120
So let's try to quantify this
statistically instead of just

460
00:29:20,560 --> 00:29:24,800
using expert knowledge, which is
what any anybody does actually,

461
00:29:25,000 --> 00:29:31,880
because every, every company
actually has sort of, you know,

462
00:29:33,240 --> 00:29:37,320
establish Noahu saying OK, for
this problem, you should use the

463
00:29:37,360 --> 00:29:41,240
KE Omega SST for this problem,
you should use the Sphalatalmers

464
00:29:41,240 --> 00:29:42,960
for this problem, you should use
that one.

465
00:29:43,160 --> 00:29:46,440
And sometimes you also know that
they retune the parameters.

466
00:29:46,800 --> 00:29:51,320
For instance, I know that people
in internal combustion engine

467
00:29:51,320 --> 00:29:56,840
used to recalibrate KE epsilon
for for having better results

468
00:29:56,840 --> 00:30:00,320
for, for, for, for, Yeah, for
internal combustion engines.

469
00:30:00,760 --> 00:30:04,240
And but all this was was tuned
by hands basically.

470
00:30:04,640 --> 00:30:08,840
And I said, OK, if there are,
you know, mathematical, clean

471
00:30:08,840 --> 00:30:12,240
mathematical techniques to do
that, let's try to use them.

472
00:30:12,240 --> 00:30:16,040
And so since I was already doing
direct uncertainty

473
00:30:16,040 --> 00:30:20,400
quantification, I started to,
you know, try to do the the

474
00:30:20,400 --> 00:30:22,440
backward uncertainty
verification.

475
00:30:23,320 --> 00:30:26,680
And at some point, well, it was
a bit hard for me because you

476
00:30:26,680 --> 00:30:30,520
know, in mechanical engineering
we are not very well trained in,

477
00:30:30,560 --> 00:30:34,440
in probability and statistics.
So eventually.

478
00:30:34,440 --> 00:30:39,360
I, I studied that myself because
it was interesting to that

479
00:30:39,360 --> 00:30:44,320
actually I also thought a little
bit, but it was not enough to,

480
00:30:44,400 --> 00:30:46,040
you know, understand all the
details.

481
00:30:46,520 --> 00:30:51,120
And so at some point I was
invited by statistics department

482
00:30:51,120 --> 00:30:55,880
in Chile and they invited me for
two months and I was there with

483
00:30:55,880 --> 00:30:59,680
statisticians and mathematicians
of probabilities.

484
00:30:59,680 --> 00:31:05,000
So and during the first month it
was impossible to understand

485
00:31:05,000 --> 00:31:08,920
each other because, you know, I
was calling things with names

486
00:31:08,920 --> 00:31:11,240
that they interpreted in another
way and so on.

487
00:31:12,240 --> 00:31:15,480
And, but at some point we
started to understand each other

488
00:31:15,480 --> 00:31:19,480
and that was great because some
papers which were, you know,

489
00:31:19,480 --> 00:31:26,600
just giraglyphs to me,
giraglyphs started to be clear.

490
00:31:26,720 --> 00:31:31,760
And, and that's where, you know,
all the, the Bayesian stuff was,

491
00:31:31,800 --> 00:31:35,640
was set and, and, and I could
start to recalibrate the

492
00:31:35,640 --> 00:31:37,880
turbulence models and and so on.
So that.

493
00:31:38,520 --> 00:31:42,080
So yeah, the fact of being in an
interdisciplinary environment

494
00:31:42,080 --> 00:31:46,880
and talk with mathematicians was
extremely useful to me, not only

495
00:31:46,880 --> 00:31:49,800
with engineers, even if I love
engineers, but you know,

496
00:31:50,680 --> 00:31:53,840
sometimes like, no, that.
That's, that's a very good

497
00:31:53,840 --> 00:31:57,800
point.
And I guess even to today,

498
00:31:57,800 --> 00:32:00,960
that's a challenge on the
machine learning side, isn't it,

499
00:32:00,960 --> 00:32:05,320
that maybe it's starting to
change in a course today.

500
00:32:05,720 --> 00:32:08,840
But I guess most people who are
doing the research, therefore

501
00:32:08,840 --> 00:32:12,840
who studied, you know, 5 to 10
years ago or more didn't have

502
00:32:12,840 --> 00:32:16,960
any of that background in
computer science or in

503
00:32:16,960 --> 00:32:20,360
statistics or, or sort of
methods that maybe people who

504
00:32:20,360 --> 00:32:23,440
did more maths, stronger maths
may, may, may do it, but

505
00:32:24,000 --> 00:32:27,200
engineering courses probably
wouldn't.

506
00:32:27,200 --> 00:32:30,240
I mean, how have you seen that
affect the machine learning

507
00:32:30,240 --> 00:32:31,400
side?
I know you've been quite

508
00:32:31,400 --> 00:32:35,360
passionate about connecting.
I know you you kicked off, for

509
00:32:35,360 --> 00:32:40,200
example, with the extrality and
Air, Air France data sets.

510
00:32:40,200 --> 00:32:43,600
And, you know, did you try and
take some of that inspiration of

511
00:32:43,600 --> 00:32:47,160
your time getting familiar with
statistics and try and do the

512
00:32:47,160 --> 00:32:49,240
same on the machine learning
side?

513
00:32:49,240 --> 00:32:53,440
Yeah, yeah, definitely.
Actually, I had the chance when,

514
00:32:53,440 --> 00:32:56,200
when I, when I moved to Sorbonne
to be in a very

515
00:32:56,200 --> 00:33:04,120
interdisciplinary environment.
And at some point I, I, well, I

516
00:33:04,120 --> 00:33:07,560
stepped into the team.
We have a very strong machine

517
00:33:07,560 --> 00:33:11,560
learning team here in the
computer science department, and

518
00:33:12,600 --> 00:33:18,080
I discovered almost by chance
that they were doing machine

519
00:33:18,080 --> 00:33:21,200
learning for physics.
So I contacted them and said,

520
00:33:21,200 --> 00:33:24,480
OK, I'm a fluid mechanicist, I'm
trying to move to machine

521
00:33:24,480 --> 00:33:26,080
learning.
I know that you're doing machine

522
00:33:26,080 --> 00:33:28,440
learning and you're trying to
move to fluid mechanics.

523
00:33:28,880 --> 00:33:32,720
Can we do something together?
And so we applied together to an

524
00:33:32,720 --> 00:33:37,280
internal funding instrument of
the university and we got

525
00:33:37,280 --> 00:33:39,200
founded.
So we had this learned fluid

526
00:33:39,400 --> 00:33:42,480
team.
So it was a little bit of money

527
00:33:42,480 --> 00:33:48,320
to have a post doc and a few
interns and little things, but

528
00:33:48,320 --> 00:33:51,240
this allowed to connect each
other.

529
00:33:51,760 --> 00:33:55,840
And then they said, Hey, we have
this guy who is going to do

530
00:33:56,080 --> 00:33:58,200
machine learning methods for
CFD.

531
00:33:58,640 --> 00:34:02,840
So we want to, but we have no
databases machine learning.

532
00:34:02,840 --> 00:34:09,440
We, we have minis.
We have, you know, many famous

533
00:34:09,440 --> 00:34:13,120
databases, but there are not bad
databases in CFD.

534
00:34:13,120 --> 00:34:16,800
We want to build 1, but actually
we don't know exactly how to do

535
00:34:16,800 --> 00:34:20,679
the meshes because they were
trying to use, you know, open

536
00:34:20,679 --> 00:34:25,880
phone and we're using, you know,
as not PX.

537
00:34:25,880 --> 00:34:28,760
Yeah.
And of course the meshes are.

538
00:34:28,760 --> 00:34:31,520
Not very good close to the wall.
And that's the point.

539
00:34:31,920 --> 00:34:35,199
So I said no, but you can also
use the structured meshes and so

540
00:34:35,199 --> 00:34:38,320
on.
And then, and then also we

541
00:34:38,320 --> 00:34:41,960
started to discuss a little bit
about the criteria we had to use

542
00:34:41,960 --> 00:34:44,600
to evaluate the results.
Not only the MSC we have

543
00:34:44,600 --> 00:34:49,400
discussed, but you know, and
that's how well eventually they

544
00:34:49,400 --> 00:34:53,480
produce this Air France.
Of course, the well, the

545
00:34:53,719 --> 00:34:56,080
marriage is essentially the,
the, the students.

546
00:34:56,480 --> 00:34:59,200
It's something interesting that
this student had the background

547
00:34:59,200 --> 00:35:02,760
in physics the beginning, so and
he moved to machine learning

548
00:35:02,760 --> 00:35:05,880
too, but he had a background in
not in CFD but in physics.

549
00:35:06,400 --> 00:35:10,880
And so, yeah, so with this
mixture of disciplines, we came

550
00:35:10,880 --> 00:35:13,920
up with this, with this
database.

551
00:35:14,200 --> 00:35:17,440
And also they could test a lot
of baselines that the way they

552
00:35:17,440 --> 00:35:20,040
do it in machine learning or
they take 1 based and two

553
00:35:20,040 --> 00:35:22,160
baselines and they test all the
baselines.

554
00:35:22,920 --> 00:35:27,000
And now we are trying to do more
actually also for unsteady

555
00:35:27,000 --> 00:35:29,520
flows.
So hopefully we will have a new

556
00:35:29,520 --> 00:35:33,040
database coming out which is on
LES.

557
00:35:33,840 --> 00:35:37,520
So this times is not steady
runs, it's more LES and it's

558
00:35:37,560 --> 00:35:43,160
actually it snapshots because
you know to to have time

559
00:35:43,160 --> 00:35:46,120
resolved predictions, yes.
And.

560
00:35:46,360 --> 00:35:50,840
And the idea is the same.
So they, they well enlighten me

561
00:35:50,840 --> 00:35:53,720
on machine learning
architectures because they know

562
00:35:53,720 --> 00:35:56,960
better.
They know, they know better.

563
00:35:56,960 --> 00:36:03,680
Also the, the, the sometimes
the, you know, the, the, the,

564
00:36:03,720 --> 00:36:08,000
the pitfalls and the you can
have in training these things

565
00:36:08,000 --> 00:36:10,840
because sometimes, you know, the
training is not an easy.

