Machine learning meets nanotechnology, award-winning implant regulates blood pressure

1
00:00:05,154 --> 00:00:06,193
Physics world.
2
00:00:07,072 --> 00:00:09,563
Hello, and welcome to the physics World Weekly
3
00:00:09,722 --> 00:00:11,397
Podcast. I'm Hamish Johnston.
4
00:00:11,875 --> 00:00:14,507
In this episode, I'm in conversation with a
5
00:00:14,507 --> 00:00:15,007
physicist
6
00:00:15,384 --> 00:00:15,884
turned
7
00:00:16,261 --> 00:00:19,705
computational scientist who applies Machine learning to a
8
00:00:19,705 --> 00:00:21,778
wide range of research problems,
9
00:00:22,496 --> 00:00:23,873
including nano technology.
10
00:00:24,808 --> 00:00:26,164
Also in this podcast,
11
00:00:26,642 --> 00:00:29,707
physics world's Tammy Freeman meets a scientist
12
00:00:30,084 --> 00:00:32,735
who has created an award winning medical
13
00:00:33,111 --> 00:00:36,139
implant that could help regulate the blood pressure
14
00:00:36,139 --> 00:00:38,625
of people with spinal cord injuries.
15
00:00:39,582 --> 00:00:41,895
But first, a word from our sponsor.
16
00:00:49,651 --> 00:00:52,548
This podcast is sponsored by ice egg, which
17
00:00:52,608 --> 00:00:55,979
specializes in the development and production. Of high
18
00:00:55,979 --> 00:00:57,257
voltage power supplies,
19
00:00:58,216 --> 00:01:01,253
especially for research, science and large experiments.
20
00:01:01,813 --> 00:01:04,870
By using modern res converter technology,
21
00:01:05,660 --> 00:01:08,994
say delivers very efficient and high precision power
22
00:01:08,994 --> 00:01:11,137
supplies in different form factors,
23
00:01:11,613 --> 00:01:13,756
with excellent electrical parameters,
24
00:01:14,169 --> 00:01:16,242
and very low ripple and noise.
25
00:01:16,960 --> 00:01:20,787
These supplies include high precision, source measure units,
26
00:01:21,425 --> 00:01:22,700
detector bias supplies,
27
00:01:23,354 --> 00:01:25,030
Dc dc converters,
28
00:01:25,589 --> 00:01:29,261
Ac dc power supplies, and multi channel high
29
00:01:29,261 --> 00:01:30,219
voltage systems.
30
00:01:30,778 --> 00:01:34,010
For more information, please visit is
31
00:01:34,469 --> 00:01:35,530
website at
32
00:01:36,310 --> 00:01:38,310
dash h v dot com.
33
00:01:45,899 --> 00:01:48,699
Amanda Barn is a senior professor in the
34
00:01:48,699 --> 00:01:52,141
school of computing at the Australian National University,
35
00:01:52,834 --> 00:01:55,814
where she is also deputy director And
36
00:01:56,187 --> 00:01:57,379
computational science lead.
37
00:01:58,253 --> 00:02:00,421
Amanda began her career as a physicist,
38
00:02:00,874 --> 00:02:04,308
but has since broadened her research interests to
39
00:02:04,308 --> 00:02:07,661
encompass many aspects of computational science,
40
00:02:08,460 --> 00:02:10,935
including the use of machine learning in nano
41
00:02:10,935 --> 00:02:11,435
technology,
42
00:02:12,292 --> 00:02:13,170
material science,
43
00:02:13,744 --> 00:02:14,244
chemistry
44
00:02:14,622 --> 00:02:15,340
and medicine.
45
00:02:16,618 --> 00:02:19,411
Amanda joins me down the line from Canberra,
46
00:02:19,890 --> 00:02:22,066
Hello and welcome to the podcast,
47
00:02:23,491 --> 00:02:25,315
Hi, Hamish. A pleasure to be with you?
48
00:02:25,553 --> 00:02:27,561
So, Amanda, can you give us a flavor
49
00:02:27,616 --> 00:02:30,496
of what you do as a computational scientist?
50
00:02:31,043 --> 00:02:32,237
Yes. That's a great question.
51
00:02:32,874 --> 00:02:36,137
The computational science in general is the design
52
00:02:36,137 --> 00:02:38,580
and use of mathematical models to
53
00:02:39,017 --> 00:02:42,621
analyze computationally demanding problems that are experienced in
54
00:02:42,679 --> 00:02:44,509
lots of areas of sites in engineering.
55
00:02:45,226 --> 00:02:48,384
And and this includes advances in computational infrastructure
56
00:02:48,583 --> 00:02:49,243
and algorithms
57
00:02:49,857 --> 00:02:50,755
that enable
58
00:02:51,210 --> 00:02:54,097
researchers across these different domains to perform large
59
00:02:54,155 --> 00:02:56,120
scale computational experiments, know
60
00:02:57,118 --> 00:02:57,678
confidential science.
61
00:02:58,477 --> 00:03:00,874
If you will, it involves research into high
62
00:03:00,874 --> 00:03:04,071
performance computing, not just research using a high
63
00:03:04,071 --> 00:03:06,158
performance computer, At a lot of the time
64
00:03:06,158 --> 00:03:08,305
we spend on algorithms are trying to figure
65
00:03:08,305 --> 00:03:10,055
out how to implement them in such a
66
00:03:10,055 --> 00:03:12,838
way that makes best use of the advanced
67
00:03:12,838 --> 00:03:15,168
hardware. And that hardware is changing all the
68
00:03:15,168 --> 00:03:15,407
time.
69
00:03:16,366 --> 00:03:19,663
This includes both sort of computational or conventional
70
00:03:19,962 --> 00:03:22,655
simulations, based on mathematical models that
71
00:03:23,014 --> 00:03:26,471
developed specifically in different scientific domains, so a
72
00:03:26,610 --> 00:03:28,628
computational chemistry that's developed by chemists
73
00:03:29,007 --> 00:03:30,945
or physics has developed by physicist
74
00:03:31,658 --> 00:03:33,494
We also spend a lot of time using
75
00:03:33,494 --> 00:03:36,129
methods in machine learning and artificial intelligence as
76
00:03:36,129 --> 00:03:36,767
you mentioned,
77
00:03:37,566 --> 00:03:39,721
which is an disappointing area, because most of
78
00:03:39,721 --> 00:03:41,807
them were developed by computer scientists.
79
00:03:42,523 --> 00:03:44,432
And this enables a whole bunch of new
80
00:03:44,432 --> 00:03:46,262
approaches to be used in all of these
81
00:03:46,262 --> 00:03:47,080
different sciences
82
00:03:48,427 --> 00:03:51,401
In at a sort of fundamental level simulation
83
00:03:51,620 --> 00:03:53,775
was born out of theoretical out of the
84
00:03:53,775 --> 00:03:55,770
theoretical aspects of each of our areas.
85
00:03:56,261 --> 00:03:58,560
And sort of added with some convenient levels
86
00:03:58,560 --> 00:03:59,511
of abstraction,
87
00:04:00,065 --> 00:04:02,998
enabled us to solve the equations. But when
88
00:04:02,998 --> 00:04:05,650
we developed all of those theories. They were
89
00:04:05,650 --> 00:04:06,629
sort of an overs
90
00:04:06,930 --> 00:04:08,769
simplification of the problem. And that was done
91
00:04:08,769 --> 00:04:11,090
either in the pursuit of, you know, mathematical
92
00:04:11,090 --> 00:04:13,009
elegance or just just being practical,
93
00:04:13,899 --> 00:04:16,204
Machine learning has the advantage of enabling us
94
00:04:16,204 --> 00:04:18,747
to recapture a lot of the complexity that
95
00:04:18,747 --> 00:04:21,846
we lost when we derive those beautiful theories.
96
00:04:22,898 --> 00:04:26,169
But, unfortunately, not machine learning works well with
97
00:04:26,169 --> 00:04:29,041
science. And so computational scientists spend a lot
98
00:04:29,041 --> 00:04:29,440
of time.
99
00:04:30,014 --> 00:04:31,868
Trying to figure out how to
100
00:04:32,244 --> 00:04:34,396
apply these algorithms that will never intended to
101
00:04:34,396 --> 00:04:36,091
be used for these kinds of datasets
102
00:04:36,467 --> 00:04:39,016
and to overcome some of the the problems
103
00:04:39,016 --> 00:04:41,972
that are experience to the interface. And and
104
00:04:41,972 --> 00:04:43,799
that's that's 1 of the exciting areas I
105
00:04:43,799 --> 00:04:43,958
like.
