WEBVTT
Language: en-US

1
00:00:02.509 --> 00:00:07.710
<v James Loach>So the job of this first part of the
system is to find things of potential interest

2
00:00:07.710 --> 00:00:13.029
<v James Loach>in a fast, efficient, clean way, and then
basically to push high quality triggers

3
00:00:13.029 --> 00:00:15.470
<v James Loach>forward to the upper brain.

4
00:00:15.779 --> 00:00:16.090
<v James Loach>Hello

5
00:00:16.090 --> 00:00:17.030
<v Speaker 2>and welcome to the

6
00:00:17.030 --> 00:00:19.120
<v Speaker 3>latest episode of the podcast powered by

7
00:00:19.120 --> 00:00:19.770
<v Speaker 2>podcast.

8
00:00:20.430 --> 00:00:25.030
<v Speaker 2>I'm really pleased to be joined by my
colleague James Loach, who is a friend of the

9
00:00:25.030 --> 00:00:25.389
<v Speaker 2>podcast.

10
00:00:25.469 --> 00:00:28.600
<v Speaker 2>He's been on many times, but I think it's
been a little while since you've been on,

11
00:00:28.629 --> 00:00:29.569
<v Speaker 2>James, so welcome back.

12
00:00:30.299 --> 00:00:30.850
<v James Loach>Yeah, I know.

13
00:00:30.930 --> 00:00:32.389
<v James Loach>Yeah, very happy to be back again.

14
00:00:32.759 --> 00:00:33.330
<v Speaker 2>Fantastic.

15
00:00:33.490 --> 00:00:37.909
<v Speaker 2>And today, we're gonna talk about
industrial AI, which is the term, if you listen to

16
00:00:38.049 --> 00:00:43.250
<v Speaker 2>any or see any materials from Siemens, you
know, industrial AI is very important to

17
00:00:43.250 --> 00:00:44.200
<v Speaker 2>what we talk about.

18
00:00:44.529 --> 00:00:47.759
<v Speaker 2>We're gonna go beyond the sort of
marketing speak, Speaking as a marketer as well,

19
00:00:47.810 --> 00:00:51.889
<v Speaker 2>we do talk a lot about industrial AI, but
James is actually doing the real stuff and

20
00:00:51.889 --> 00:00:56.990
<v Speaker 2>the real implementations of it with our,
our customers and, and behind the scenes.

21
00:00:57.459 --> 00:01:00.290
<v Speaker 2>So, yeah, it's gonna be a great
conversation, but I guess before we, we start,

22
00:01:00.380 --> 00:01:04.339
<v Speaker 2>James, it'd be good to, for you to
introduce yourself to the audience and remind us

23
00:01:04.339 --> 00:01:06.580
<v Speaker 2>what your role is in Siemens.

24
00:01:07.050 --> 00:01:08.069
<v James Loach>Yeah, no, no problem.

25
00:01:08.209 --> 00:01:13.489
<v James Loach>OK, so I'm head of research at Siemens
Sensai Predictive Maintenance.

26
00:01:13.769 --> 00:01:18.970
<v James Loach>That's a role I've had for many years, 8
or 9 years now, since this was a very small

27
00:01:18.970 --> 00:01:19.589
<v James Loach>startup.

28
00:01:20.010 --> 00:01:25.279
<v James Loach>My role, yeah, head of the research group
and part of the product leadership team,

29
00:01:25.599 --> 00:01:30.480
<v James Loach>so responsible for developing all of the
algorithms, the analytics, and the

30
00:01:30.480 --> 00:01:33.269
<v James Loach>conceptual structures that those things
sit within.

31
00:01:33.739 --> 00:01:34.860
<v Speaker 2>Fantastic, fantastic.

32
00:01:34.940 --> 00:01:35.419
<v Speaker 2>Thank you for that.

33
00:01:35.540 --> 00:01:40.440
<v Speaker 2>First of all, let's start with that term,
industrial AI, uh, and what it actually

34
00:01:40.889 --> 00:01:42.300
<v Speaker 2>means in reality.