566
00:36:12,760 --> 00:36:17,280
But on the other side, I say,
OK, maybe you should look at

567
00:36:17,280 --> 00:36:20,600
this and that you should use
this physical criteria and so

568
00:36:20,600 --> 00:36:22,760
on.
And it's very, very instructive,

569
00:36:22,800 --> 00:36:25,720
I think, for both sides.
So I'm very happy with this

570
00:36:25,720 --> 00:36:30,480
collaboration.
And yeah, that's that's how we.

571
00:36:32,360 --> 00:36:36,720
I mean, how much have you
struggled, though to get

572
00:36:36,720 --> 00:36:41,240
acceptance in that the from when
you started to now?

573
00:36:41,240 --> 00:36:43,000
Have you how have you seen
things progress?

574
00:36:43,000 --> 00:36:47,680
You know, the the argument of
how well is the model just a

575
00:36:47,680 --> 00:36:51,440
fancy interpolation versus
actually learning the physics?

576
00:36:51,440 --> 00:36:55,160
You know, how much is that is a
a good thing that this is hard

577
00:36:55,160 --> 00:36:59,400
questions and how much of it is
almost holding things back a

578
00:36:59,400 --> 00:37:01,160
little bit?
Well, the.

579
00:37:01,160 --> 00:37:03,440
Hardest thing, Well, I started
with the turbulence for this

580
00:37:03,440 --> 00:37:06,240
right?
And it was hard beginning

581
00:37:06,280 --> 00:37:11,320
because, you know, well,
turbulence models have a lot of

582
00:37:11,320 --> 00:37:13,640
knowledge.
They are really incredible.

583
00:37:13,640 --> 00:37:19,080
They have this physics sense,
which is, I don't know, but they

584
00:37:19,080 --> 00:37:24,480
are also very, you know, fond of
their methodology.

585
00:37:24,480 --> 00:37:27,680
They, you know, the fact that
you have do you have to do the

586
00:37:27,680 --> 00:37:32,200
things like in a certain way?
And also the parameters has

587
00:37:32,200 --> 00:37:36,760
been, have been the model
parameters been tuned making a

588
00:37:36,760 --> 00:37:40,640
lot of compromises and so on.
So they don't like that you

589
00:37:40,640 --> 00:37:43,480
start playing with the
parameters, playing with the

590
00:37:43,480 --> 00:37:47,640
terms and so on and so on.
And for sure at the beginning

591
00:37:48,520 --> 00:37:53,000
the community of people who were
who was playing with machine

592
00:37:53,000 --> 00:37:56,560
learning or, or calibration also
calibrations of of turbulence

593
00:37:56,560 --> 00:38:00,520
model was not a community of
turbulence models was a

594
00:38:00,640 --> 00:38:03,840
community of people coming like
me from numerical schemes

595
00:38:04,080 --> 00:38:06,160
actually or from numerics in
general.

596
00:38:07,200 --> 00:38:13,160
And so we didn't have the right
codes for provenance modelling,

597
00:38:13,480 --> 00:38:17,200
which is, you know, there is a
lot of knowledge accumulated for

598
00:38:17,400 --> 00:38:21,320
for decades.
And so many people were doing

599
00:38:21,320 --> 00:38:24,920
things that were not acceptable
actually from a strict

600
00:38:24,920 --> 00:38:26,320
turbulence modeling point of
view.

601
00:38:26,320 --> 00:38:30,000
So they were not using the right
features as the input of the

602
00:38:30,000 --> 00:38:32,000
turbulent model.
For instance, some people were

603
00:38:32,000 --> 00:38:36,280
using velocities at the
beginning, which, you know, it's

604
00:38:36,320 --> 00:38:39,280
it's non Galilean invariants and
so on.

605
00:38:39,800 --> 00:38:45,240
And and however, well, still
this model looked interesting.

606
00:38:45,240 --> 00:38:47,840
So everybody was intrigued with
the with the with the

607
00:38:47,840 --> 00:38:50,000
possibility of having it in this
model.

608
00:38:50,000 --> 00:38:52,360
Because in the end, if you look
at even if the to the

609
00:38:52,400 --> 00:38:56,880
traditional turbulence models,
they remain data-driven.

610
00:38:57,080 --> 00:39:01,120
They yeah, data-driven with a
human, you know, tuning the

611
00:39:01,120 --> 00:39:02,600
parameters.
But they are data-driven

612
00:39:02,720 --> 00:39:06,360
actually because you are you are
calibrating on, on some data

613
00:39:06,360 --> 00:39:08,200
sets which are the canonical
flows.

614
00:39:08,480 --> 00:39:12,440
But OK, so at some point, well
you had Chris Ramsey, which you

615
00:39:12,440 --> 00:39:19,600
know who, you know who decides
to organize a meeting at it was

616
00:39:19,600 --> 00:39:23,720
in Virginia, it was at the was
the name, the light, the

617
00:39:23,720 --> 00:39:25,200
lighthouse, right, the
lighthouse.

618
00:39:25,840 --> 00:39:30,240
So there was this mythical
meeting in 2022 where he said,

619
00:39:30,240 --> 00:39:33,400
OK, we are going to put around
the table classical turbulence

620
00:39:33,400 --> 00:39:36,560
modeler and these guys, these,
you know, these power venues

621
00:39:36,560 --> 00:39:37,640
with their machine learning
stuff.

622
00:39:39,680 --> 00:39:43,760
And and it was really
instructive because you had

623
00:39:43,760 --> 00:39:48,040
incredible guys there.
So it was a in honor of the 60th

624
00:39:48,040 --> 00:39:51,440
birthday of Lips Palat, who is a
great guy by the way.

625
00:39:52,000 --> 00:39:56,600
And you had, you know, Phillip,
you had Paul Durbin, you had all

626
00:39:56,600 --> 00:40:00,400
these guys, you know, you know,
plenty of things on turbulence

627
00:40:00,400 --> 00:40:02,640
modelling and they were
explaining things.

628
00:40:03,000 --> 00:40:06,200
And on the other hand, we were
trying to defend the idea of

629
00:40:06,200 --> 00:40:09,800
using machine learning.
So we were a little bit, you

630
00:40:09,800 --> 00:40:15,160
know, like how, how do you say
you attack?

631
00:40:16,680 --> 00:40:19,160
Attack.
But it was instructive because I

632
00:40:19,160 --> 00:40:22,440
think thanks to that, we
progressed a lot because we, we

633
00:40:22,440 --> 00:40:24,320
started with the idea, OK, it's
nice.

634
00:40:24,320 --> 00:40:29,320
We just fine tune a model for a
very narrow set of flows.

635
00:40:29,360 --> 00:40:31,200
We get better results.
We are happy with that.

636
00:40:31,760 --> 00:40:35,880
And now we are moving more and
more toward unifying models.

637
00:40:36,280 --> 00:40:40,280
And maybe, maybe, maybe I, I
maybe show ambition, but never

638
00:40:40,280 --> 00:40:43,400
know.
Maybe in 10 years more we could

639
00:40:43,400 --> 00:40:47,640
have a foundation model that's a
dream and turbulence model which

640
00:40:47,640 --> 00:40:52,200
could, you know, actually
realize the the the dream of

641
00:40:52,200 --> 00:40:55,600
turbulence model of a universal
turbulence model.

642
00:40:56,160 --> 00:41:01,120
I'm not sure we will get there,
but you know it is.

643
00:41:01,120 --> 00:41:07,280
Interesting though, because in
some ways turbulence modelling

644
00:41:08,400 --> 00:41:14,240
was going out of fashion.
And I know that it used to be

645
00:41:14,240 --> 00:41:16,880
the joke, didn't it?
I think that for a European

646
00:41:16,880 --> 00:41:19,000
project, if you said you're
going to work on turbans

647
00:41:19,000 --> 00:41:21,640
modelling, it was almost like
guaranteed to be rejected

648
00:41:22,160 --> 00:41:26,080
because it was seen as a solved
problem or you know what's new?

649
00:41:26,600 --> 00:41:32,760
And but there were the the irony
is is the industry still uses

650
00:41:32,760 --> 00:41:37,040
mainly RANS and are stuck using
methods from the 80s or or early

651
00:41:37,040 --> 00:41:39,800
90s seems like with machine
learning.

652
00:41:39,800 --> 00:41:43,960
Then there was this spike of of
of potential again.

653
00:41:44,640 --> 00:41:46,920
But I don't know about you, but
I almost felt that.

654
00:41:48,600 --> 00:41:51,200
Maybe some of the?
Use of machine learning for

655
00:41:51,200 --> 00:41:55,440
turbans modeling was a little
bit early and the expectation

656
00:41:55,440 --> 00:42:00,400
was so big that when it didn't
meet that expectation, it sort

657
00:42:00,400 --> 00:42:05,120
of then dropped down again.
And, and I, I, but I agree with

658
00:42:05,120 --> 00:42:11,400
you that in some ways with the
idea that we want to build also

659
00:42:11,400 --> 00:42:14,520
foundation models from a
surrogate modelling side, having

660
00:42:14,520 --> 00:42:17,600
to do everything with LES or
whatever, it's just so

661
00:42:17,600 --> 00:42:21,080
computationally expensive.
Ironically, if you could use

662
00:42:21,080 --> 00:42:24,440
machine learning to come up with
the ultimate turbots model, you

663
00:42:24,440 --> 00:42:29,520
would actually make it much more
affordable to, to, to actually

664
00:42:29,520 --> 00:42:33,080
run the simulations to achieve.
So I feel like now there is

665
00:42:33,080 --> 00:42:39,160
maybe a little bit more economic
or relevant again of the turbots

666
00:42:39,160 --> 00:42:45,840
modelling because it would save
so much on the date generation

667
00:42:45,840 --> 00:42:51,000
side.
But yeah, it it does seem to be,

668
00:42:51,600 --> 00:42:55,800
I don't know if you've noticed,
but I saw in CFD anyway that

669
00:42:55,800 --> 00:42:58,760
initially everybody was focused
on machine learning for turbans

670
00:42:58,760 --> 00:43:04,920
models, where now it seems to be
far more about machine learning

671
00:43:04,920 --> 00:43:10,800
for surrogate models.
And that seems to be much more

672
00:43:10,800 --> 00:43:15,120
focused on and you don't hear as
much on maybe the terms

673
00:43:15,120 --> 00:43:18,360
modelling in such a strong way.
Would would, would you tend to

674
00:43:18,360 --> 00:43:21,400
agree with that, that the
community sort of shifted a

675
00:43:21,400 --> 00:43:25,960
little bit?
Some communities, yes, I think

676
00:43:25,960 --> 00:43:31,840
in the, in the well in, in
complex engineering applications

677
00:43:32,000 --> 00:43:38,600
like well car industry or even
in solid mechanics for instance,

678
00:43:38,600 --> 00:43:40,760
they are shifting in that
direction.