106
00:04:45,309 --> 00:04:47,314
And, amanda, you began your career is a
107
00:04:47,314 --> 00:04:50,535
physicist. What made you make the move to
108
00:04:50,675 --> 00:04:51,314
computational science?
109
00:04:52,435 --> 00:04:54,115
I did begin as a physicist. I think
110
00:04:54,115 --> 00:04:56,275
it's a great starting point for virtually anything.
111
00:04:56,529 --> 00:04:58,514
I think businesses can do virtual anything.
112
00:04:59,387 --> 00:05:01,372
I think I was always on the path
113
00:05:01,372 --> 00:05:04,152
to computational slides, so maybe didn't realize it.
114
00:05:04,723 --> 00:05:07,030
But from my very first research project as
115
00:05:07,030 --> 00:05:08,643
of students, which used
116
00:05:09,018 --> 00:05:11,246
computational methods, I was I instantly hooked.
117
00:05:12,293 --> 00:05:14,200
I love the code, all the way from
118
00:05:14,200 --> 00:05:16,106
the code to the final results, I I
119
00:05:16,106 --> 00:05:19,362
kinda instantly knew that supercomputer were destined to
120
00:05:19,362 --> 00:05:20,871
be my scientific instrument.
121
00:05:22,483 --> 00:05:23,921
It was exciting to think about, you know,
122
00:05:24,081 --> 00:05:26,238
what could a material scientist do if they
123
00:05:26,238 --> 00:05:28,395
could make perfect samples every time.
124
00:05:28,969 --> 00:05:30,723
Or what could a chemist do if they
125
00:05:30,723 --> 00:05:31,780
could remove all
126
00:05:32,157 --> 00:05:33,672
contamination and do perfect reactions.
127
00:05:34,628 --> 00:05:36,881
What could we do if we could explore
128
00:05:37,114 --> 00:05:40,411
harsh or dangerous environments without risk of injuring
129
00:05:40,471 --> 00:05:42,709
anyone. And more importantly, what if we could
130
00:05:42,709 --> 00:05:44,807
do all of these things simultaneously,
131
00:05:45,665 --> 00:05:46,225
on demand,
132
00:05:46,878 --> 00:05:47,835
every time we tried.
133
00:05:48,711 --> 00:05:51,819
Well, supercomputer are the only instrument that enables
134
00:05:51,819 --> 00:05:52,616
us to do that,
135
00:05:53,333 --> 00:05:53,833
and
136
00:05:54,463 --> 00:05:56,213
I find that the beauty of them and
137
00:05:56,213 --> 00:05:59,021
what activates me most using this instrument
138
00:05:59,396 --> 00:06:01,464
is not that I can reproduce what my
139
00:06:01,464 --> 00:06:02,499
colleagues can do in the lab.
140
00:06:03,549 --> 00:06:05,147
But that I can do everything they can't
141
00:06:05,147 --> 00:06:05,787
do in the lab.
142
00:06:06,825 --> 00:06:09,702
So from the very early days, my my
143
00:06:09,702 --> 00:06:10,261
computational physics,
144
00:06:11,460 --> 00:06:14,745
was on a computer, like, computational chemistry then
145
00:06:14,745 --> 00:06:14,985
evolved,
146
00:06:15,862 --> 00:06:19,051
through now to materials, materials informatics and pretty
147
00:06:19,051 --> 00:06:20,986
much now exclusively machine learning
148
00:06:21,376 --> 00:06:23,145
But I've always focused on the methods
149
00:06:23,517 --> 00:06:25,102
in each of these areas.
150
00:06:25,578 --> 00:06:26,054
And I think,
151
00:06:27,084 --> 00:06:29,798
foundation in physics enabled, meet it to think
152
00:06:29,798 --> 00:06:30,695
very creatively
153
00:06:31,467 --> 00:06:33,636
about our approach all of these up hearing
154
00:06:34,011 --> 00:06:36,396
computationally. And you mentioned machine learning.
155
00:06:36,968 --> 00:06:38,962
What is machine learning and how do you
156
00:06:38,962 --> 00:06:41,355
use it in your research? Most of my
157
00:06:41,355 --> 00:06:42,950
research is now machine learning, probably,
158
00:06:43,828 --> 00:06:45,923
80 percent of it. I still do some
159
00:06:46,141 --> 00:06:46,595
conventional
160
00:06:47,194 --> 00:06:50,011
simulations. And I use these 2 approaches because
161
00:06:50,470 --> 00:06:52,787
they are very different, and they give me
162
00:06:52,787 --> 00:06:54,600
selling very differently in my science so
163
00:06:55,755 --> 00:06:58,146
simulations fundamentally are a a bottom up approach.
164
00:06:58,784 --> 00:07:00,161
We start with some
165
00:07:00,936 --> 00:07:02,371
understanding of a system or problem.
166
00:07:02,944 --> 00:07:05,417
We run a simulation, and then we get
167
00:07:05,417 --> 00:07:06,215
some data at the end.
168
00:07:07,412 --> 00:07:09,487
Informatics or machine learning is a top down
169
00:07:09,487 --> 00:07:09,806
approach.
170
00:07:10,539 --> 00:07:12,530
It we start with the data. We run
171
00:07:12,530 --> 00:07:13,884
a model, and then we end up with
172
00:07:13,884 --> 00:07:15,795
a better understanding of the system problem.
173
00:07:17,468 --> 00:07:19,553
Simulation is based on rules that's you know,
174
00:07:19,712 --> 00:07:20,746
a is our science,
175
00:07:21,303 --> 00:07:24,246
whereas machine learning is based on experiences and
176
00:07:24,246 --> 00:07:24,724
history.
177
00:07:26,394 --> 00:07:28,065
Simulations often, a lot of them are largely,
178
00:07:28,238 --> 00:07:29,928
deter stick, although there are some
179
00:07:30,538 --> 00:07:32,521
some examples of s cast methods such as
180
00:07:32,679 --> 00:07:33,234
Monte carlo.
181
00:07:33,869 --> 00:07:35,851
And machine learning as light s acid,
182
00:07:36,583 --> 00:07:38,975
although there are any examples that are deter
183
00:07:38,975 --> 00:07:40,250
as well. But for the most part.
184
00:07:41,446 --> 00:07:44,019
With simulations, I'm able to do very good
185
00:07:44,316 --> 00:07:44,476
extrapolation,
186
00:07:45,608 --> 00:07:47,363
a lot of the theories that would derived
187
00:07:47,363 --> 00:07:49,118
that form the foundation for simulations,
188
00:07:49,757 --> 00:07:52,331
enable us to explore areas of
189
00:07:53,123 --> 00:07:55,119
configuration space or areas of the science problem
190
00:07:55,119 --> 00:07:58,173
that we haven't observed any data or any
191
00:07:58,232 --> 00:07:58,392
information,
192
00:07:59,110 --> 00:08:01,639
whereas informatics is really good at inter in
193
00:08:01,759 --> 00:08:03,751
f in all the gaps. It's very good
194
00:08:03,751 --> 00:08:04,389
for inputs.
195
00:08:05,983 --> 00:08:08,135
The 2 methods are based on very different
196
00:08:08,135 --> 00:08:08,852
kinds of logic,
197
00:08:10,064 --> 00:08:12,221
simulation is based on an if then else
198
00:08:12,221 --> 00:08:14,139
logic, which means, you know, if I have
199
00:08:14,139 --> 00:08:14,799
a certain
200
00:08:15,178 --> 00:08:18,069
problem or a certain set of conditions. Then
201
00:08:18,148 --> 00:08:19,686
I'll get us a determined
202
00:08:20,702 --> 00:08:21,899
answer or else well.
203
00:08:23,017 --> 00:08:25,991
Computationally, it'll probably crash get on. Whereas
204
00:08:26,464 --> 00:08:28,705
Machine learning is based on an estimate improve
205
00:08:28,705 --> 00:08:31,105
repeat logic, which means it will always give
206
00:08:31,105 --> 00:08:33,764
an answer that answers always improve.
207
00:08:34,716 --> 00:08:36,390
But it may not always be right.