35
00:01:42.500 --> 00:01:46.220
<v Speaker 2>So, not that there is a marketing reality,
but there's also what's happening on the

36
00:01:46.220 --> 00:01:46.480
<v Speaker 2>ground.

37
00:01:47.099 --> 00:01:49.699
<v Speaker 2>So, what's your sort of perspective on
that, James?

38
00:01:50.260 --> 00:01:52.779
<v James Loach>Yeah, I guess maybe I'll put across a very
simple view of this.

39
00:01:52.860 --> 00:01:56.099
<v James Loach>AI is obviously this sort of suite of
technologies that we all know about.

40
00:01:56.339 --> 00:02:00.819
<v James Loach>Industrial AI, yeah, it's just applying
all of those technologies within an

41
00:02:00.819 --> 00:02:01.819
<v James Loach>industrial setting.

42
00:02:02.069 --> 00:02:04.580
<v James Loach>This makes it automatically not one thing,
right?

43
00:02:04.739 --> 00:02:07.860
<v James Loach>There's all kinds of different things that
AI.

44
00:02:07.955 --> 00:02:13.274
<v James Loach>It's very suitable for everything from
supporting coding all the way through to

45
00:02:13.274 --> 00:02:18.755
<v James Loach>agentic decision making or generative
tasks or simpler things like summarization or

46
00:02:18.755 --> 00:02:23.375
<v James Loach>how you mind things in disordered data,
all of these different kinds of things, but

47
00:02:23.375 --> 00:02:28.315
<v James Loach>those applications of the technology start
in the industrial world all over the

48
00:02:28.315 --> 00:02:28.815
<v James Loach>place.

49
00:02:29.259 --> 00:02:29.800
<v James Loach>OK.

50
00:02:30.179 --> 00:02:33.660
<v James Loach>And it's a huge area of opportunities and
what everybody is working out at the

51
00:02:33.660 --> 00:02:37.080
<v James Loach>moment really is how best to apply these
technologies.

52
00:02:37.160 --> 00:02:39.460
<v James Loach>What are the situations where it works
really well?

53
00:02:39.899 --> 00:02:41.970
<v James Loach>What are the situations where we have to
wait.

54
00:02:42.259 --> 00:02:42.539
<v James Loach>Yeah.

55
00:02:43.000 --> 00:02:43.020
<v James Loach>Uh,

56
00:02:43.139 --> 00:02:46.539
<v Speaker 2>I guess just to cover first of all,
because we all know, and you'll know obviously

57
00:02:46.539 --> 00:02:50.080
<v Speaker 2>better than most people, the explosion in
terms of generative AI, but of course,

58
00:02:50.710 --> 00:02:52.679
<v Speaker 2>Senseii, we talk about Senseii
specifically.

59
00:02:52.860 --> 00:02:55.580
<v Speaker 2>AI has been the backbone of Senseii.

60
00:02:56.130 --> 00:02:57.369
<v Speaker 2>From the beginning, right?

61
00:02:57.449 --> 00:03:00.320
<v Speaker 2>So do you think it's almost been forgotten
about these more traditional, I'm gonna

62
00:03:00.320 --> 00:03:02.410
<v Speaker 2>say traditional AI technologies.

63
00:03:02.490 --> 00:03:05.470
<v Speaker 2>Has it been forgotten about in this new
age that we're in at the moment?

64
00:03:05.729 --> 00:03:06.089
<v Speaker 2>Yeah, I

65
00:03:06.089 --> 00:03:06.809
<v James Loach>don't know, it's interesting.

66
00:03:06.889 --> 00:03:11.509
<v James Loach>There's a long history of using machine
learning, which is what you're talking about

67
00:03:11.509 --> 00:03:17.210
<v James Loach>there in the pre-generative AI period to
do lots of interesting tasks and S I, for

68
00:03:17.210 --> 00:03:20.009
<v James Loach>many years before generative AI was doing
things with.