679
00:43:41,760 --> 00:43:44,280
I think that in aerospace they
are still interesting to

680
00:43:44,280 --> 00:43:47,360
turbulence models, especially
for complex application like

681
00:43:47,360 --> 00:43:52,800
hypersonics or transition
models, all that.

682
00:43:54,320 --> 00:44:01,040
That's because, well, even if
you can, well what you would

683
00:44:01,040 --> 00:44:06,520
like to have in in for instance
in aerospace where we have what

684
00:44:06,520 --> 00:44:11,200
you would like to have is a high
fidelity model, so like Elias

685
00:44:11,200 --> 00:44:13,960
quality and to perform your
optimization with that.

686
00:44:13,960 --> 00:44:16,120
Why?
Because we know that France has

687
00:44:16,120 --> 00:44:20,960
flows, because we want to
explore extreme, extreme

688
00:44:20,960 --> 00:44:26,240
operating conditions, because we
are moving away from known paths

689
00:44:26,400 --> 00:44:29,480
like we are changing the fuels,
we are changing the

690
00:44:29,480 --> 00:44:34,720
architectures and so on.
So you cannot just design, I

691
00:44:34,720 --> 00:44:38,520
don't know, a new propeller or a
new wing based on epsilon

692
00:44:38,520 --> 00:44:41,960
modifications of something you
know and for which you know that

693
00:44:41,960 --> 00:44:44,680
the turbulence model is going to
be wrong, but you know more or

694
00:44:44,680 --> 00:44:48,000
less how it's going to fail and
how you should correct it.

695
00:44:48,280 --> 00:44:52,000
OK.
And the problem is that even if

696
00:44:52,000 --> 00:44:55,760
right now with the, the, the,
the Portage to GPUs and so on,

697
00:44:56,760 --> 00:45:00,720
high fidelity simulation are
becoming more affordable, still

698
00:45:00,720 --> 00:45:04,200
that's one simulation.
If you want to perform an

699
00:45:04,200 --> 00:45:08,680
optimization, you need thousands
of LES.

700
00:45:08,840 --> 00:45:13,120
And even if you can run a
complex world model LES on on a

701
00:45:13,160 --> 00:45:18,960
GPU, what you know, chatting,
you know, well, you know this,

702
00:45:20,440 --> 00:45:23,320
but still it's 1.
So if you want to perform, I

703
00:45:23,320 --> 00:45:28,160
don't know, 2000 simulations,
you need 2000 GPUs, which is

704
00:45:28,160 --> 00:45:31,240
very costly.
I don't know how many companies

705
00:45:31,240 --> 00:45:35,920
can, you know, afford 2000 GPUs,
But there are and also, if you

706
00:45:35,920 --> 00:45:38,920
have, you know, few GPUs, well,
you have to wait for many, many

707
00:45:38,920 --> 00:45:43,720
days and the design cycles to be
shorter for for economical

708
00:45:43,720 --> 00:45:46,600
reasons and also for
environmental reasons.

709
00:45:46,600 --> 00:45:51,080
I mean, OK, so for that you
cannot rely on MES alone.

710
00:45:51,080 --> 00:45:55,480
So what you can try to do is to
distill the knowledge of this

711
00:45:55,480 --> 00:45:58,440
high fidelity method into lower
order modes.

712
00:45:59,000 --> 00:46:03,040
So the how can you do it?
That's the way I tried to do it.

713
00:46:03,400 --> 00:46:08,360
One possibility is to distill
this into augmented runs models.

714
00:46:08,480 --> 00:46:12,160
And then once you have augmented
your runs model, you use that

715
00:46:12,160 --> 00:46:15,360
one to perform the optimization
cycle, which is much more

716
00:46:15,360 --> 00:46:20,600
affordable, provided that your
model generalizes well enough,

717
00:46:20,600 --> 00:46:26,120
at least on your design space.
Because if the model fails as

718
00:46:26,120 --> 00:46:28,960
you move a little bit far from
the baseline, well, that's of

719
00:46:28,960 --> 00:46:32,680
course that's useless.
The other point you can do is to

720
00:46:32,880 --> 00:46:35,840
surrogate models.
OK, but the problem with

721
00:46:35,840 --> 00:46:40,520
surrogate model is that you will
never have enough NES to train a

722
00:46:40,520 --> 00:46:44,000
surrogate model on it on NES.
So you are trying to do that

723
00:46:44,040 --> 00:46:45,600
right?
You are trying to produce high

724
00:46:45,600 --> 00:46:50,680
fidelity databases but they are
still relatively limited and I

725
00:46:50,680 --> 00:46:54,040
don't know how many of them will
be possible to produce

726
00:46:54,040 --> 00:46:56,920
especially for very complex very
high Reynolds number flows.

727
00:46:57,720 --> 00:47:01,200
And if we have so many data at
some point, what's the point of

728
00:47:01,200 --> 00:47:05,120
having surrogates if we can run
thousands and 10s of thousands

729
00:47:05,120 --> 00:47:07,560
of LES, what's the point of
having a surrogate?

730
00:47:07,560 --> 00:47:10,800
So I don't know.
So maybe what's what we are

731
00:47:10,800 --> 00:47:14,920
trying to do is to have multi
fidelity models.

732
00:47:15,200 --> 00:47:19,320
So where actually you train the
model using data of different

733
00:47:19,320 --> 00:47:24,080
origins provided that you have a
clear hierarchy and you can

734
00:47:24,080 --> 00:47:30,480
actually try to to train using
plenty of low fidelity data like

735
00:47:30,480 --> 00:47:32,840
runs.
So that's where having a runs,

736
00:47:32,840 --> 00:47:35,520
which is not too bad remains
useful.

737
00:47:36,240 --> 00:47:40,840
And then you try to learn the
gap between the runs and the

738
00:47:41,120 --> 00:47:43,720
high fidelity and then you use
your surrogate.

739
00:47:44,120 --> 00:47:49,560
And this is something which is
not yet so used in the surrogate

740
00:47:49,560 --> 00:47:56,080
modelling community because they
either use plenty runs data or

741
00:47:56,120 --> 00:47:59,720
well, whoever some databases,
high fidelity database like the

742
00:47:59,720 --> 00:48:01,920
one you you have contributed to
produce.

743
00:48:02,160 --> 00:48:07,200
But still in a database you have
what, 300 cases for a family of

744
00:48:07,200 --> 00:48:10,480
cases, But it's not enough to,
you know, to have something

745
00:48:10,480 --> 00:48:13,640
which can be reused for
anything.

746
00:48:14,000 --> 00:48:18,320
So this means that every time
you change, you need to run 300

747
00:48:18,320 --> 00:48:21,560
and yes, or, or one model that
yes or whatever.

748
00:48:21,560 --> 00:48:25,600
And that's costly.
And yeah, So no, I, I.

749
00:48:25,840 --> 00:48:29,480
That's kind of why I think there
is a connection between the two

750
00:48:29,480 --> 00:48:35,720
for sure that the the the
underlying CFD is still the key

751
00:48:36,000 --> 00:48:39,600
and making that more affordable.
Yeah, you're right.

752
00:48:39,600 --> 00:48:42,880
There are start-ups like like
Volcano and others who are

753
00:48:42,880 --> 00:48:45,520
trying to come up with very
computationally efficient codes.

754
00:48:46,080 --> 00:48:51,160
But that is kind of the whole
point of turbans modelling in

755
00:48:51,160 --> 00:48:54,560
some ways is to, is to try to
come up with a way of modelling

756
00:48:54,560 --> 00:48:57,600
it in a, in a, in a, in a clever
way.

757
00:48:57,720 --> 00:49:02,000
So, yeah, if it, if it can be
done, then that would be then

758
00:49:02,000 --> 00:49:05,600
that would be good.
I, I did want to ask you on the,

759
00:49:05,600 --> 00:49:09,080
I mean, you're an editor in
chief of Computers and fluids.

760
00:49:09,080 --> 00:49:13,800
You're extremely active in the,
you know, academic world when it

761
00:49:13,800 --> 00:49:18,040
comes to publishing.
You know, how how are things

762
00:49:18,040 --> 00:49:20,920
changing with the rise of like
ML based papers?

763
00:49:21,320 --> 00:49:24,800
You know, what standards should
they be having?

764
00:49:24,800 --> 00:49:26,920
You know, you would previously
we'd always say, I want to see a

765
00:49:26,920 --> 00:49:28,640
mesh requirement study.
I want to see proof of

766
00:49:28,640 --> 00:49:32,400
convergence.
I want to see, you know what,

767
00:49:33,520 --> 00:49:36,280
what standards are you wanting
and seeing?

768
00:49:36,280 --> 00:49:38,840
And and yeah, I'd be interested
to hear your thoughts as a

769
00:49:38,840 --> 00:49:40,960
journal editor.
Yeah, what we are.

770
00:49:40,960 --> 00:49:46,120
Doing right now in computers and
fluids is to ask to motivate

771
00:49:46,120 --> 00:49:51,320
very well why you do need
machine learning and why this is

772
00:49:51,320 --> 00:49:55,560
improving over, I would say,
standard approaches.

773
00:49:56,680 --> 00:50:00,280
To give an example, many people
do OK, we were talking about

774
00:50:00,280 --> 00:50:04,680
surrogates and they say, OK, I
want to design a new airfoil.

775
00:50:05,120 --> 00:50:10,160
So new airfoil and for that I
generated a database of 10,000

776
00:50:10,160 --> 00:50:14,040
run simulations and then I
trained the surrogate and then

777
00:50:14,240 --> 00:50:21,520
here goes my my, my, my optimal
profile optimized using the

778
00:50:21,520 --> 00:50:24,120
surrogate.
But if you have the

779
00:50:24,160 --> 00:50:29,680
computational power to run
10,000 runs, you can do brute

780
00:50:29,680 --> 00:50:32,880
force optimization.
You don't need actually the

781
00:50:32,880 --> 00:50:36,640
surrogates, right.
So what we are trying to ask is

782
00:50:36,640 --> 00:50:41,080
to show clearly that you are
learning something more than.