208
00:08:37,346 --> 00:08:37,983
And so that's,
209
00:08:38,620 --> 00:08:40,612
another difference. And, as I mentioned before,
210
00:08:41,662 --> 00:08:42,162
inter
211
00:08:42,535 --> 00:08:43,035
intra
212
00:08:43,806 --> 00:08:44,782
simulations are inter,
213
00:08:45,870 --> 00:08:48,967
very close relation to the domain knowledge and
214
00:08:48,967 --> 00:08:50,397
relies on human intelligence,
215
00:08:51,210 --> 00:08:53,389
whereas machine learning is inter
216
00:08:54,809 --> 00:08:58,090
using the models developed outside of the original
217
00:08:58,090 --> 00:08:59,024
domain and is
218
00:08:59,462 --> 00:09:02,431
agnostic to domain knowledge and provides heavily on
219
00:09:02,568 --> 00:09:03,284
artificial intelligence.
220
00:09:04,319 --> 00:09:05,753
So this is why I like to combine
221
00:09:05,753 --> 00:09:06,788
these 2 approaches,
222
00:09:07,678 --> 00:09:10,065
And and what machine learning in particular, I
223
00:09:10,065 --> 00:09:12,133
think is it important for science is that
224
00:09:12,133 --> 00:09:14,065
before the advent of machine learning scientists.
225
00:09:14,933 --> 00:09:19,024
Had to pretty much understand the the relationships
226
00:09:19,162 --> 00:09:21,498
between the airports and the outputs between the
227
00:09:21,875 --> 00:09:23,960
at actually had to have the structure of
228
00:09:23,960 --> 00:09:26,188
the model predetermined before we were able to
229
00:09:26,188 --> 00:09:26,586
solve it.
230
00:09:27,541 --> 00:09:28,973
It kind of means we need to have
231
00:09:28,973 --> 00:09:30,883
the answer before we could solve for an
232
00:09:30,883 --> 00:09:31,042
answer.
233
00:09:31,853 --> 00:09:33,206
Whereas using machine learning,
234
00:09:34,320 --> 00:09:36,866
because it the idea is the machines they
235
00:09:36,866 --> 00:09:40,470
use statistical techniques and historical information to basically
236
00:09:40,527 --> 00:09:41,579
program them, cells.
237
00:09:42,298 --> 00:09:43,597
What this means is
238
00:09:44,136 --> 00:09:46,374
that we are able to develop the structure
239
00:09:46,374 --> 00:09:48,611
of an expression or an equation and solve
240
00:09:48,611 --> 00:09:49,910
it at the same time
241
00:09:50,865 --> 00:09:53,665
that represents an acceleration is a scientific method,
242
00:09:54,305 --> 00:09:55,985
and it's another reason I like to use
243
00:09:55,985 --> 00:09:58,465
it. The kind of machine learning I use,
244
00:09:58,719 --> 00:10:01,508
are, are quite diverse. There's a lot of
245
00:10:01,508 --> 00:10:03,818
different places and types of machine learning just
246
00:10:03,818 --> 00:10:05,945
that there are lots of different types
247
00:10:06,543 --> 00:10:09,018
computation physics or experimental physics.
248
00:10:10,375 --> 00:10:11,355
I use
249
00:10:11,972 --> 00:10:15,086
unsupervised learning, which is based entirely on input
250
00:10:15,086 --> 00:10:15,586
variables
251
00:10:16,460 --> 00:10:16,960
and
252
00:10:17,420 --> 00:10:18,720
it looks at
253
00:10:19,100 --> 00:10:22,300
developing hidden patents or trying to find representative
254
00:10:22,300 --> 00:10:22,700
data,
255
00:10:23,514 --> 00:10:27,669
And that's useful entire for materials in that
256
00:10:27,669 --> 00:10:30,946
science when we haven't done the experiments to
257
00:10:30,946 --> 00:10:33,593
perhaps measure a property. But we know quite
258
00:10:33,593 --> 00:10:35,981
a bit about the equal conditions that we
259
00:10:35,981 --> 00:10:37,493
put in to develop material.
260
00:10:38,369 --> 00:10:40,916
Unsupervised learning can be useful in looking at
261
00:10:40,916 --> 00:10:41,951
finding hidden patterns,
262
00:10:42,603 --> 00:10:43,502
such as
263
00:10:44,039 --> 00:10:46,033
similarities in our high dimensional space.
264
00:10:47,708 --> 00:10:50,920
I also use supervised machine learning to find
265
00:10:51,217 --> 00:10:51,935
relationships and trends,
266
00:10:52,590 --> 00:10:55,550
such as structured property relationships, which are important
267
00:10:55,550 --> 00:10:57,309
in materials and in nano neuroscience.
268
00:10:57,790 --> 00:10:59,170
And this includes our
269
00:10:59,550 --> 00:10:59,790
classification,
270
00:11:00,444 --> 00:11:02,679
where we have a discrete label. So we
271
00:11:02,679 --> 00:11:04,615
already have maybe different categories
272
00:11:05,233 --> 00:11:05,733
of
273
00:11:06,351 --> 00:11:08,745
nano particles, for example, and we wanna be
274
00:11:08,745 --> 00:11:11,339
able to based on their characteristics
275
00:11:12,759 --> 00:11:13,259
automatically
276
00:11:14,440 --> 00:11:16,735
assign them to either 1 category or another
277
00:11:17,094 --> 00:11:19,430
and make sure that we can easily separate
278
00:11:19,568 --> 00:11:21,266
between these classes based on
279
00:11:21,644 --> 00:11:22,782
input data alone
280
00:11:23,240 --> 00:11:23,719
or
281
00:11:24,278 --> 00:11:27,406
another example is regression and that uses continuous
282
00:11:27,406 --> 00:11:27,885
variables,
283
00:11:28,444 --> 00:11:30,301
and this looks at relationships
284
00:11:31,079 --> 00:11:34,512
such as, a temperature independent relationship and being
285
00:11:34,512 --> 00:11:35,684
able to predict
286
00:11:36,762 --> 00:11:39,976
the an output property or AAA
287
00:11:40,115 --> 00:11:41,951
target label as we would call it,
288
00:11:42,669 --> 00:11:44,505
by evaluating a cost function,
289
00:11:45,876 --> 00:11:49,067
I use statistical learning and semi supervised learning
290
00:11:49,067 --> 00:11:49,625
as well.
291
00:11:50,662 --> 00:11:52,815
Statistical learning in particular is quite useful in
292
00:11:52,815 --> 00:11:54,784
science although it's not widely used
293
00:11:55,221 --> 00:11:55,459
yet.
294
00:11:56,251 --> 00:11:59,131
It... We think of that as causal
295
00:11:59,820 --> 00:12:02,199
that is used in medical diagnostics quite a
296
00:12:02,199 --> 00:12:02,437
lot.
297
00:12:03,166 --> 00:12:05,568
And this can be applied to effectively diagnose
298
00:12:05,940 --> 00:12:06,440
how
299
00:12:06,892 --> 00:12:09,904
a material, for example, might be created rather
300
00:12:09,904 --> 00:12:12,004
than just... Why it is created.
301
00:12:12,930 --> 00:12:14,911
I don't really use a lot of reinforcement
302
00:12:14,911 --> 00:12:17,288
learning, which is another type of machine learning
303
00:12:17,288 --> 00:12:20,581
that's designed to find hidden behaviors and strategies
304
00:12:21,034 --> 00:12:23,177
as an agent navigate an environment.
305
00:12:23,971 --> 00:12:26,432
And so, amanda, your research group includes people
306
00:12:26,432 --> 00:12:29,398
with a wide range of scientific interests? Can
307
00:12:29,398 --> 00:12:31,792
you give us a a flavor of some
308
00:12:31,792 --> 00:12:34,106
of the things that that they're studying?
309
00:12:34,919 --> 00:12:37,319
Yes. It's very generous diverse. And that when
310
00:12:37,480 --> 00:12:39,720
I started him physically Never a thought that
311
00:12:39,879 --> 00:12:42,120
I'd be surrounded by such an amazing group
312
00:12:42,120 --> 00:12:46,290
of. Different scientific areas and and inks and
313
00:12:46,290 --> 00:12:46,929
smart people.
314
00:12:47,728 --> 00:12:48,468
So the
315
00:12:48,847 --> 00:12:50,626
computational science cluster at A.
316
00:12:51,576 --> 00:12:53,507
Has a team that includes
317
00:12:54,120 --> 00:12:55,177
environmental scientists,
318
00:12:55,630 --> 00:12:56,528
earth scientists
319
00:12:57,061 --> 00:12:57,538
by
320
00:12:58,095 --> 00:12:59,923
computational biologists bio politicians,
321
00:13:00,654 --> 00:13:01,534
a generalist assist,
322
00:13:02,574 --> 00:13:04,754
computational on neuroscience, quantum chemistry,
323
00:13:05,615 --> 00:13:07,134
material science, plasma physics,
324
00:13:07,869 --> 00:13:08,369
astro,
325
00:13:08,826 --> 00:13:12,038
astronomy engineering and me and technology. So Rick
326
00:13:12,176 --> 00:13:13,293
were quite a diverse bunch.