69
00:03:20.115 --> 00:03:25.104
<v James Loach>This kind of technology, but there have
always been issues with scaling, basically,

70
00:03:25.235 --> 00:03:25.434
<v James Loach>right?

71
00:03:25.615 --> 00:03:28.955
<v James Loach>Being able to use this technology for
particular applications but not being able to

72
00:03:28.955 --> 00:03:33.315
<v James Loach>push it out, yeah, at scale, right, across
applications or different machine types

73
00:03:33.315 --> 00:03:34.735
<v James Loach>or all of that sort of thing.

74
00:03:34.835 --> 00:03:38.634
<v James Loach>And one special thing that AI is letting
people do and definitely in the case of

75
00:03:38.634 --> 00:03:43.384
<v James Loach>Sensai, is to take what now seem like sort
of traditional technology as a

76
00:03:43.384 --> 00:03:44.095
<v James Loach>foundation.

77
00:03:44.320 --> 00:03:48.320
<v James Loach>And then, yeah, using AI to make the most
of the things that you were doing before

78
00:03:48.320 --> 00:03:53.240
<v James Loach>and in particular for addressing that sort
of generalization problem or other

79
00:03:53.240 --> 00:03:56.070
<v James Loach>aspects as well, communication of results
and things like that.

80
00:03:56.479 --> 00:04:01.000
<v James Loach>Yeah, but many systems do use a
combination of these things and S I also uses a

81
00:04:01.000 --> 00:04:02.500
<v James Loach>combination of even older stuff, right?

82
00:04:02.600 --> 00:04:06.559
<v James Loach>We have a set of statistical systems, we
have a set of machine learning systems.

83
00:04:06.669 --> 00:04:08.289
<v James Loach>We now have a set of AI systems.

84
00:04:08.369 --> 00:04:11.309
<v James Loach>And one thing which is nice about Sensai
is that these things do are quite

85
00:04:11.309 --> 00:04:13.460
<v James Loach>complementary and work nicely together.

86
00:04:13.710 --> 00:04:17.649
<v Speaker 2>That's really well explained, actually,
the sort of foundations of Sensai with AI.

87
00:04:17.989 --> 00:04:22.070
<v Speaker 2>And what I'm wondering is, obviously,
generative AI has opened a world of

88
00:04:22.070 --> 00:04:22.470
<v Speaker 2>opportunity.

89
00:04:22.589 --> 00:04:26.390
<v Speaker 2>So, what was simply impossible all those
years ago when you started at Sensai, which

90
00:04:26.390 --> 00:04:27.700
<v Speaker 2>is now possible today?

91
00:04:28.670 --> 00:04:34.179
<v James Loach>Yeah, So Sensii has always been about
trying to look at sensor data from all

92
00:04:34.179 --> 00:04:39.000
<v James Loach>different kinds of sensors attached to all
different kinds of machines and do that

93
00:04:39.000 --> 00:04:43.679
<v James Loach>at scale over hundreds or thousands of
assets and try to find things that the user

94
00:04:43.679 --> 00:04:48.260
<v James Loach>would be interested in that essentially
help them with their maintenance, right, so

95
00:04:48.260 --> 00:04:51.260
<v James Loach>typically developing problems and all of
that kind of thing.

96
00:04:51.799 --> 00:04:57.200
<v James Loach>So traditional techniques, if you apply
them well, do allow you to do the first part

97
00:04:57.200 --> 00:05:01.279
<v James Loach>of that pretty well, right, which is to be
looking over all of this data,

98
00:05:01.480 --> 00:05:05.549
<v James Loach>identifying things, essentially really
unusual things or things that seem to have

99
00:05:05.549 --> 00:05:08.910
<v James Loach>some implication, triggering a system and
getting those things to the user.

100
00:05:09.160 --> 00:05:10.420
<v James Loach>But then it becomes very difficult.