783
00:50:41,720 --> 00:50:45,320
So your method is, is bringing
something more either in term of

784
00:50:46,320 --> 00:50:49,240
total computational time,
including the time required to

785
00:50:49,240 --> 00:50:55,560
generate the database and in
terms of improved design or

786
00:50:55,560 --> 00:51:00,600
showing that you, you know, you
are making up for some failure

787
00:51:00,600 --> 00:51:04,720
of the plastical model because
otherwise showing OK, I train

788
00:51:05,120 --> 00:51:08,200
surrogate, it works well.
Here's the optimum.

789
00:51:08,240 --> 00:51:12,200
It's, you know, it's, it's not
I, I believe it, it's, it's,

790
00:51:12,200 --> 00:51:14,920
it's better.
But you know, you are not

791
00:51:15,840 --> 00:51:19,840
telling the community what's new
with respect to things we, we

792
00:51:20,040 --> 00:51:24,080
already are able to do.
For instance, in some cases we

793
00:51:24,080 --> 00:51:28,920
have tested some business on Air
Force, you can generate an Air

794
00:51:28,920 --> 00:51:31,960
Force solution of full sealed
runs in 5 minutes.

795
00:51:31,960 --> 00:51:34,240
You know using free FAM or any
any code.

796
00:51:34,400 --> 00:51:42,800
OK to train some farrogates you
need 20 or 25 or 30 hours of of

797
00:51:42,800 --> 00:51:46,840
GPU time.
So in 20 hours I have largely

798
00:51:46,840 --> 00:51:49,800
finished my optimization with my
old runs code.

799
00:51:50,080 --> 00:51:54,760
So the point is to show that
either you go to more complex

800
00:51:54,760 --> 00:51:56,760
models and you can do that
fastly.

801
00:51:57,120 --> 00:52:00,600
So you have, you know, each run
simulation would take, I don't

802
00:52:00,600 --> 00:52:05,920
know, 100 hours and thanks to
the surrogates, I'm speeding up

803
00:52:05,920 --> 00:52:09,840
and I can have it in, I don't
know, 20 hours or 24 hours or

804
00:52:09,840 --> 00:52:12,360
it's not new.
So what we are asking is that

805
00:52:12,360 --> 00:52:18,080
OK, not only show that your
model has been trained well and

806
00:52:18,080 --> 00:52:22,800
it performs better, but also try
to compare to classical methods,

807
00:52:22,800 --> 00:52:27,280
show what's wrong with the
classical method, what's what

808
00:52:27,280 --> 00:52:30,440
what your new machine learning
method is bringing.

809
00:52:30,880 --> 00:52:34,560
And also show eventually where
your machine learning method is

810
00:52:34,560 --> 00:52:37,720
going to fail because that's
also important, not only

811
00:52:37,720 --> 00:52:41,880
showing, you know, positive
results, but also negative

812
00:52:41,880 --> 00:52:43,480
results.
And that's something the

813
00:52:43,480 --> 00:52:46,360
community doesn't like to do too
much because it's easier to

814
00:52:46,360 --> 00:52:48,320
publish when you have good
results.

815
00:52:48,320 --> 00:52:51,120
You know that's.
That's a very good point

816
00:52:51,120 --> 00:52:53,520
actually.
I, I don't, I think if you look

817
00:52:53,520 --> 00:53:00,960
at like Nurips or those papers
conferences, they mandate like a

818
00:53:00,960 --> 00:53:05,000
limitations section where you
have to put down all the bits

819
00:53:05,000 --> 00:53:08,320
where you didn't do it.
And unless I'm mistaken, that

820
00:53:09,080 --> 00:53:12,280
that precedent of very clearly
pointing out where the

821
00:53:12,280 --> 00:53:15,640
limitation is not something
that's traditionally done right

822
00:53:15,720 --> 00:53:21,680
in not in such a clear before
the conclusions limitations.

823
00:53:21,680 --> 00:53:25,560
And yeah, I've often felt that
there needs to be more

824
00:53:25,560 --> 00:53:28,920
transparency also from just the
CFD side where things are bad,

825
00:53:29,400 --> 00:53:33,800
you know, show me examples of
machine learning, for example,

826
00:53:33,800 --> 00:53:37,800
where the where if I pick a
certain split or, or I pick a

827
00:53:37,800 --> 00:53:40,760
thing, I get really bad results.
It's actually quite important.

828
00:53:40,760 --> 00:53:45,160
Or else the conclusion is that
these models are amazing and

829
00:53:45,160 --> 00:53:50,480
standard CF DS gone rather than
pointing out they're positives

830
00:53:50,480 --> 00:53:54,400
and and where they you failed.
The classic one I guess is in

831
00:53:54,400 --> 00:53:58,760
distribution and out of
distribution where like yeah.

832
00:53:59,840 --> 00:54:02,120
Exactly.
And the other point is to

833
00:54:02,400 --> 00:54:05,040
evaluate the models is as also
discussion.

834
00:54:05,040 --> 00:54:12,120
We, we have had as aware compare
the model on CFD criteria, not

835
00:54:12,120 --> 00:54:16,240
only on machine learning
criteria, because OK, mean

836
00:54:16,240 --> 00:54:18,640
squared errors are the standard
in machine learning.

837
00:54:18,640 --> 00:54:23,160
They are useful, whatever, but
they can be also, you know,

838
00:54:23,800 --> 00:54:27,920
misleading, because in
particular in external

839
00:54:27,920 --> 00:54:32,400
aerodynamics, where most of the
flow is uniform, basically you

840
00:54:32,400 --> 00:54:36,720
are training a neural network to
capture a constant function, and

841
00:54:36,720 --> 00:54:39,800
the regions where something's
going on which are close to the

842
00:54:39,800 --> 00:54:43,840
wall are very tiny.
So if you are wrong in that

843
00:54:43,840 --> 00:54:48,320
region comparatively to the
whole with all the points you

844
00:54:48,320 --> 00:54:52,680
have, maybe it's not enough to
have a, a clear signal on the

845
00:54:52,680 --> 00:54:55,120
MSC.
But you see it if you plot

846
00:54:55,120 --> 00:54:58,640
velocity profiles, if you plot
pressure distribution in

847
00:54:58,640 --> 00:55:02,760
general, they are very bumpy.
If you plot skin friction, skin

848
00:55:02,760 --> 00:55:05,400
friction is terrible because in
that case you are not only

849
00:55:05,400 --> 00:55:08,800
reconstructing the flow field,
the the the velocity field, you

850
00:55:08,800 --> 00:55:11,480
are taking the derivatives.
And it's well known that

851
00:55:11,480 --> 00:55:14,280
approximating the derivative of
a function is more difficult

852
00:55:14,280 --> 00:55:16,440
than approximating the function
itself, right?

853
00:55:17,000 --> 00:55:21,680
And so in the machine learning
community is not used to this

854
00:55:21,680 --> 00:55:24,440
criteria, which are the standard
criteria in CFD.

855
00:55:24,800 --> 00:55:30,520
So again, I I think that if you
pretend to, if you, if you say

856
00:55:30,520 --> 00:55:33,880
that you are going to replace
standard CFD approaches with

857
00:55:33,880 --> 00:55:37,280
machine learning, then you have
to show that machine learning is

858
00:55:37,280 --> 00:55:40,240
able to provide the same thing
that the CFD group is able to

859
00:55:40,240 --> 00:55:42,520
provide and funny.
Enough.

860
00:55:42,520 --> 00:55:46,320
I was having this discussion on
a, on a, on a paper we're

861
00:55:46,320 --> 00:55:51,640
putting out where it's also the
case, isn't it, that in machine

862
00:55:51,640 --> 00:55:58,040
learning, because you have,
let's say, 300 or 100 or 50 test

863
00:55:58,600 --> 00:56:05,120
cases, you'll average over them.
And as your number where that

864
00:56:05,800 --> 00:56:11,920
has the same risk of some are
really bad, some are really

865
00:56:11,920 --> 00:56:13,240
good.
And then you have lots in the

866
00:56:13,240 --> 00:56:15,440
middle.
Or is it you have lots, you

867
00:56:15,440 --> 00:56:18,000
know, how, how bad is bad, how
good is good?

868
00:56:18,000 --> 00:56:23,120
And, and this isn't fully
explained just by a single

869
00:56:23,120 --> 00:56:26,320
number.
So this, as you say, almost

870
00:56:27,000 --> 00:56:33,560
doing a deeper analysis and
showing I, I guess that what

871
00:56:33,560 --> 00:56:38,320
someone described to me is they
to, to the early point of like

872
00:56:38,320 --> 00:56:42,560
hype because of the way the
results are presented today,

873
00:56:43,120 --> 00:56:48,360
where an R-squared value or even
l ^2 gives such low errors, it

874
00:56:48,400 --> 00:56:51,200
looks like they're perfect.
But then when they try and

875
00:56:51,200 --> 00:56:54,480
practice and they don't see as
good, it almost creates A

876
00:56:54,480 --> 00:56:59,080
disappointment, which if the
paper had shown it, they

877
00:56:59,080 --> 00:57:01,000
probably would not be
disappointed because it would

878
00:57:01,000 --> 00:57:03,920
just be meeting the expectation
where yes.

879
00:57:03,920 --> 00:57:07,560
So I would agree with you, part
of that is down to the way that

880
00:57:07,880 --> 00:57:09,840
the results have been presented.
Yeah.

881
00:57:10,480 --> 00:57:13,240
There is quite a lot of
overselling, but that's because

882
00:57:13,240 --> 00:57:16,720
of the economical model also to
continue, you know, and, and

883
00:57:16,720 --> 00:57:20,480
the, and the publication system,
which is a bit too, you know,

884
00:57:21,240 --> 00:57:23,880
especially in, in some
countries, that's not the case

885
00:57:23,880 --> 00:57:26,040
in France, but in some
countries, there's this huge

886
00:57:26,040 --> 00:57:29,080
pressure for publication for
being the 1st for, you know,

887
00:57:29,520 --> 00:57:32,840
this is the store.
And I think this, this, this

888
00:57:33,800 --> 00:57:38,320
rush to publication publishing
more and more in higher impact

889
00:57:38,320 --> 00:57:44,280
in juveniles and so on, in the
end is increasing in the, the

890
00:57:44,280 --> 00:57:50,360
signal to noise ratio at, at
such a point that people don't

891
00:57:50,360 --> 00:57:53,640
even read the papers.
They just read reviews of paper

892
00:57:53,640 --> 00:57:56,240
which are automatically
generated by AIS.