327
00:13:14,250 --> 00:13:16,164
To give you an idea of some of
328
00:13:16,164 --> 00:13:17,654
the work that's going on, I thought I
329
00:13:18,091 --> 00:13:20,473
maybe mention some very small things and then
330
00:13:20,473 --> 00:13:21,585
maybe some very big things.
331
00:13:22,935 --> 00:13:25,690
Giuseppe Barker is a junior academic in the
332
00:13:25,808 --> 00:13:27,870
school who is in the area of quantum
333
00:13:27,870 --> 00:13:28,370
chemistry.
334
00:13:28,980 --> 00:13:32,651
And his work is developing algorithms at under
335
00:13:33,340 --> 00:13:36,215
Google their kind of quantum chemistry software packages
336
00:13:36,215 --> 00:13:37,646
that are used all around the world.
337
00:13:39,235 --> 00:13:40,799
So his work is
338
00:13:41,477 --> 00:13:43,812
focused on how we can leverage new processes
339
00:13:43,950 --> 00:13:45,008
such as accelerators,
340
00:13:45,784 --> 00:13:48,597
and how we can rethink how large molecules
341
00:13:48,736 --> 00:13:50,273
can be petition and fragmented.
342
00:13:50,984 --> 00:13:53,402
So that we can strategically combine
343
00:13:53,781 --> 00:13:57,377
massively parallel workflows, so being able to do
344
00:13:57,377 --> 00:13:59,555
much more sophisticated quantum chemistry
345
00:14:00,107 --> 00:14:03,524
and embarrassing parallel workflows that enables us to
346
00:14:03,524 --> 00:14:05,431
break it up into smaller pieces so that
347
00:14:05,431 --> 00:14:06,861
we can do bigger molecules.
348
00:14:08,311 --> 00:14:10,386
And his work will facilitate the use of
349
00:14:10,386 --> 00:14:11,365
highly accurate
350
00:14:11,742 --> 00:14:14,476
post hydro frog methods to be applied to
351
00:14:14,789 --> 00:14:16,463
large bio molecules, for example,
352
00:14:17,420 --> 00:14:20,848
he could also help us to use supercomputer
353
00:14:20,848 --> 00:14:22,841
more efficiently, which saves energy,
354
00:14:23,573 --> 00:14:25,474
And for the last 2 years, he's held
355
00:14:25,474 --> 00:14:26,370
the world record
356
00:14:26,742 --> 00:14:29,198
in the best scaling quantum chemistry algorithm. So
357
00:14:29,198 --> 00:14:30,253
he's to be congratulate.
358
00:14:32,236 --> 00:14:33,353
Also on a small scale,
359
00:14:34,071 --> 00:14:35,609
small in terms of
360
00:14:36,625 --> 00:14:39,020
the the scale of the the science,
361
00:14:39,754 --> 00:14:41,454
is min b, who's a bio
362
00:14:41,754 --> 00:14:44,075
invitation, and he works in the area of
363
00:14:44,075 --> 00:14:44,575
phylogenetic.
364
00:14:46,075 --> 00:14:49,124
He's working on new statistical models for phylogenetic
365
00:14:49,124 --> 00:14:52,137
genomic systems such as petition models, iso small
366
00:14:52,137 --> 00:14:53,009
physical rare models,
367
00:14:54,198 --> 00:14:55,942
distribution tree models and models,
368
00:14:56,830 --> 00:14:59,769
And he's looking at applications in areas that
369
00:14:59,769 --> 00:15:00,667
include photosynthesis
370
00:15:01,120 --> 00:15:04,083
synthetic enzymes or deep insect.
371
00:15:05,667 --> 00:15:06,146
Transcription data,
372
00:15:06,785 --> 00:15:09,578
and has done work around you looking into,
373
00:15:09,977 --> 00:15:11,515
Algae as well as via
374
00:15:12,212 --> 00:15:15,445
viruses such H hiv at Covid 19.
375
00:15:17,024 --> 00:15:19,184
He's... A lot of his work has been
376
00:15:19,184 --> 00:15:20,785
captured in a software package,
377
00:15:21,439 --> 00:15:23,991
that has... He's been developing since 20 11,
378
00:15:24,150 --> 00:15:26,623
over more than 10 years now called Iq
379
00:15:26,623 --> 00:15:28,935
tree, and he has tens of thousands of
380
00:15:28,935 --> 00:15:29,435
users
381
00:15:29,892 --> 00:15:30,131
worldwide,
382
00:15:30,944 --> 00:15:31,444
And
383
00:15:31,820 --> 00:15:33,334
he's actually 1 of for the last 2
384
00:15:33,334 --> 00:15:35,883
years has been a highly cited research worldwide
385
00:15:35,883 --> 00:15:36,680
for his working,
386
00:15:37,477 --> 00:15:40,106
phylogenetic genomics and the Tree software.
387
00:15:41,393 --> 00:15:42,903
So now flipping to something big,
388
00:15:43,698 --> 00:15:46,718
final example is traveling Dang, who is also
389
00:15:46,718 --> 00:15:47,672
a junior academic.
390
00:15:48,244 --> 00:15:50,634
In the group, and he is a mathematician.
391
00:15:50,953 --> 00:15:53,901
He's working on Math be modeling and simulation
392
00:15:53,901 --> 00:15:54,401
for
393
00:15:55,907 --> 00:15:58,525
large scale media such as oceans and atmosphere
394
00:15:58,525 --> 00:16:01,167
dynamics and looks at modeling antarctic
395
00:16:01,539 --> 00:16:01,936
flows.
396
00:16:02,983 --> 00:16:06,240
And his work uses, scientific computing and numerical
397
00:16:06,240 --> 00:16:06,716
solve,
398
00:16:07,749 --> 00:16:08,702
including parallel,
399
00:16:09,655 --> 00:16:10,768
computing and pre conditioners,
400
00:16:11,418 --> 00:16:14,683
and Pd all such as if vm, fun
401
00:16:14,683 --> 00:16:17,470
element sorry. That many people might be familiar
402
00:16:17,470 --> 00:16:17,709
with.
403
00:16:18,505 --> 00:16:18,823
And,
404
00:16:20,432 --> 00:16:22,506
so that they... And also machine learning a
405
00:16:22,506 --> 00:16:24,819
little bit as well. And so there's a
406
00:16:24,819 --> 00:16:26,893
kind of an an example of the... From
407
00:16:26,893 --> 00:16:28,488
the very very small to very big,
408
00:16:29,858 --> 00:16:32,079
of the different kinds of applications that can
409
00:16:32,079 --> 00:16:34,698
be all clustered together in computational science.
410
00:16:35,808 --> 00:16:37,554
I think the best part about it all
411
00:16:37,554 --> 00:16:38,783
is that
412
00:16:39,556 --> 00:16:42,022
we sometimes discover where a problem experienced in
413
00:16:42,022 --> 00:16:44,488
1 domain has actually been already solved in
414
00:16:44,488 --> 00:16:44,727
another.
415
00:16:45,378 --> 00:16:47,627
And when would a computational neuroscientist
416
00:16:48,001 --> 00:16:49,910
ever work with a plasma businesses that just
417
00:16:49,910 --> 00:16:50,864
wouldn't normally happen.
418
00:16:51,818 --> 00:16:54,362
But even more exciting is when we discover
419
00:16:54,362 --> 00:16:56,847
scientific problems that are experiencing 1 and domain
420
00:16:56,847 --> 00:16:59,319
are actually also experienced in another.
421
00:16:59,956 --> 00:17:02,109
And if we solve them, our work and
422
00:17:02,109 --> 00:17:02,928
scale super,
423
00:17:03,559 --> 00:17:05,179
1 solution, more
424
00:17:05,559 --> 00:17:07,640
areas of impact, and that's that's the best.
425
00:17:09,160 --> 00:17:11,240
As well as working with your research group,
426
00:17:11,480 --> 00:17:15,109
amanda, you're also serving as deputy director of
427
00:17:15,167 --> 00:17:18,515
the Australian National University university's school of computing.
428
00:17:18,914 --> 00:17:20,428
Can can you tell us a bit about
429
00:17:20,428 --> 00:17:20,906
that role?