101
00:05:10.880 --> 00:05:15.799
<v James Loach>So we can, in the past were able to
identify that a particular machine seemed to be

102
00:05:15.799 --> 00:05:20.130
<v James Loach>having an issue, and we can describe
what's going on, but It was very difficult for

103
00:05:20.130 --> 00:05:25.329
<v James Loach>us to really say what that particular
issue was, and we had to make best use of the

104
00:05:25.329 --> 00:05:29.450
<v James Loach>techniques available to us at the time,
which were things like mapping the data that

105
00:05:29.450 --> 00:05:33.529
<v James Loach>we were seeing to matching the data before
previous failures in the past identified

106
00:05:33.529 --> 00:05:36.809
<v James Loach>from maintenance management systems, these
kinds of techniques.

107
00:05:36.929 --> 00:05:41.440
<v James Loach>But what generative AI really lets you do
in Sensei is to take those triggers, so

108
00:05:41.440 --> 00:05:46.369
<v James Loach>the understanding that there's something
up, pass that to a part of the system.

109
00:05:46.829 --> 00:05:51.600
<v James Loach>Which in our case, we may touch on a bit,
is highly agentic and very smart, that can

110
00:05:51.600 --> 00:05:54.000
<v James Loach>take that trigger as an indication of what
it is.

111
00:05:54.109 --> 00:05:59.000
<v James Loach>There seems to be something going on, and
then to call on all kinds of supplementary

112
00:05:59.000 --> 00:06:00.709
<v James Loach>information about what it might be.

113
00:06:00.959 --> 00:06:05.920
<v James Loach>Everything from the documents to looking
at images of the time series to actually

114
00:06:05.920 --> 00:06:09.730
<v James Loach>getting stuck in and doing some data
analysis and changing the view of the data,

115
00:06:10.160 --> 00:06:12.660
<v James Loach>whatever, all of that kind of information,
bring it together.

116
00:06:12.920 --> 00:06:14.579
<v James Loach>And confirm that there is an issue.

117
00:06:15.160 --> 00:06:19.859
<v James Loach>Explain to the user what it is with some
kind of confidence and go through and say

118
00:06:19.859 --> 00:06:23.200
<v James Loach>what we think you need to do and what's
likely to happen if you don't do it, right,

119
00:06:23.440 --> 00:06:25.299
<v James Loach>and provide all of the evidence for that.

120
00:06:25.600 --> 00:06:29.079
<v James Loach>So that last bit completely unimaginable
before generative AI.

121
00:06:29.119 --> 00:06:32.079
<v James Loach>You would just be surfacing objects of
context.

122
00:06:32.179 --> 00:06:34.279
<v James Loach>So you'd say, oh, here's the thing, it's
similar.

123
00:06:34.630 --> 00:06:38.029
<v James Loach>To that occasion where you had a bearing
problem, this kind of thing, right?

124
00:06:38.149 --> 00:06:41.029
<v James Loach>But yeah, we can be massively more
sophisticated now.

125
00:06:41.149 --> 00:06:41.429
<v James Loach>That's,

126
00:06:41.709 --> 00:06:45.890
<v Speaker 2>you know, some amazing opportunities very
much into the prescriptive side of things.

127
00:06:46.149 --> 00:06:49.230
<v Speaker 2>So it's not just this is happening and
this is what we spotted, but we suggest

128
00:06:49.230 --> 00:06:52.850
<v Speaker 2>you're, you're doing is a lot, is stronger
maybe than it was before.

129
00:06:53.239 --> 00:06:56.920
<v James Loach>Yeah, taking the super intelligences,
which are now available for hire and

130
00:06:56.920 --> 00:07:01.399
<v James Loach>connecting them up to all of this sort of
information that we have and ways of

131
00:07:01.399 --> 00:07:02.679
<v James Loach>interacting with that information.

132
00:07:02.760 --> 00:07:04.309
<v Speaker 2>I guess to balance up a little bit as
well.