893
00:57:56,520 --> 00:57:59,640
So there is something
fundamentally wrong in this

894
00:57:59,680 --> 00:58:03,240
because, you know, which should
remain at a reasonable level

895
00:58:03,240 --> 00:58:07,560
where humans are talking to
humans because, you know, we're

896
00:58:07,560 --> 00:58:10,160
trying to produce human
knowledge in the end.

897
00:58:10,240 --> 00:58:14,640
Because what's the point if all
the AI is told to each other,

898
00:58:14,920 --> 00:58:17,680
hey, what are we going to do as
you?

899
00:58:18,600 --> 00:58:23,640
I'm going to look at the beach.
What the beach is not really not

900
00:58:23,640 --> 00:58:27,560
bad, but stay on the beach all
the year.

901
00:58:29,040 --> 00:58:30,520
You have to find something else
to do.

902
00:58:31,760 --> 00:58:36,240
So, yeah, no, I think that we
should maybe slow down a little

903
00:58:36,240 --> 00:58:41,680
bit and do what well people used
to do some years ago, which

904
00:58:41,680 --> 00:58:47,600
means publish less, but publish
more thoughtful papers and wait

905
00:58:47,600 --> 00:58:52,240
a little bit before publishing.
And when you publish, publish

906
00:58:52,240 --> 00:58:54,600
something which is more
complete, you have looked to all

907
00:58:54,600 --> 00:59:00,480
the consequences and you know,
thought to all the possible yes.

908
00:59:01,040 --> 00:59:04,480
Do you know what you mean?
This is the, the good and the

909
00:59:04,480 --> 00:59:07,840
bad thing of things like archive
where there's almost a sense of

910
00:59:08,680 --> 00:59:11,680
let's get it out there.
We want people to see it.

911
00:59:11,680 --> 00:59:14,560
It's a preprint.
The preprint supposed to be

912
00:59:14,560 --> 00:59:17,360
taken as well.
You know, this hasn't been

913
00:59:17,360 --> 00:59:21,000
reviewed yet.
But unfortunately people just

914
00:59:21,000 --> 00:59:25,560
take it as the main thing.
And because things are moving so

915
00:59:25,560 --> 00:59:28,920
fast, the, the fact that it
hasn't been reviewed is

916
00:59:28,920 --> 00:59:32,840
sometimes almost ignored, you
know, by, by, by people in a

917
00:59:32,840 --> 00:59:36,640
way.
So it's, I agree with you there,

918
00:59:37,960 --> 00:59:42,160
the pace necessitates sometimes
to put it out, but at the same

919
00:59:42,160 --> 00:59:45,600
time that feeling in your back
at the head of going, well, I

920
00:59:45,600 --> 00:59:48,440
could really do with another
month or two to really deeply

921
00:59:48,440 --> 00:59:51,680
investigate this.
Is.

922
00:59:52,080 --> 00:59:55,600
Balanced against the desire just
to get something out, it's.

923
00:59:56,320 --> 00:59:58,080
It's a tricky 1 and.
On the other.

924
00:59:58,080 --> 01:00:02,560
Hand the editorial The standard
editorial system is being

925
01:00:02,640 --> 01:00:06,000
flooded by papers.
So it's becoming more and more

926
01:00:06,000 --> 01:00:10,000
difficult to keep quality in the
review process.

927
01:00:11,920 --> 01:00:17,160
You know when when you are
flooded by 10s of papers every

928
01:00:17,160 --> 01:00:20,440
day, OK.
And you have to process them and

929
01:00:20,440 --> 01:00:22,520
you have to send them to
reviewers.

930
01:00:22,520 --> 01:00:25,840
You have to find reviewers who
are also flooded by other

931
01:00:25,840 --> 01:00:29,400
journals because there are many
journals, OK, many, many

932
01:00:29,400 --> 01:00:32,960
journals and more journals are,
are being created and so on.

933
01:00:33,440 --> 01:00:39,280
So at some point it's just, you
know, just an escalation and

934
01:00:39,280 --> 01:00:42,000
it's very difficult to control
the quality of the papers and

935
01:00:42,320 --> 01:00:47,120
what in, in free mechanics,
things are becoming difficult,

936
01:00:47,120 --> 01:00:50,640
but they are still under control
to some extent because it's a

937
01:00:50,800 --> 01:00:53,960
small community.
But you have seen the examples

938
01:00:53,960 --> 01:00:58,280
of these big conferences like
Europe's where they have 10s of

939
01:00:58,280 --> 01:01:03,120
thousands of papers, where most
of the papers are written by AI

940
01:01:03,120 --> 01:01:08,920
and reviewed by AI.
And so what's, what's the

941
01:01:08,920 --> 01:01:11,840
meaning of the review process in
that case?

942
01:01:11,840 --> 01:01:15,240
You don't review it at all.
You just put it for free on the

943
01:01:15,240 --> 01:01:17,840
website and who wants to read
it?

944
01:01:18,840 --> 01:01:22,880
So I think that the, the, the
rush to having more and more and

945
01:01:22,880 --> 01:01:27,640
more papers, which is also due
to the, you know, to the

946
01:01:27,640 --> 01:01:31,480
publishers is in some sense
killing system.

947
01:01:31,480 --> 01:01:36,600
Because if the review level
falls below a given pressure, at

948
01:01:36,600 --> 01:01:39,920
some point, there is no point in
going through a standard review

949
01:01:39,920 --> 01:01:43,680
process because there is no
added values in going through

950
01:01:43,680 --> 01:01:47,440
that review.
So you just produce your your

951
01:01:47,440 --> 01:01:50,800
paper, you put it somewhere and
the people look at it if they

952
01:01:50,800 --> 01:01:54,400
want and that's it.
And if you are in the hype, you

953
01:01:54,400 --> 01:01:58,080
are in, let's say, famous
scheme, very visible and so on,

954
01:01:58,080 --> 01:02:01,960
people will have followers who
will read your paper because

955
01:02:01,960 --> 01:02:04,520
it's you.
So we are lifting from

956
01:02:04,520 --> 01:02:08,320
scientists to influencers.
And that's a bit, you know,

957
01:02:08,320 --> 01:02:10,120
scary for me.
Yeah.

958
01:02:10,560 --> 01:02:15,760
Yeah, no, no the.
So how does this, you know,

959
01:02:15,760 --> 01:02:19,440
translate and what it what's
what are your trying to achieve

960
01:02:19,440 --> 01:02:23,560
with with, you know, the new
position that Sobhan the AI

961
01:02:23,560 --> 01:02:24,840
centre?
Could you maybe tell a little

962
01:02:24,840 --> 01:02:28,920
bit more about your place and
and kind of your vision to how

963
01:02:28,920 --> 01:02:33,200
to do, I guess, AI for science
and AI for engineering and and

964
01:02:33,200 --> 01:02:37,080
and maybe take some of those
rigorous things that you've

965
01:02:37,640 --> 01:02:41,320
taken from the fluid mechanics
world in into this side?

966
01:02:41,320 --> 01:02:43,320
That'd be.
I love your shiny new Offit So I

967
01:02:43,320 --> 01:02:46,760
wanted to hear more.
Yeah.

968
01:02:47,280 --> 01:02:50,440
Well, actually, well, in
Sorbonne we have this centre,

969
01:02:50,440 --> 01:02:53,080
Sky Sorbonne.
The beginning, the name was

970
01:02:53,080 --> 01:02:56,320
Sorbonne Centre for Artificial
Intelligence, which was created

971
01:02:56,320 --> 01:03:04,360
seven years ago at the beginning
of the AI hype, because we had

972
01:03:04,360 --> 01:03:08,160
the strong teams in computer
science and also in applied

973
01:03:08,160 --> 01:03:11,160
mathematics working on that.
So basically at the beginning,

974
01:03:11,240 --> 01:03:17,040
Sky has been created by applied
mathematicians and and computer

975
01:03:17,040 --> 01:03:21,480
scientists alongside with some
people from humanities and from

976
01:03:21,480 --> 01:03:24,400
other sciences, for instance of
computational biology, this kind

977
01:03:24,400 --> 01:03:28,720
of thing.
And well, it has helped a lot in

978
01:03:28,760 --> 01:03:33,040
in making.
So in making AI, it's a one

979
01:03:33,040 --> 01:03:36,560
visible.
They have generated a lot of

980
01:03:36,960 --> 01:03:39,760
activities we have.
So at at some point we have been

981
01:03:39,760 --> 01:03:42,400
labeled by the French government
as clusters.

982
01:03:42,400 --> 01:03:45,320
So now the sea in sky stands for
cluster.

983
01:03:47,040 --> 01:03:50,360
And these clusters are sort of
centres of excellence in France.

984
01:03:50,360 --> 01:03:55,680
There are nine of them which are
supposed to foster research in

985
01:03:55,680 --> 01:04:01,680
AI training and transfer to to
companies.

986
01:04:02,520 --> 01:04:07,080
And well, every cluster has its
own specificities.

987
01:04:07,480 --> 01:04:11,680
But when?
Well, based on my own experience

988
01:04:11,680 --> 01:04:15,520
and based on the very large
scientific community at

989
01:04:15,520 --> 01:04:20,520
Sorbonne, I said, well, right
now everybody is looking at AI

990
01:04:20,520 --> 01:04:25,720
for science because science is
actually going to be deeply

991
01:04:25,720 --> 01:04:27,400
modified.
The way of doing science is

992
01:04:27,400 --> 01:04:31,680
going to be deeply modified by
the introduction of these AI

993
01:04:32,600 --> 01:04:36,240
approaches, which is not the
machine learning approaches for,

994
01:04:36,240 --> 01:04:39,680
you know, improving predictions
or speeding up predictions, but

995
01:04:39,680 --> 01:04:48,000
also reasoning tools and, you
know, models for, you know, for,

996
01:04:48,280 --> 01:04:53,800
for automatizing lab,
automatizing labs, AI agents

997
01:04:53,800 --> 01:04:56,800
and, and all that.
And I said, well, the, the

998
01:04:56,800 --> 01:05:00,640
difficulty with that is that
when you apply this to science,

999
01:05:00,640 --> 01:05:03,240
you have several promos.
First of all, you are not

1000
01:05:03,240 --> 01:05:08,320
dealing with random data.
It's not images, it's not video.