430
00:17:21,798 --> 00:17:23,734
Yes. Well, unfortunately, it's largely
431
00:17:24,191 --> 00:17:26,185
administrative, but the best part is that I
432
00:17:26,185 --> 00:17:28,817
have the opportunity to work with not only
433
00:17:28,817 --> 00:17:30,891
an amazing group of computer scientists across.
434
00:17:31,862 --> 00:17:32,657
Data science,
435
00:17:33,770 --> 00:17:35,701
foundational areas, in languages,
436
00:17:36,553 --> 00:17:37,586
software developments,
437
00:17:38,223 --> 00:17:39,256
site security,
438
00:17:39,828 --> 00:17:42,290
computer vision, robotics and a lot of exciting
439
00:17:42,290 --> 00:17:43,005
areas. Like, well,
440
00:17:43,720 --> 00:17:46,183
I also get to create opportunities when you
441
00:17:46,183 --> 00:17:48,104
people, to join the school and to be
442
00:17:48,104 --> 00:17:48,604
the
443
00:17:49,057 --> 00:17:49,692
of themselves.
444
00:17:50,327 --> 00:17:52,233
So modify my work in the a leadership
445
00:17:52,233 --> 00:17:54,559
role is around if is without the people
446
00:17:54,695 --> 00:17:57,970
to And this includes recruitment looking after our
447
00:17:57,970 --> 00:18:00,514
team track program and our career development,
448
00:18:01,150 --> 00:18:02,661
professional development program as well.
449
00:18:03,473 --> 00:18:06,187
Oh, I've also had the opportunity to start
450
00:18:06,187 --> 00:18:08,742
some brand new programs or areas that I
451
00:18:08,742 --> 00:18:11,456
thought were in need of particular attention.
452
00:18:12,189 --> 00:18:14,179
And 1 example is during Covid,
453
00:18:15,931 --> 00:18:17,762
we all experience, you know, a lot of
454
00:18:17,762 --> 00:18:18,399
a shutdown.
455
00:18:18,957 --> 00:18:21,916
And with not... Accessing our labs and wondering
456
00:18:21,916 --> 00:18:23,348
what we can do. And I took the
457
00:18:23,348 --> 00:18:26,133
opportunity to develop a program called the J
458
00:18:26,133 --> 00:18:27,406
joint fellowship program,
459
00:18:28,535 --> 00:18:29,035
which
460
00:18:29,490 --> 00:18:32,219
support research is working right at the interface
461
00:18:32,674 --> 00:18:33,471
between site...
462
00:18:34,824 --> 00:18:36,988
Computer science, and some other domain,
463
00:18:38,022 --> 00:18:40,330
where they're solving grand challenges in their domain,
464
00:18:40,489 --> 00:18:43,274
but also where they're using that domain knowledge
465
00:18:43,274 --> 00:18:46,319
to inform new types computer sites as well.
466
00:18:46,720 --> 00:18:48,960
And 5 of them was supported across different
467
00:18:48,960 --> 00:18:49,460
areas
468
00:18:50,400 --> 00:18:50,900
in
469
00:18:51,440 --> 00:18:53,299
in 20 21
470
00:18:54,089 --> 00:18:54,747
I've also
471
00:18:55,359 --> 00:18:57,502
cheering the pioneering women program,
472
00:18:57,979 --> 00:19:01,154
which has, a, combination of scholarships and ships
473
00:19:01,154 --> 00:19:01,949
and fellowship.
474
00:19:02,599 --> 00:19:05,560
To support women entering the field of computing
475
00:19:06,092 --> 00:19:06,592
and
476
00:19:07,522 --> 00:19:09,745
make sure that they're successful throughout their with
477
00:19:09,745 --> 00:19:10,245
us
478
00:19:11,353 --> 00:19:12,949
And of course, 1 of my other roles
479
00:19:12,949 --> 00:19:15,264
is deputy directors to look after to computing
480
00:19:15,264 --> 00:19:18,058
facilities for our school, as you could imagine,
481
00:19:18,617 --> 00:19:20,940
school of computer science. Needs to have quite
482
00:19:20,940 --> 00:19:22,368
extensive computer facilities.
483
00:19:23,003 --> 00:19:24,909
And well, that's always been a favorite thing
484
00:19:24,909 --> 00:19:26,973
of mine. So I look after the facilities,
485
00:19:27,211 --> 00:19:30,118
those school looking at ways that we can
486
00:19:30,336 --> 00:19:32,271
diversify our pipeline of resources
487
00:19:32,649 --> 00:19:34,962
to get through tough times like in Covid
488
00:19:34,962 --> 00:19:37,376
where we couldn't actually order any new equipment
489
00:19:38,403 --> 00:19:40,789
and also how we can be more energy
490
00:19:40,789 --> 00:19:41,743
efficient because,
491
00:19:42,856 --> 00:19:45,082
computing uses up enormous amount of energy.
492
00:19:45,814 --> 00:19:46,214
And
493
00:19:47,093 --> 00:19:49,410
if that's not enough, Amanda, you're you're also
494
00:19:49,410 --> 00:19:52,786
editor in chief of the Io Journal Nano
495
00:19:53,005 --> 00:19:55,088
futures. Could can you tell us a bit
496
00:19:55,486 --> 00:19:57,471
about this journal? What are its aims?
497
00:19:58,028 --> 00:19:58,822
Yes. Thanks, Hamish.
498
00:19:59,696 --> 00:20:01,523
That's great because that's where a lot of
499
00:20:01,523 --> 00:20:02,873
my passions will get to come together.
500
00:20:03,604 --> 00:20:06,236
And I learn more about the cutting edge
501
00:20:06,236 --> 00:20:08,810
of nano science and technology, which is my
502
00:20:09,028 --> 00:20:10,703
research area in computational science.
503
00:20:11,420 --> 00:20:13,095
So the nano features journal.
504
00:20:13,589 --> 00:20:17,178
Publishes innovative, urgent work that reflects diverse and
505
00:20:17,178 --> 00:20:19,194
multi fields in data science
506
00:20:19,810 --> 00:20:20,948
and brings together
507
00:20:21,326 --> 00:20:23,375
researchers from a lot of the areas that
508
00:20:23,494 --> 00:20:25,650
I experience in my day to day work
509
00:20:25,650 --> 00:20:28,225
in the school of computing from physics, chemistry,
510
00:20:28,683 --> 00:20:31,876
bio medicine materials, engineering and also people at
511
00:20:31,876 --> 00:20:32,674
work. Extensive
512
00:20:33,248 --> 00:20:33,408
Industry.
513
00:20:34,446 --> 00:20:36,383
This covers biotechnology technology,
514
00:20:37,480 --> 00:20:40,753
quantum the data materials phenomenon, data our energy,
515
00:20:41,152 --> 00:20:42,290
then electronics
516
00:20:42,844 --> 00:20:45,182
low dimensional materials and my favorite
517
00:20:45,721 --> 00:20:47,398
computational nano materials design.
518
00:20:48,437 --> 00:20:51,565
In each case, what differentiates this journal from
519
00:20:51,565 --> 00:20:54,903
our sister General technology and other journals in
520
00:20:54,903 --> 00:20:56,335
the Nano neuroscience area,
521
00:20:56,811 --> 00:20:59,753
is the Nano futures is specifically forward looking.
522
00:21:00,150 --> 00:21:02,799
It's and publishes work that anticipates to set
523
00:21:02,799 --> 00:21:05,374
a new direction in emerging fields
524
00:21:06,071 --> 00:21:08,466
with the expectation that these are the important
525
00:21:08,466 --> 00:21:11,197
areas have long term scientific value an impact.
526
00:21:12,236 --> 00:21:15,032
You mentioned the the future, looking towards the
527
00:21:15,032 --> 00:21:17,030
future. It must be a very exciting time.
528
00:21:17,603 --> 00:21:21,020
For, people doing research in in machine learning,
529
00:21:21,577 --> 00:21:24,517
the technology is finding so many different uses.
530
00:21:24,755 --> 00:21:26,400
And, you know, it's even in the headline
531
00:21:26,599 --> 00:21:29,571
I was listening to to Bbc radio 4
532
00:21:29,710 --> 00:21:32,581
this morning. And once again, they were talking
533
00:21:32,581 --> 00:21:35,133
about so the implications of of,
534
00:21:36,984 --> 00:21:38,902
artificial intelligence on society.