133
00:07:04.399 --> 00:07:09.399
<v Speaker 2>So, is it, what areas are still, like,
what challenges do we still have?

134
00:07:09.480 --> 00:07:14.320
<v Speaker 2>What's just as hard as it was before,
maybe that the moment it generative AI hasn't

135
00:07:14.600 --> 00:07:15.140
<v Speaker 2>solved.

136
00:07:15.519 --> 00:07:19.480
<v James Loach>The way I like to look at this, we just
described a little bit of.

137
00:07:19.540 --> 00:07:25.640
<v James Loach>About how Sensii works, but many other
systems have the same kind of form, right?

138
00:07:25.720 --> 00:07:28.869
<v James Loach>Especially, I think, industrial AI
systems.

139
00:07:29.190 --> 00:07:33.209
<v James Loach>Did you have some kind of live information
about what's going on?

140
00:07:33.500 --> 00:07:37.619
<v James Loach>This can be time series, uh, different
kinds of vibration data.

141
00:07:37.700 --> 00:07:40.109
<v James Loach>You can imagine even images, video,
whatever, right?

142
00:07:40.220 --> 00:07:43.320
<v James Loach>You've got something which is information
about what's happening, OK?

143
00:07:43.529 --> 00:07:45.739
<v James Loach>And then you have information about the
setting.

144
00:07:46.100 --> 00:07:50.000
<v James Loach>Say like what the machine is, the history
of the machine, you mean, engineering

145
00:07:50.000 --> 00:07:52.239
<v James Loach>drawings, all of this kind of thing,
right?

146
00:07:52.359 --> 00:07:56.119
<v James Loach>And so you're taking the data that you're
seeing and you're trying to make sense of

147
00:07:56.119 --> 00:08:01.369
<v James Loach>that data fundamentally with what you know
about the context, all of that additional

148
00:08:01.369 --> 00:08:01.579
<v James Loach>information.

149
00:08:01.670 --> 00:08:08.630
<v James Loach>And I think ultimately, what generative AI
allows is to do that kind optimally,

150
00:08:08.760 --> 00:08:09.109
<v James Loach>right?

151
00:08:09.279 --> 00:08:12.920
<v James Loach>So given the data you have, given the
context you have, you know.

152
00:08:13.290 --> 00:08:17.420
<v James Loach>Like what is the most clear and confident
things that you can say, right, that are

153
00:08:17.420 --> 00:08:18.559
<v James Loach>most useful to the user.

154
00:08:18.670 --> 00:08:21.820
<v James Loach>So there's a sense in which you set up the
system properly, you can solve that.

155
00:08:21.940 --> 00:08:26.679
<v James Loach>But the situation that I guess Sensi is
starting to touch, and I think many of these

156
00:08:27.290 --> 00:08:30.500
<v James Loach>applications will, is you have limitations
on that input information.

157
00:08:30.619 --> 00:08:32.140
<v James Loach>So the AI is letting you do.

158
00:08:32.809 --> 00:08:37.409
<v James Loach>Ultimately perfect reasoning over the
input information, but your input information

159
00:08:37.409 --> 00:08:40.609
<v James Loach>compared to the amount of information
that's out there in the world is typically

160
00:08:40.609 --> 00:08:41.549
<v James Loach>very small, right?

161
00:08:41.890 --> 00:08:45.729
<v James Loach>And then this is ultimately going to set
the limit, and there'll be a point where it

162
00:08:45.729 --> 00:08:49.969
<v James Loach>makes no sense to keep upgrading the
models in these kinds of systems without doing

163
00:08:49.969 --> 00:08:51.440
<v James Loach>anything about the information, right?

164
00:08:51.969 --> 00:08:53.809
<v James Loach>And so there will be a lot of
differentiation.

165
00:08:54.309 --> 00:08:57.630
<v James Loach>In systems in how they approach this
problem of getting context.