1001
01:05:08,400 --> 01:05:14,240
It's something which has a deep
physical meaning, which he's

1002
01:05:14,280 --> 01:05:16,520
known only to experts from the
discipline.

1003
01:05:16,520 --> 01:05:20,720
So you can train a surrogate or
whatever, or a foundation model

1004
01:05:20,720 --> 01:05:22,280
using, I don't know, chemistry
data.

1005
01:05:22,480 --> 01:05:26,240
But if you don't have a
chemistry by you to explain, if

1006
01:05:26,240 --> 01:05:30,840
what's coming out of the model
is you know, makes sense from a

1007
01:05:30,840 --> 01:05:34,560
chemical point of view or is
just an hallucination, you don't

1008
01:05:34,560 --> 01:05:37,320
know.
So that's where it's extremely

1009
01:05:37,320 --> 01:05:41,360
important to connect people from
computer science and the

1010
01:05:41,360 --> 01:05:44,320
disciplines and also people from
mathematics because they can

1011
01:05:44,320 --> 01:05:46,800
help with the mathematical
foundations.

1012
01:05:47,080 --> 01:05:50,520
What I dream about of is the
equivalence, you know, LACS

1013
01:05:50,520 --> 01:05:52,760
equivalence theorem for machine
learning.

1014
01:05:52,760 --> 01:05:55,520
That would be the growl, you
know, something that ensures

1015
01:05:55,520 --> 01:05:58,840
that you, if you put more data
in your model, at some point you

1016
01:05:58,840 --> 01:06:00,480
are going to converge to
something.

1017
01:06:00,480 --> 01:06:03,160
But you know, that would be
great.

1018
01:06:03,160 --> 01:06:06,360
But so the idea would be to have
all these people discussing

1019
01:06:06,360 --> 01:06:08,200
together.
And then I realized that some

1020
01:06:08,200 --> 01:06:11,560
problems you can have in one
discipline, for instance, in

1021
01:06:11,560 --> 01:06:14,080
high energy physics, you have
big data.

1022
01:06:14,280 --> 01:06:19,960
This data are structured, they
encode causalities because the

1023
01:06:19,960 --> 01:06:24,240
particle I'm capturing here at
this moment comes from a jet

1024
01:06:24,240 --> 01:06:28,640
which has originated there and
the particle have undergone some

1025
01:06:28,640 --> 01:06:31,560
decay process and so on.
And this, they call this

1026
01:06:31,600 --> 01:06:34,520
actually jets of particles.
And they say, oh, this is the

1027
01:06:34,520 --> 01:06:37,120
same you can have in a turbulent
jet because the jet is

1028
01:06:37,200 --> 01:06:42,040
originating from, you know, as
lice or something as law.

1029
01:06:42,440 --> 01:06:46,520
And then you have causalities of
turbulent, of lamina structures

1030
01:06:46,520 --> 01:06:48,480
which eventually become
turbulent and then they are

1031
01:06:48,480 --> 01:06:51,680
modified, they are transported.
So we have the same problems and

1032
01:06:51,680 --> 01:06:54,760
we have to encode these
causalities because trying to

1033
01:06:54,760 --> 01:06:58,560
learn all the physics, all the
chemistry, all the biology from

1034
01:06:58,560 --> 01:07:02,640
scratch, using brute force
training on data, it's not

1035
01:07:02,640 --> 01:07:04,760
efficient because we know the
physics.

1036
01:07:04,960 --> 01:07:06,800
If we know the physics, let's
use it.

1037
01:07:07,160 --> 01:07:10,800
But the problem is then how do
you really encode that physics?

1038
01:07:10,800 --> 01:07:12,200
First of all, you need to know
it.

1039
01:07:12,480 --> 01:07:16,280
So it's not the computer science
guy who is going to know which

1040
01:07:16,280 --> 01:07:20,520
are the constraints which are
the most suitable for high

1041
01:07:20,520 --> 01:07:23,920
energy jets, right?
So we need the communities to

1042
01:07:23,920 --> 01:07:26,960
work together and also some
problems which have been solved

1043
01:07:26,960 --> 01:07:29,240
about causalities.
I don't know in high energy in

1044
01:07:29,240 --> 01:07:32,960
the energy physics community can
be useful for free mechanics and

1045
01:07:32,960 --> 01:07:36,800
cover city.
For instance, I have talked with

1046
01:07:36,960 --> 01:07:41,520
people from plasma physics.
So in some cases they have

1047
01:07:41,520 --> 01:07:43,560
Navier Stokes plus the magnetic
fields.

1048
01:07:43,560 --> 01:07:46,880
They have problems of closures
because they cannot resolve all

1049
01:07:46,880 --> 01:07:49,840
the scales they need to cause
brain models.

1050
01:07:49,840 --> 01:07:54,760
So they have closure problems,
and maybe something of what we

1051
01:07:54,760 --> 01:07:57,920
are doing with turbulence
modeling can be reduced in that

1052
01:07:57,920 --> 01:07:59,400
field.
I'm collaborating with

1053
01:07:59,400 --> 01:08:02,280
volcanologists because in
volcanoes you need to

1054
01:08:02,280 --> 01:08:05,960
characterize the biology of
magmas and you cannot measure it

1055
01:08:06,040 --> 01:08:10,320
and you have simplified models
which are not accurate enough.

1056
01:08:10,640 --> 01:08:14,800
So you can use the techniques we
use for turbulence modelling to

1057
01:08:14,800 --> 01:08:18,399
improve the rheology of magmas.
So that's where it becomes

1058
01:08:18,399 --> 01:08:23,240
exciting to share knowledge
across the, the different

1059
01:08:23,240 --> 01:08:27,880
scientific communities instead
of having every guy in its

1060
01:08:27,880 --> 01:08:32,520
isolated department reinvent
everything from scratch because

1061
01:08:32,520 --> 01:08:35,640
we are reinventing the wheel
otherwise, you know, and that's

1062
01:08:35,920 --> 01:08:39,200
how we are trying to, you know,
I'm trying to, to, to, to

1063
01:08:39,640 --> 01:08:43,319
animate this community.
And we have a seminar now called

1064
01:08:43,319 --> 01:08:47,520
the EI for Science.
And hopefully we are trying to

1065
01:08:47,520 --> 01:08:51,840
apply to funding opportunities,
even if they are not funding,

1066
01:08:52,520 --> 01:08:54,960
just the fact that we are around
the table and we are thinking

1067
01:08:54,960 --> 01:08:57,880
about what we could do together.
It's already something which

1068
01:08:57,880 --> 01:09:01,520
helps building the community.
And then what people start to

1069
01:09:01,520 --> 01:09:04,840
collaborate and I think it's
beneficial to everybody.

1070
01:09:04,840 --> 01:09:07,479
That's, that's my hope at least.
No, it sounds.

1071
01:09:07,479 --> 01:09:09,399
Fantastic.
And I totally agree there's lots

1072
01:09:09,399 --> 01:09:13,439
of I like the ideas that these
multidisciplinary centres,

1073
01:09:13,439 --> 01:09:16,800
particularly in the age of AII
think it makes complete sense.

1074
01:09:16,800 --> 01:09:20,920
And as you say as well helping
be a bridge to I guess the

1075
01:09:20,920 --> 01:09:23,640
industry and startups and and
and things like that.

1076
01:09:23,640 --> 01:09:28,120
That makes makes complete sense.
I guess maybe to, as we come to

1077
01:09:28,120 --> 01:09:33,000
the end, all of these new
things, all of these topics that

1078
01:09:33,000 --> 01:09:36,200
have come up since, since I
guess you did your original

1079
01:09:36,200 --> 01:09:41,439
studies, you know, how does this
affect the education system?

1080
01:09:41,680 --> 01:09:46,399
You know, what should young
fluid mechanics researchers be

1081
01:09:47,720 --> 01:09:50,920
be picking today as their PhD
topics or the undergraduate what

1082
01:09:50,920 --> 01:09:57,560
what will prepare them for the
next 10 to 20 years of of the

1083
01:09:57,560 --> 01:09:59,360
world we live in?
Yeah, I.

1084
01:09:59,360 --> 01:10:03,640
Think we are in a very critical
period for education because

1085
01:10:03,640 --> 01:10:08,040
well you and me have been
educated in a few without AI.

1086
01:10:08,240 --> 01:10:14,880
So we have been taught how to to
solve problems without the help

1087
01:10:14,880 --> 01:10:21,880
of AI to how to like to
interpret critically the results

1088
01:10:22,320 --> 01:10:27,680
and how to look for information
and, you know, how to, to check

1089
01:10:27,680 --> 01:10:30,200
if what we are doing is good or
not and so on.

1090
01:10:30,560 --> 01:10:33,520
The problem is that the new
generation is AI native.

1091
01:10:33,760 --> 01:10:41,440
So I, I was hearing, yeah, yeah.
Because yesterday at the radio

1092
01:10:41,440 --> 01:10:46,320
were saying that many teenagers
now who are going to vote for

1093
01:10:46,320 --> 01:10:50,400
the first time are asking AIS
what should they vote?

1094
01:10:50,800 --> 01:10:55,520
Because trust AI is more than
politicians or more than the

1095
01:10:55,520 --> 01:10:57,200
classical media.
OK.

1096
01:10:57,600 --> 01:11:00,440
And they're doing a little bit
the same with the with the

1097
01:11:00,440 --> 01:11:04,520
school, with school teachers and
with university professor who

1098
01:11:04,520 --> 01:11:07,920
are not prepared to that
because, you know, the students

1099
01:11:07,920 --> 01:11:11,680
have smarter than we are and
they know how to, you know, to

1100
01:11:11,680 --> 01:11:14,920
use AI to solve their trade
mechanics problems, to solve

1101
01:11:14,920 --> 01:11:19,160
their thermodynamics problems.
And you have still teachers who

1102
01:11:20,160 --> 01:11:24,080
six months ago, we're not even
imagining that the problem had

1103
01:11:24,080 --> 01:11:27,360
been solved by an AI.
So right now, many, many

1104
01:11:27,360 --> 01:11:31,160
colleagues are scared about the
fact that students are cheating

1105
01:11:31,200 --> 01:11:33,680
with AI.
But the for me, the problem is

1106
01:11:33,680 --> 01:11:36,640
not only cheating.
The fact is that some students

1107
01:11:36,640 --> 01:11:42,040
are basically, you know,
delegating the task of thinking

1108
01:11:42,040 --> 01:11:44,360
to AI.
That's more much more dangerous.