535
00:21:39,461 --> 00:21:42,598
What new applications of machine learning are you
536
00:21:42,658 --> 00:21:45,295
looking forward to in your fields of research?
537
00:21:46,266 --> 00:21:48,096
Well probably some of the ones you're already
538
00:21:48,096 --> 00:21:49,289
hearing about this morning.
539
00:21:49,926 --> 00:21:52,950
While, there are risks associated with the eye
540
00:21:52,950 --> 00:21:54,914
ai. It is also enormous
541
00:21:55,353 --> 00:21:57,906
opportunity. And I think the generative Ai is
542
00:21:57,906 --> 00:21:59,603
going to be particularly important
543
00:21:59,980 --> 00:22:01,597
in the coming years for science
544
00:22:02,467 --> 00:22:04,212
not just my area of science but many
545
00:22:04,212 --> 00:22:05,085
areas of science,
546
00:22:06,037 --> 00:22:07,624
provided we can come over some of the
547
00:22:07,624 --> 00:22:08,838
issues within hall.
548
00:22:09,607 --> 00:22:09,924
And,
549
00:22:10,654 --> 00:22:12,489
But no matter what area of science we're
550
00:22:12,489 --> 00:22:14,642
in whether we're doing computational or whether we're
551
00:22:14,642 --> 00:22:15,599
doing experiments.
552
00:22:16,078 --> 00:22:18,311
We're all at the moment in a position
553
00:22:18,311 --> 00:22:20,519
where we're suffering under a number of restriction
554
00:22:20,798 --> 00:22:22,813
were restricted by the time we have
555
00:22:23,189 --> 00:22:25,261
by the money that resources the equipment we
556
00:22:25,261 --> 00:22:26,138
have access to.
557
00:22:26,776 --> 00:22:28,784
And what this is means, pensions Is that
558
00:22:28,784 --> 00:22:32,000
scientists are compromising our science to fit these
559
00:22:32,219 --> 00:22:34,695
limitations rather than focusing on overcoming them.
560
00:22:35,509 --> 00:22:37,501
And I really believe that the infrastructure shouldn't
561
00:22:37,501 --> 00:22:39,333
dictate what we do, which it shouldn't really
562
00:22:39,333 --> 00:22:40,289
be the other way around.
563
00:22:41,007 --> 00:22:42,281
I think generated Ai,
564
00:22:43,093 --> 00:22:45,323
actually has come at the right to enable
565
00:22:45,323 --> 00:22:47,075
us to find me overcome some of these
566
00:22:47,075 --> 00:22:47,575
problems
567
00:22:48,349 --> 00:22:50,420
because it has a lot of potential to
568
00:22:50,420 --> 00:22:51,455
be able to fill in the gaps,
569
00:22:52,186 --> 00:22:54,493
and provide us with an idea of what
570
00:22:54,493 --> 00:22:56,481
science we could have done if we had
571
00:22:56,481 --> 00:22:58,629
had those resources, and and that we could
572
00:22:58,629 --> 00:22:59,663
do for whatever reason.
573
00:23:01,028 --> 00:23:03,406
It could enable us to get, basically get
574
00:23:03,406 --> 00:23:05,943
more by doing less and avoiding some of
575
00:23:05,943 --> 00:23:07,687
the pitfalls like selection bias.
576
00:23:08,418 --> 00:23:10,571
That is a really big problem when applying
577
00:23:10,571 --> 00:23:12,267
machine learning to science datasets.
578
00:23:13,362 --> 00:23:15,037
There... There's a lot more work that needs
579
00:23:15,037 --> 00:23:17,244
to go into this area to ensure that
580
00:23:17,443 --> 00:23:18,556
generative methods
581
00:23:18,954 --> 00:23:21,578
producing meaningful science, not hallucinations.
582
00:23:23,009 --> 00:23:25,076
But... And this is particularly important if they're
583
00:23:25,076 --> 00:23:27,319
going to form the foundation for large pre
584
00:23:27,319 --> 00:23:29,075
trade models, but I think this is going
585
00:23:29,075 --> 00:23:31,810
to be a really exciting era of science
586
00:23:32,108 --> 00:23:34,683
where we're working collaboratively with Ai
587
00:23:35,061 --> 00:23:37,550
rather than in just performing a task for
588
00:23:37,550 --> 00:23:37,710
us.
589
00:23:38,588 --> 00:23:40,503
Well, that's great. Thanks so much for for
590
00:23:40,503 --> 00:23:42,338
coming on the podcast demand. It sounds like
591
00:23:42,338 --> 00:23:43,695
you've got a an amazing...
592
00:23:44,349 --> 00:23:45,169
Research program
593
00:23:45,548 --> 00:23:47,146
they're in Canberra, and
594
00:23:48,664 --> 00:23:50,502
thanks again for talking about it with us.
595
00:23:50,902 --> 00:23:52,500
You're welcome. Pleasure Be here.
596
00:23:59,787 --> 00:24:02,278
That was amanda barn of the Us Australian
597
00:24:02,597 --> 00:24:03,416
National University,
598
00:24:04,192 --> 00:24:06,505
and she's also editor in chief of the
599
00:24:06,664 --> 00:24:09,855
Journal Nano futures, which you can find on
600
00:24:09,855 --> 00:24:11,664
the Io science website
601
00:24:12,422 --> 00:24:16,251
Up next is physics world's medical physics expert
602
00:24:16,410 --> 00:24:18,665
Tammy Freeman, and she's in
603
00:24:19,042 --> 00:24:22,009
conversation with the inventor of an implant that
604
00:24:22,009 --> 00:24:24,904
could help regulate the blood pressure of people
605
00:24:24,963 --> 00:24:26,559
with spinal cord injuries.
606
00:24:34,640 --> 00:24:38,235
Spinal cord injuries can impact or completely eliminate.
607
00:24:38,489 --> 00:24:41,382
A person's ability to control movement or feel
608
00:24:41,600 --> 00:24:41,839
sensations,
609
00:24:42,716 --> 00:24:44,173
a less well known consequence,
610
00:24:44,551 --> 00:24:47,023
which affects over 90 percent of people with
611
00:24:47,023 --> 00:24:47,741
such injuries.
612
00:24:48,474 --> 00:24:50,468
The inability to regulate blood pressure.
613
00:24:51,026 --> 00:24:53,259
This can result in dizziness, nausea,
614
00:24:53,818 --> 00:24:54,296
faint,
615
00:24:54,855 --> 00:24:56,870
or in some cases, can keep a person
616
00:24:57,008 --> 00:24:57,886
completely bedridden driven.
617
00:24:59,096 --> 00:25:01,985
Aiming to solve this problem is Jordan Square
618
00:25:02,201 --> 00:25:03,020
from Ep,
619
00:25:03,396 --> 00:25:06,263
the Swiss Federal Institute of Technology in Los.
620
00:25:07,311 --> 00:25:10,272
Jordan has developed an implant that delivers electrical
621
00:25:10,803 --> 00:25:14,057
stimulation to spinal neurons to treat dangerously low
622
00:25:14,057 --> 00:25:14,613
blood pressure.
623
00:25:15,739 --> 00:25:18,282
The significance of this work is highlighted by
624
00:25:18,282 --> 00:25:21,405
his recent award of the 20 23 Bio
625
00:25:21,541 --> 00:25:24,164
innovation Institute and science price for innovation.
626
00:25:25,133 --> 00:25:27,122
I'm speaking to Jordan to find out more
627
00:25:27,122 --> 00:25:28,156
about this new therapy.
628
00:25:28,951 --> 00:25:31,281
Hello, Jordan, and welcome to the podcast.
629
00:25:32,451 --> 00:25:33,247
Thanks for having me.
630
00:25:34,375 --> 00:25:37,000
So first of all, can you briefly explain
631
00:25:37,000 --> 00:25:39,226
the clinical problem that you are aiming to
632
00:25:39,226 --> 00:25:39,465
solve?
633
00:25:40,181 --> 00:25:41,453
Yeah. Absolutely. So...
634
00:25:42,264 --> 00:25:45,060
People with spinal cord injuries, specifically,
635
00:25:46,259 --> 00:25:48,256
severe ones or ones that are very high
636
00:25:48,256 --> 00:25:48,815
up in the neck.
637
00:25:49,548 --> 00:25:51,872
And in addition to other people at different
638
00:25:51,927 --> 00:25:55,257
neuro gender disorders like multiple system atrophy in
639
00:25:55,416 --> 00:25:56,843
Parkinson's disease suffer from...
640
00:25:57,494 --> 00:25:59,409
Very unstable blood pressure.