166
00:08:57.909 --> 00:09:02.030
<v James Loach>Need to get information about the machine,
what the machines are doing, what's

167
00:09:02.030 --> 00:09:06.229
<v James Loach>happened to the machine in the past, how
the machine works, and specifically what

168
00:09:06.229 --> 00:09:07.599
<v James Loach>it's doing now, right?

169
00:09:08.070 --> 00:09:08.190
<v James Loach>Yeah,

170
00:09:08.299 --> 00:09:12.150
<v Speaker 2>and it's interesting because we, you know,
in our discussions before, I think we

171
00:09:12.150 --> 00:09:14.549
<v Speaker 2>knew where this podcast was going to lead,
and you mentioned context.

172
00:09:14.630 --> 00:09:18.030
<v Speaker 2>The first mention with 12 minutes in, so I
thought it was gonna actually be front

173
00:09:18.030 --> 00:09:18.479
<v Speaker 2>and center.

174
00:09:18.630 --> 00:09:21.330
<v Speaker 2>But you mentioned it there about it's all
very well changing models and getting

175
00:09:21.330 --> 00:09:23.989
<v Speaker 2>better models, but if the context isn't
there.

176
00:09:24.409 --> 00:09:27.619
<v Speaker 2>So that, what you're saying maybe, James,
is actually that will become the

177
00:09:27.619 --> 00:09:28.619
<v Speaker 2>competitive advantage.

178
00:09:28.700 --> 00:09:33.739
<v Speaker 2>What context can you bring to these models
to, to, for the outputs to be improved?

179
00:09:34.179 --> 00:09:34.340
<v Speaker 2>Yeah,

180
00:09:34.460 --> 00:09:34.979
<v James Loach>I think so.

181
00:09:35.099 --> 00:09:38.179
<v James Loach>And yeah, so, I could have mentioned
context in response to all of the questions

182
00:09:38.179 --> 00:09:40.169
<v James Loach>earlier, later question.

183
00:09:40.580 --> 00:09:41.150
<v James Loach>But yeah.

184
00:09:41.510 --> 00:09:45.880
<v James Loach>Yeah, it's what you bring as an
organization, all the people developing the software

185
00:09:45.880 --> 00:09:49.559
<v James Loach>to the software, and in the case of
Siemens, you can think about all of this

186
00:09:49.559 --> 00:09:54.020
<v James Loach>collective industrial knowledge, not all
of which is available to these standard

187
00:09:54.020 --> 00:09:58.739
<v James Loach>foundation models, and then, yeah, all
kinds of systems and tricks and techniques

188
00:09:58.739 --> 00:10:04.309
<v James Loach>and ways of getting context basically out
of users' heads, right?

189
00:10:04.840 --> 00:10:08.099
<v James Loach>And as we said, technical stuff about the
machines and different kinds of data.

190
00:10:08.400 --> 00:10:09.419
<v James Loach>And you can imagine.

191
00:10:09.780 --> 00:10:12.729
<v James Loach>And there's now generator AI is also
helpful to this, right?

192
00:10:13.020 --> 00:10:17.659
<v James Loach>So again, it's become a very easy task,
for example, to photograph nameplates on

193
00:10:17.659 --> 00:10:20.849
<v James Loach>machines, pull the text out of that and,
and use it, right?

194
00:10:21.140 --> 00:10:24.940
<v James Loach>So that could be done at some level before
generative AI, but stuff like this just

195
00:10:24.940 --> 00:10:27.260
<v James Loach>becomes very accurate and very easy,
right?

196
00:10:27.700 --> 00:10:31.500
<v James Loach>But yeah, it is exactly as you said, I
think this will be differentiate lots of

197
00:10:31.500 --> 00:10:32.640
<v James Loach>these products, how they approach.

198
00:10:33.400 --> 00:10:37.909
<v James Loach>That context problem and data collection
problem on the input, and then the second

199
00:10:37.909 --> 00:10:39.580
<v James Loach>area really would be on the output.