1109
01:11:44,720 --> 01:11:49,760
So what I think is that we have
to rethink the whole system and

1110
01:11:49,760 --> 01:11:54,000
many tasks which were, you know,
I teach you how to solve a

1111
01:11:54,000 --> 01:11:59,000
second order linear ordinary
differential equations and you

1112
01:11:59,000 --> 01:12:01,000
just have the technique to solve
that.

1113
01:12:01,000 --> 01:12:05,080
That's no longer useful because
you know, ChatGPT knows to do

1114
01:12:05,080 --> 01:12:08,280
that very well.
The problem, we should teach the

1115
01:12:08,280 --> 01:12:12,800
students how to analyze the
results and how to understand if

1116
01:12:12,800 --> 01:12:18,880
the result is correct, if it's
an hallucination, how to

1117
01:12:18,880 --> 01:12:22,000
formulate the questions also.
Because if you don't know how to

1118
01:12:22,000 --> 01:12:27,720
ask AI correctly, what to do, AI
is going to do something you

1119
01:12:27,720 --> 01:12:31,560
know you don't control.
And if you are not cultivated

1120
01:12:31,560 --> 01:12:35,920
enough to understand the answer,
you are just going to accept

1121
01:12:35,920 --> 01:12:41,600
whatever comes out without,
without, you know, without

1122
01:12:41,680 --> 01:12:43,760
criticizing what what's coming
out.

1123
01:12:44,160 --> 01:12:46,560
So I think that we have to move
the shift.

1124
01:12:46,560 --> 01:12:50,480
So instead of teaching
techniques and, you know,

1125
01:12:50,480 --> 01:12:57,080
techniques or, or just notions,
we should teach critical

1126
01:12:57,080 --> 01:12:59,560
thinking.
And that's harder because it,

1127
01:13:00,120 --> 01:13:05,320
it, it, it implies a, a
considerable modification of our

1128
01:13:05,320 --> 01:13:07,480
teaching problems and new
teaching approaches.

1129
01:13:07,760 --> 01:13:10,920
It's not easy to do, especially
in large universities as we

1130
01:13:10,920 --> 01:13:16,560
have, I, we have courses with,
you know, 5500 students.

1131
01:13:16,840 --> 01:13:21,760
It's not easy to, you know, to
have to, to change all the

1132
01:13:21,760 --> 01:13:24,560
teaching procedure.
But it's something we need to do

1133
01:13:25,120 --> 01:13:29,680
because without critical
thinking, basically humans are

1134
01:13:29,680 --> 01:13:35,920
going to lose competence.
And yeah, and, and, and, and,

1135
01:13:36,280 --> 01:13:39,920
and in case, in case AI is not
working, what do we do?

1136
01:13:40,240 --> 01:13:44,520
And I had a very nice metaphor
from a person from the European

1137
01:13:44,520 --> 01:13:50,800
committee who said, well, when
you, when you are, when are you

1138
01:13:50,800 --> 01:13:58,040
are a pilot of an aircraft, you
are 9, I say 8000 or 99900.

1139
01:13:58,040 --> 01:14:01,840
Ninety, 100% of the time you are
going to use the automatic pipe.

1140
01:14:02,200 --> 01:14:07,080
But in case of accident, the
pilot needs to know how to drive

1141
01:14:07,320 --> 01:14:12,640
the plane down to safety, right?
And that's exactly that.

1142
01:14:12,640 --> 01:14:17,720
So we, we need to be critical
because we don't know where the

1143
01:14:17,720 --> 01:14:20,680
AI is going to hallucinate even
we are.

1144
01:14:20,720 --> 01:14:25,040
Also, we need to, to understand
if, for instance, sometimes AI

1145
01:14:25,040 --> 01:14:28,960
is able to bring us notions from
other disciplines which are not

1146
01:14:28,960 --> 01:14:32,160
our discipline.
But if we have methodology, we

1147
01:14:32,160 --> 01:14:38,080
know how to analyze if what the,
the, the, the LMM is saying is

1148
01:14:38,080 --> 01:14:41,680
reasonable or not.
We know how to verify facts and

1149
01:14:41,680 --> 01:14:45,680
so on.
And we have culture, general

1150
01:14:45,680 --> 01:14:50,680
culture and language capacity to
understand then we can control

1151
01:14:50,680 --> 01:14:54,680
and we can be enriched as
professionals, as scientists, as

1152
01:14:54,680 --> 01:14:59,160
humans and whatever instead of
being just, you know, replaced

1153
01:14:59,520 --> 01:15:01,920
like in Matrix.
That's my, my nightmare.

1154
01:15:01,920 --> 01:15:05,480
You know, Matrix humans are just
there to be pumped energy.

1155
01:15:08,960 --> 01:15:09,480
Yeah.
No.

1156
01:15:09,480 --> 01:15:14,040
It does.
It does seem that it's not

1157
01:15:14,040 --> 01:15:16,600
universally accepted.
How best to do this from an

1158
01:15:16,600 --> 01:15:19,960
education point of view?
I've heard some go the other way

1159
01:15:19,960 --> 01:15:24,000
and be almost well for an
undergraduate, we will do lots

1160
01:15:24,000 --> 01:15:29,720
of in person examinations and
basically no homework anymore

1161
01:15:29,760 --> 01:15:34,080
because you this is no way of
knowing the homework.

1162
01:15:35,480 --> 01:15:39,680
And, and as you say, there's the
other side, which is maybe a

1163
01:15:39,680 --> 01:15:45,760
little bit more ex, you know, do
a presentation to explain how

1164
01:15:45,760 --> 01:15:49,000
you got there.
I, I kind of lean a little bit

1165
01:15:49,000 --> 01:15:52,400
more to the first one at the
right stage of your career.

1166
01:15:52,840 --> 01:15:56,160
I guess this is more like even a
secondary school, let's say, you

1167
01:15:56,160 --> 01:15:59,960
know, before university, you
know, the certain things where I

1168
01:15:59,960 --> 01:16:03,880
just feel you have to know it
because as you say, what if

1169
01:16:03,880 --> 01:16:06,560
there is a moment where you
don't, you're on a, you know,

1170
01:16:06,560 --> 01:16:08,280
you don't have access to the AI
tools.

1171
01:16:09,480 --> 01:16:13,200
I feel there's like a transition
point when you can expect.

1172
01:16:14,040 --> 01:16:15,680
Like if you're doing a P.
HD.

1173
01:16:17,000 --> 01:16:18,400
Or certainly a.
Postdoc.

1174
01:16:18,480 --> 01:16:22,400
I actually think AI is a
fantastic help and will make you

1175
01:16:22,400 --> 01:16:26,640
more productive, make you be
able to do more things, help you

1176
01:16:26,640 --> 01:16:28,960
achieve what you want to do
faster and better, and

1177
01:16:28,960 --> 01:16:32,560
ultimately get a, you know what
would have taken you a month to

1178
01:16:32,560 --> 01:16:36,440
write some coding, which wasn't
your main task, it was just you

1179
01:16:36,440 --> 01:16:39,480
had to do it to get there.
AI will accelerate it.

1180
01:16:39,480 --> 01:16:43,240
But if you're at the
undergraduate level or or still

1181
01:16:43,240 --> 01:16:48,240
at school or something, I feel
that's probably where AI should

1182
01:16:48,240 --> 01:16:53,320
only be used to help you learn
something, but not it shouldn't

1183
01:16:53,320 --> 01:16:57,120
be used like a calculator almost
or or or you will.

1184
01:16:57,120 --> 01:16:58,760
Just not.
Understand the fundamental

1185
01:16:59,320 --> 01:17:04,120
theory, it can be used.
As a mate you know because you

1186
01:17:04,120 --> 01:17:11,560
can no say tell the EI ask me
questions and give the answer

1187
01:17:11,560 --> 01:17:14,000
and then the EI can analyze the
tutor.

1188
01:17:14,240 --> 01:17:18,000
Yes, and it.
Makes, you know, learning more

1189
01:17:18,000 --> 01:17:20,200
interactive.
In my research, sometimes I feel

1190
01:17:20,200 --> 01:17:23,200
alone and I don't have a
colleague to discuss about the

1191
01:17:23,200 --> 01:17:25,440
point.
So I take an AI and say, what do

1192
01:17:25,440 --> 01:17:28,480
you think about this idea?
And so we start the conversation

1193
01:17:28,480 --> 01:17:31,080
and that's useful.
And then, you know, the AI is

1194
01:17:31,080 --> 01:17:33,120
bringing some ideas.
So, yeah, yeah.

1195
01:17:33,120 --> 01:17:35,000
Or.
Explaining new topic.

1196
01:17:35,000 --> 01:17:38,040
I often do that when I'm trying
to and you can ask it.

1197
01:17:38,040 --> 01:17:40,440
And the great thing I find is
that it has these different

1198
01:17:40,440 --> 01:17:44,480
levels where you could say, OK,
sometimes I'll ask like, can you

1199
01:17:44,480 --> 01:17:48,000
show me that in a code like how
would you actually code this up?

1200
01:17:48,320 --> 01:17:52,840
Which would be hard for a tutor
to do instantly like a person to

1201
01:17:52,920 --> 01:17:57,120
be that flexible to move from
reasoning.

1202
01:17:57,120 --> 01:18:00,840
To code and back and so on.
Because you are, yeah, easily

1203
01:18:00,840 --> 01:18:05,000
lost in, you know, coding
details and so on That that

1204
01:18:05,080 --> 01:18:07,200
yeah, that's that's actually
great.

1205
01:18:07,840 --> 01:18:11,840
But the problem is to explain to
the young generations that

1206
01:18:12,040 --> 01:18:16,440
that's the way probably to use
the AI more like a companion

1207
01:18:16,720 --> 01:18:22,880
then like an entity that is
going to do the work in my stead

1208
01:18:23,080 --> 01:18:28,320
while I'm, I don't know, playing
video games or something.

1209
01:18:28,320 --> 01:18:32,720
That's that's not the point.
And, and they need so the

1210
01:18:32,720 --> 01:18:37,320
problem is to explain them what
are the risks and why they are

1211
01:18:37,560 --> 01:18:41,680
missing something if they use AI
only for that.