641
00:26:00,047 --> 00:26:01,265
This means that
642
00:26:02,201 --> 00:26:04,435
when they sit up in the morning,
643
00:26:05,329 --> 00:26:07,650
instead of all of the blood vessels constrict
644
00:26:07,650 --> 00:26:09,569
as they should to keep the blood going
645
00:26:09,569 --> 00:26:10,929
up to the brain and the heart,
646
00:26:11,650 --> 00:26:14,069
the blood starts to pool down and gravity
647
00:26:14,130 --> 00:26:14,863
takes its toll.
648
00:26:15,581 --> 00:26:18,133
And this has really significant consequences not only
649
00:26:18,133 --> 00:26:20,287
for their health, but just for their day
650
00:26:20,287 --> 00:26:21,882
to day life and being able to sit
651
00:26:21,882 --> 00:26:23,738
up and not feel dizzy and nauseous
652
00:26:24,195 --> 00:26:26,035
toll, and this is a condition that's called
653
00:26:26,035 --> 00:26:29,845
ortho static hypotension. Okay. So you've developed this
654
00:26:29,845 --> 00:26:30,345
neuro
655
00:26:30,718 --> 00:26:31,591
stimulation implants.
656
00:26:32,162 --> 00:26:34,069
How does this actually work to control a
657
00:26:34,069 --> 00:26:35,260
person's blood pressure?
658
00:26:35,816 --> 00:26:37,667
So we did a lot of experiments
659
00:26:38,040 --> 00:26:39,709
in animal models.
660
00:26:41,794 --> 00:26:44,822
Which effectively allowed us to figure out that
661
00:26:44,822 --> 00:26:47,132
there's a very specific spot of the spinal
662
00:26:47,132 --> 00:26:47,372
cord,
663
00:26:48,104 --> 00:26:50,417
in the what's called the lower Jurassic spinal
664
00:26:50,417 --> 00:26:51,954
cord would be kind of,
665
00:26:52,730 --> 00:26:54,746
where the ribs are ending approximately.
666
00:26:55,761 --> 00:26:57,696
And this part of the spinal cord
667
00:26:58,489 --> 00:27:00,267
just so happens to control
668
00:27:00,966 --> 00:27:02,425
most of the blood vessels
669
00:27:02,804 --> 00:27:05,201
that are in the abdomen. So this is
670
00:27:05,201 --> 00:27:06,559
all of the ones in the gut.
671
00:27:07,369 --> 00:27:11,105
And this is where generally approximately 30 percent
672
00:27:11,105 --> 00:27:13,727
of our blood volume would be at any
673
00:27:13,727 --> 00:27:16,229
given time. And so because we can... Stimulate
674
00:27:16,285 --> 00:27:18,057
that part of the spot bird, which then
675
00:27:18,431 --> 00:27:20,974
stimulates these blood vessels to constrict.
676
00:27:21,689 --> 00:27:24,470
We can control that, you know approximately 30
677
00:27:24,470 --> 00:27:25,763
percent of the blood volume,
678
00:27:26,162 --> 00:27:28,080
and this has a very, very big effect
679
00:27:28,080 --> 00:27:30,796
on blood pressure. So we're kind of hijacking
680
00:27:30,796 --> 00:27:31,455
the system
681
00:27:32,166 --> 00:27:34,890
After you initially published your findings.
682
00:27:35,582 --> 00:27:38,227
I gather that neurologists at a local hospital
683
00:27:38,363 --> 00:27:40,844
asked, to test your implant to treat a
684
00:27:40,844 --> 00:27:42,914
patient in their care. So can you tell
685
00:27:42,914 --> 00:27:44,028
us a bit about this case?
686
00:27:45,142 --> 00:27:46,120
Absolutely. So
687
00:27:46,814 --> 00:27:48,030
1 of the neuro surgeons
688
00:27:49,297 --> 00:27:50,253
that we work with,
689
00:27:51,289 --> 00:27:53,360
who's also the head of the center. This
690
00:27:53,360 --> 00:27:55,352
research was done in at Jocelyn Block,
691
00:27:56,482 --> 00:27:58,474
the neurologist came to her and and we
692
00:27:58,474 --> 00:28:00,705
started to discuss a a patient under their
693
00:28:00,705 --> 00:28:03,309
care who was suffered from quite severe multiple
694
00:28:03,429 --> 00:28:05,123
system. So this is a neuro
695
00:28:05,657 --> 00:28:06,055
disorder.
696
00:28:06,772 --> 00:28:09,159
That's very well known to have very severe
697
00:28:09,159 --> 00:28:11,961
ortho static type hypertension system And so our
698
00:28:11,961 --> 00:28:14,749
original publication was in people's spinal trees. So
699
00:28:14,749 --> 00:28:16,284
is slightly different, but
700
00:28:17,377 --> 00:28:19,768
we thought that maybe the same principles would
701
00:28:19,768 --> 00:28:20,166
apply.
702
00:28:20,897 --> 00:28:22,727
And so after they contacted us, we did
703
00:28:22,727 --> 00:28:24,397
a few experiments in the lab to see
704
00:28:24,397 --> 00:28:26,227
if it might be reasonable that it would
705
00:28:26,227 --> 00:28:26,545
work.
706
00:28:27,754 --> 00:28:30,729
And when those were results came back positive
707
00:28:31,666 --> 00:28:33,901
because her condition was so severe, she was
708
00:28:33,901 --> 00:28:34,401
bedridden
709
00:28:34,779 --> 00:28:37,015
unable to even sit up at all, unable
710
00:28:37,015 --> 00:28:40,057
to participate in life despite being cognitive very
711
00:28:40,057 --> 00:28:40,376
intact.
712
00:28:41,491 --> 00:28:44,758
We decided to move forward and and and
713
00:28:44,917 --> 00:28:45,417
J
714
00:28:47,004 --> 00:28:49,558
did the surgery, and we implanted the system.
715
00:28:50,197 --> 00:28:52,193
And I think it... You know, we can
716
00:28:52,193 --> 00:28:54,268
say quite confidently. It it changed her life.
717
00:28:54,428 --> 00:28:55,726
She was able to
718
00:28:57,074 --> 00:28:58,665
to sit up, and she was actually even
719
00:28:58,665 --> 00:29:00,494
able to walk for hundreds of meters years,
720
00:29:00,653 --> 00:29:02,562
something that she wasn't able to do before.
721
00:29:04,012 --> 00:29:06,246
Excellent. So I mean, is it quite a
722
00:29:06,246 --> 00:29:08,719
large surgical procedure? Is it quite a big
723
00:29:08,719 --> 00:29:10,888
deal to get the implant put in? So
724
00:29:10,888 --> 00:29:13,041
it's... It is a surgery with which comes
725
00:29:13,041 --> 00:29:13,360
with
726
00:29:14,078 --> 00:29:16,471
any... The risks of any any surgery, but
727
00:29:16,471 --> 00:29:18,465
this is a surgery that is fairly routine
728
00:29:18,465 --> 00:29:19,821
for neuro neurosurgery to do.
729
00:29:20,394 --> 00:29:23,211
It's very commonly done. It's a very similar
730
00:29:23,351 --> 00:29:26,547
almost identical procedure so that done to treat
731
00:29:26,547 --> 00:29:27,825
things like neuro pain.
732
00:29:28,719 --> 00:29:31,185
And so this is what would be considered
733
00:29:31,185 --> 00:29:32,777
a a low risk procedure.
734
00:29:33,573 --> 00:29:36,199
Okay. And and has your device been tested
735
00:29:36,199 --> 00:29:37,791
in a a clinical trial yet?
736
00:29:38,364 --> 00:29:40,439
So this entire system and all of the
737
00:29:40,439 --> 00:29:42,035
patents related to this work in,
738
00:29:43,631 --> 00:29:45,786
licensed to onward medical, which was a...
739
00:29:46,439 --> 00:29:48,993
Startup up company that originally came out of
740
00:29:48,993 --> 00:29:51,946
the lab of Gu team, who is also
741
00:29:51,946 --> 00:29:54,101
the, head of the center that I'm working
742
00:29:54,101 --> 00:29:54,181
in,
743
00:29:55,074 --> 00:29:56,370
And this
744
00:29:56,826 --> 00:29:57,883
company is now
745
00:29:58,259 --> 00:30:00,910
moving this forward into clinical trials
746
00:30:01,525 --> 00:30:03,436
and into eventual quick commercialization.