1212
01:18:42,000 --> 01:18:44,080
So I can understand that
sometimes you need to

1213
01:18:44,080 --> 01:18:46,960
accelerate, you know, you have
to give a report or you are a

1214
01:18:46,960 --> 01:18:49,440
little bit in a hurry.
But if you do that

1215
01:18:49,440 --> 01:18:52,720
systematically, then basically
you are missing your education

1216
01:18:52,720 --> 01:18:56,040
because it's not you who is
going to be educated, but the AI

1217
01:18:56,040 --> 01:18:59,760
and and I and I think.
In some ways it'll end up

1218
01:18:59,760 --> 01:19:04,960
creating a bit of a so A2 tier
system where in some ways people

1219
01:19:04,960 --> 01:19:10,560
will become it's quite capable
of doing some jobs where it

1220
01:19:10,560 --> 01:19:16,000
requires them just to be, you
know, a functional user of these

1221
01:19:16,000 --> 01:19:20,520
tools.
But it will stop you from maybe

1222
01:19:20,520 --> 01:19:23,040
being the the inventor or the
innovator.

1223
01:19:23,160 --> 01:19:26,880
And there it there feels like
you could, to be fair, you could

1224
01:19:26,880 --> 01:19:30,760
probably still do quite good at
a job if you because these AI

1225
01:19:30,760 --> 01:19:34,240
tools are so good.
But I feel there must be some

1226
01:19:34,240 --> 01:19:38,200
points where that lack of deep
understanding will come and bite

1227
01:19:38,200 --> 01:19:44,320
you.
And so, yeah, it's it's a

1228
01:19:44,320 --> 01:19:45,240
challenge.
Yeah.

1229
01:19:45,640 --> 01:19:49,840
We're not in very.
Specialistic fields as as as so

1230
01:19:49,840 --> 01:19:51,560
we are in very specialistic
fields.

1231
01:19:52,000 --> 01:19:56,280
So I don't think that the EI has
been trained enough.

1232
01:19:56,400 --> 01:20:01,280
So that's where you know, you
bring the new ideas, maybe the

1233
01:20:01,280 --> 01:20:06,560
AI strengthens you, accelerates
you with the coding or allows

1234
01:20:06,560 --> 01:20:10,160
you to make connections more
quickly and so on.

1235
01:20:10,160 --> 01:20:12,920
But it's still new.
We're bringing the ideas.

1236
01:20:13,320 --> 01:20:17,480
And if you're you know although
that.

1237
01:20:17,840 --> 01:20:25,360
I do suspect will also be, you
know, succeeded by AI because

1238
01:20:25,360 --> 01:20:30,960
the, you know, the, the AI
scientist or the AI engineer.

1239
01:20:31,920 --> 01:20:35,480
I have to say, I, I, I still
suspect actually that it will be

1240
01:20:35,480 --> 01:20:40,240
able to come up with as many
ideas as as we, we will be able

1241
01:20:40,240 --> 01:20:42,080
to do.
Maybe not the, you know,

1242
01:20:42,720 --> 01:20:48,960
Einstein new completely
breakthrough thing, but it's

1243
01:20:48,960 --> 01:20:51,320
it's certainly going to be an
interesting test.

1244
01:20:52,400 --> 01:20:54,840
Of given.
That the whole, like your centre

1245
01:20:54,840 --> 01:20:58,840
is a good example of it.
I, I do get the sense that AI

1246
01:20:58,840 --> 01:21:02,000
for science or AI for
engineering is now becoming the

1247
01:21:02,000 --> 01:21:04,840
frontier.
It wasn't five years ago.

1248
01:21:05,280 --> 01:21:07,120
It was sort of seen as a niche
area.

1249
01:21:07,840 --> 01:21:12,240
I, I get the sense now that the,
the world of sort of physical AI

1250
01:21:12,240 --> 01:21:15,680
and AI for engineering and AI
for science is becoming a hot

1251
01:21:15,680 --> 01:21:18,240
topic.
And whenever it's a hot topic

1252
01:21:18,320 --> 01:21:21,200
and there's lots of investment
and money put into it, it, you

1253
01:21:21,200 --> 01:21:24,320
know, yeah.
There is, yeah, some

1254
01:21:24,320 --> 01:21:29,560
overselling, yeah.
But but yes, I guess, I guess

1255
01:21:29,560 --> 01:21:33,200
that's the front here.
And I, well, I want to believe

1256
01:21:33,200 --> 01:21:35,200
that.
Well, even if the guys are

1257
01:21:35,200 --> 01:21:39,280
becoming more and more powerful
and potentially they could even

1258
01:21:39,360 --> 01:21:43,720
won a Nobel Prize at some point
if you think about alpha fold

1259
01:21:43,720 --> 01:21:46,880
for instance.
It's well.

1260
01:21:48,000 --> 01:21:51,680
What's new is the fact that in a
few years they could discover

1261
01:21:51,680 --> 01:21:55,400
thousands and hundreds of
thousands of proteins which are

1262
01:21:55,400 --> 01:21:59,080
much more than the 2000 and
something proteins they had

1263
01:21:59,080 --> 01:22:01,520
discovered during the past 50
years.

1264
01:22:02,280 --> 01:22:08,920
So every protein structure was
PhD thesis and now you can

1265
01:22:08,920 --> 01:22:12,600
generate plenty of them, you
know, just with with with these

1266
01:22:12,720 --> 01:22:17,000
AR.
So what was new was the fact of

1267
01:22:17,000 --> 01:22:19,440
being able to accelerate so
much.

1268
01:22:19,440 --> 01:22:23,440
But still without the knowledge
of the proteins from the past,

1269
01:22:23,680 --> 01:22:25,800
it wouldn't have been possible
to have that.

1270
01:22:26,000 --> 01:22:29,600
And the same for climate.
If now climate models are so

1271
01:22:29,600 --> 01:22:34,880
good, that's because for 10s of
years people have developed

1272
01:22:34,880 --> 01:22:38,960
better and better climate model,
the classical ones, and they

1273
01:22:39,040 --> 01:22:43,840
have done data simulation and
then they have done all the

1274
01:22:43,840 --> 01:22:46,880
reanalysis from the beginning.
And with that they have

1275
01:22:46,880 --> 01:22:50,480
generated the huge databases
which can be used for by the

1276
01:22:50,480 --> 01:22:53,000
models.
And also, well, everything is

1277
01:22:53,000 --> 01:22:55,320
like that.
That's because we have huge

1278
01:22:55,640 --> 01:23:00,120
amounts of human knowledge
accumulated over years, which

1279
01:23:00,120 --> 01:23:05,480
can be injected in these models.
But what once all the knowledge

1280
01:23:05,480 --> 01:23:07,880
will be generated by by these
things?

1281
01:23:08,160 --> 01:23:11,080
Probably, I don't know, Maybe at
some point they will become so

1282
01:23:11,080 --> 01:23:14,840
intelligent with these emergent
phenomena, they will be able to

1283
01:23:14,840 --> 01:23:18,280
create something new that
happens in very complex systems.

1284
01:23:19,200 --> 01:23:24,360
But I, I, I I I'm.
But I, I hope humans has still a

1285
01:23:24,360 --> 01:23:26,160
role to play.
Oh yes.

1286
01:23:27,040 --> 01:23:29,840
And that.
Well, and that instead of being

1287
01:23:29,840 --> 01:23:35,000
replaced by these AIS, well, one
of the person who participates,

1288
01:23:35,000 --> 01:23:38,480
a Jesse Taylor, he's a, he's a
physicist from MIT and he

1289
01:23:38,480 --> 01:23:41,640
participated to our AI for
science launch play.

1290
01:23:41,920 --> 01:23:45,560
And he came out with his
metaphor of the Centaur

1291
01:23:45,680 --> 01:23:48,720
scientists.
And I found that as a beautiful

1292
01:23:48,720 --> 01:23:53,960
metaphor because as you know, if
you are two men and not enough

1293
01:23:53,960 --> 01:23:57,360
horse, you are too slow.
If you are two horse and not

1294
01:23:57,360 --> 01:24:02,160
enough man, you cannot reason.
But if you have 1/2 man half

1295
01:24:02,160 --> 01:24:05,360
horse, you can go as fast as a
horse while having all, you

1296
01:24:05,360 --> 01:24:08,840
know, the creativity and and
width of a human.

1297
01:24:09,160 --> 01:24:13,800
And so, well, I, I adopted this
idea and say that, well, our

1298
01:24:13,880 --> 01:24:17,760
role would be to generate to,
to, to train the next generation

1299
01:24:17,760 --> 01:24:22,000
of Centaur scientists and
Centaur professionals and not

1300
01:24:22,000 --> 01:24:26,320
just horse.
I like that.

1301
01:24:26,640 --> 01:24:29,480
You know, and that's, that's
probably a great way to end this

1302
01:24:29,480 --> 01:24:32,600
in the, in the sense of I agree
with you that it's, it's about

1303
01:24:33,040 --> 01:24:36,320
the coupling of, of human and
AI.

1304
01:24:36,320 --> 01:24:41,920
It's about being progressive and
optimistic about AI, but not

1305
01:24:41,920 --> 01:24:45,800
forgetting the, the sort of
human role role in it.

1306
01:24:45,800 --> 01:24:50,480
And, and I would argue that
today the human specialist is

1307
01:24:50,480 --> 01:24:54,760
needed even more because, you
know, to, to help develop these

1308
01:24:54,760 --> 01:24:57,080
AI models and make them accurate
and point them in the right

1309
01:24:57,080 --> 01:25:00,080
direction.
But maybe we need to talk again

1310
01:25:00,080 --> 01:25:04,320
in two years time and, and see
where things have gone.

1311
01:25:04,320 --> 01:25:08,040
It'd be interesting to to listen
back and see as it shot off like

1312
01:25:08,040 --> 01:25:10,160
this is, is it the same?
Is it less?

1313
01:25:10,200 --> 01:25:13,520
That will be an interesting, but
I really enjoyed this discussion

1314
01:25:13,520 --> 01:25:16,960
with you and I personally always
love working with you on on

1315
01:25:16,960 --> 01:25:20,680
various committees and things.
And yeah, been a pleasure to

1316
01:25:20,680 --> 01:25:22,160
have to spend this time with
you.

1317
01:25:23,160 --> 01:25:24,920
OK.
Thank you for the interview.

1318
01:25:24,920 --> 01:25:28,120
It was very, very nice to
discuss about all this.

1319
01:25:28,880 --> 01:25:29,640
Great.
Thank you.