747
00:30:04,567 --> 00:30:06,423
Okay. Yeah. I was gonna ask if if
748
00:30:06,960 --> 00:30:09,033
commercialization is going ahead. What sort of time
749
00:30:09,033 --> 00:30:10,150
scale do you think that could be on?
750
00:30:10,389 --> 00:30:12,941
Yeah. Absolutely. So, of course, for commercialization, the
751
00:30:12,941 --> 00:30:14,791
the the step that has to come before
752
00:30:14,791 --> 00:30:17,207
that is these clinical trials and so to
753
00:30:17,265 --> 00:30:19,340
execute these large scale multi center clinical trials
754
00:30:19,340 --> 00:30:21,756
is obviously a a big undertaking. And so
755
00:30:22,869 --> 00:30:24,230
we are working with the company,
756
00:30:24,789 --> 00:30:27,190
and they are are moving forward, you know,
757
00:30:27,269 --> 00:30:29,029
as fast as we can with this. And
758
00:30:29,029 --> 00:30:29,529
so
759
00:30:30,004 --> 00:30:31,838
we are we're very much hoping to start
760
00:30:31,838 --> 00:30:33,194
these trials within the year.
761
00:30:34,152 --> 00:30:36,725
And so we would hope to have the
762
00:30:36,863 --> 00:30:39,336
conclusion of those trials and a potential move
763
00:30:39,336 --> 00:30:40,729
towards if d approval
764
00:30:41,426 --> 00:30:43,741
in the coming 2 to 3 years. You
765
00:30:43,741 --> 00:30:47,253
were recently awarded an innovation price from the
766
00:30:47,333 --> 00:30:50,376
Bio innovation Institute. And the journal science.
767
00:30:50,854 --> 00:30:53,161
So what's the significance of this award you?
768
00:30:53,320 --> 00:30:54,752
Why do you think you you were chosen
769
00:30:54,752 --> 00:30:55,968
as this year's winner,
770
00:30:56,517 --> 00:30:58,267
I mean, it's obviously a huge honor to
771
00:30:58,267 --> 00:30:59,800
in award like this. And
772
00:31:00,334 --> 00:31:02,083
I think the thing that is is most
773
00:31:02,083 --> 00:31:04,150
exciting to me is just the recognition that
774
00:31:04,150 --> 00:31:05,184
this work is important.
775
00:31:05,820 --> 00:31:08,239
Well when we think of people with
776
00:31:08,690 --> 00:31:09,190
paralysis,
777
00:31:10,198 --> 00:31:12,656
I think what most people's minds immediately go
778
00:31:12,656 --> 00:31:15,056
to is is the recovery. Of walking, getting
779
00:31:15,056 --> 00:31:16,491
people up and out of their chairs of
780
00:31:16,491 --> 00:31:19,201
walking again. And I think the recognition of
781
00:31:19,201 --> 00:31:20,259
this work on
782
00:31:20,636 --> 00:31:22,571
on blood pressure and these
783
00:31:22,962 --> 00:31:25,428
the impact this has on people's quality of
784
00:31:25,428 --> 00:31:28,292
life in this quite severe medical condition. I
785
00:31:28,292 --> 00:31:30,612
think it's really exciting to see their recognition
786
00:31:30,612 --> 00:31:32,620
for that work and see it move forward
787
00:31:33,549 --> 00:31:34,978
on on this kind of,
788
00:31:36,328 --> 00:31:37,598
kind of high profile,
789
00:31:39,283 --> 00:31:40,960
award so I think it's really exciting.
790
00:31:41,678 --> 00:31:43,194
And this is something... I mean, you've been
791
00:31:43,194 --> 00:31:45,748
working in this area since your Phd, so
792
00:31:45,748 --> 00:31:47,185
you've been working in this for a while.
793
00:31:47,519 --> 00:31:49,937
Yeah. Absolutely. I started my Phd
794
00:31:52,234 --> 00:31:55,430
working in in autonomic control, blood pressure control.
795
00:31:56,403 --> 00:31:57,918
In people with spinal cord.
796
00:31:58,317 --> 00:32:00,071
And so I've been in this field for,
797
00:32:01,107 --> 00:32:03,181
almost a decade now. And so it's it's
798
00:32:03,181 --> 00:32:04,718
very exciting to see
799
00:32:05,908 --> 00:32:07,903
to see a treatment moving forward in this
800
00:32:07,903 --> 00:32:08,403
area
801
00:32:09,180 --> 00:32:12,373
that, is very, very needed. Excellent. So finally,
802
00:32:12,532 --> 00:32:14,380
I mean, I mean, what's next? What you
803
00:32:14,380 --> 00:32:16,707
actually working on now what are you looking
804
00:32:17,558 --> 00:32:19,544
to do in the future? So I I
805
00:32:19,544 --> 00:32:21,450
think 1 of the things that's been very
806
00:32:21,450 --> 00:32:23,198
clear through working in this field over the
807
00:32:23,198 --> 00:32:25,924
past decade and and working on this work
808
00:32:25,924 --> 00:32:27,142
and seeing this
809
00:32:27,520 --> 00:32:29,993
move all the way from doing mechanistic work
810
00:32:29,993 --> 00:32:31,371
in animals to
811
00:32:32,002 --> 00:32:34,411
to clinical trials and commercialization
812
00:32:34,786 --> 00:32:36,775
is is that we need to think about
813
00:32:36,775 --> 00:32:38,150
the next big step
814
00:32:38,605 --> 00:32:40,139
in this in this
815
00:32:41,166 --> 00:32:43,339
line of work for these people who have
816
00:32:43,557 --> 00:32:45,389
spinal cord injuries and and I think the
817
00:32:45,389 --> 00:32:45,889
next
818
00:32:46,265 --> 00:32:48,359
big problem that we now have a technology
819
00:32:48,496 --> 00:32:49,633
to begin to tackle
820
00:32:50,742 --> 00:32:51,936
is to repair the cord.
821
00:32:52,893 --> 00:32:54,804
Is to repair the injured spinal cord and
822
00:32:54,804 --> 00:32:57,685
working on biological strategies to do that. And
823
00:32:57,685 --> 00:32:59,272
so I think that's that's the next big
824
00:32:59,272 --> 00:33:01,175
step, and that's what we're gonna keep working
825
00:33:01,175 --> 00:33:01,572
on now.
826
00:33:02,365 --> 00:33:04,943
Excellent. Well, I mean, as he says really
827
00:33:05,000 --> 00:33:07,145
fascinating to see these things come from sort
828
00:33:07,145 --> 00:33:10,244
of laboratory experience, and and hopefully for the
829
00:33:10,244 --> 00:33:12,231
for the benefit of of patients in the
830
00:33:12,231 --> 00:33:15,501
clinic. So it's That's great. Well, thanks very
831
00:33:15,501 --> 00:33:16,930
much for speaking to us today.
832
00:33:17,961 --> 00:33:19,469
Of course. Thanks for having me.
833
00:33:26,726 --> 00:33:28,399
I'm afraid that's all the time we have
834
00:33:28,399 --> 00:33:29,457
for this week's podcast
835
00:33:29,849 --> 00:33:31,367
thanks to Amanda Barn,
836
00:33:31,847 --> 00:33:35,044
Jordan Square and Tammy Freeman for joining me
837
00:33:35,044 --> 00:33:37,441
today, and is special thanks to our producer
838
00:33:37,601 --> 00:33:38,320
Fred Isles.
839
00:33:39,131 --> 00:33:41,117
This podcast is brought to you by ice
840
00:33:41,117 --> 00:33:44,215
egg. High voltage power supplies made by ice
841
00:33:44,215 --> 00:33:45,987
egg means high voltage
842
00:33:46,519 --> 00:33:46,916
exactly.
843
00:33:47,409 --> 00:33:50,382
For more information, please visit the
844
00:33:51,079 --> 00:33:52,137
website at
845
00:33:53,153 --> 00:33:54,110
dash h v.
846
00:33:54,763 --> 00:33:55,321
Dot com.
847
00:33:56,039 --> 00:33:58,112
We'll be back again next week when I'll
848
00:33:58,112 --> 00:34:00,743
be chatting with an x ray astronomer about
849
00:34:00,743 --> 00:34:02,918
the challenges and opportunities
850
00:34:03,549 --> 00:34:05,957
associated with observing the cosmos
851
00:34:06,332 --> 00:34:07,388
using telescopes
852
00:34:07,843 --> 00:34:08,661
on satellites
853
00:34:14,999 --> 00:34:16,031
Physics world.