200
00:10:39.840 --> 00:10:43.760
<v James Loach>Yeah, how you're getting this stuff to
users, how you're interacting with customers

201
00:10:43.760 --> 00:10:45.039
<v James Loach>or, or stuff like that.

202
00:10:45.320 --> 00:10:45.580
<v Speaker 2>Yeah.

203
00:10:46.000 --> 00:10:49.559
<v Speaker 2>And it, I think you touched upon the
elements there, James, or maybe we could spell

204
00:10:49.559 --> 00:10:53.880
<v Speaker 2>it out clearly for your, from your point
of view, what information would an AI

205
00:10:53.880 --> 00:10:56.919
<v Speaker 2>system need to understand what's happening
with the machine?

206
00:10:57.460 --> 00:11:01.979
<v Speaker 2>In an ideal world, which we're not, I
know, but if we had every bit of context

207
00:11:01.979 --> 00:11:06.489
<v Speaker 2>available, what should we be feeding into
the system for it to provide the, the best

208
00:11:06.489 --> 00:11:08.059
<v Speaker 2>output it can do?

209
00:11:08.580 --> 00:11:11.659
<v James Loach>Yeah, I think there's probably a massive
difference here between what's sensible and

210
00:11:11.659 --> 00:11:15.650
<v James Loach>what the perfect real world is, you know,
you know, in the real world, you, no, no,

211
00:11:15.739 --> 00:11:18.460
<v James Loach>I just mean in a perfect real world, it's
you're interested in the physics of

212
00:11:18.460 --> 00:11:19.169
<v James Loach>things, right?

213
00:11:19.340 --> 00:11:23.619
<v James Loach>You know, the complete digital detailed
digital twins, right?

214
00:11:23.700 --> 00:11:29.250
<v James Loach>So very great detail so that the AI can
essentially have access and know that

215
00:11:29.250 --> 00:11:33.530
<v James Loach>information and then also test its
hypothesis against the models derived from that

216
00:11:33.530 --> 00:11:33.700
<v James Loach>information.

217
00:11:33.979 --> 00:11:34.919
<v James Loach>So really, of course.

218
00:11:35.320 --> 00:11:41.159
<v James Loach>In an ideal world, this can be absurdly
detailed and, and impractical, right?

219
00:11:41.559 --> 00:11:45.900
<v James Loach>But if you're going to being more
practical, yeah, you need to know the, the, the,

220
00:11:45.960 --> 00:11:49.390
<v James Loach>the, the basic things, right, from what
the, the machine is, right?

221
00:11:49.559 --> 00:11:54.440
<v James Loach>It's, it's make and a model and what it's
doing down through the details of its

222
00:11:54.440 --> 00:11:59.880
<v James Loach>construction and deployment, different
components inside it, all of that kind of

223
00:11:59.880 --> 00:12:00.239
<v James Loach>thing.

224
00:12:00.630 --> 00:12:01.020
<v James Loach>Yeah.

225
00:12:01.349 --> 00:12:04.669
<v James Loach>And it, it's strange to talk about things
as obvious as the make and model of the

226
00:12:04.669 --> 00:12:09.109
<v James Loach>machine, but in, in many settings, yeah,
getting this information users and getting

227
00:12:09.109 --> 00:12:11.630
<v James Loach>it into the app requires work even for
basic things.

228
00:12:11.840 --> 00:12:12.090
<v James Loach>Yeah.

229
00:12:12.219 --> 00:12:15.750
<v James Loach>And of course, you want to go all the way
down into the physics and engineering as

230
00:12:15.750 --> 00:12:16.690
<v James Loach>deep as you can.

231
00:12:17.080 --> 00:12:20.950
<v James Loach>But yeah, so if you want information at
different levels about what the thing is,

232
00:12:21.349 --> 00:12:23.469
<v James Loach>and then you need to know what it's doing.

233
00:12:23.849 --> 00:12:27.530
<v James Loach>And then you need to how it relates to
other machines around it and the processes,

234
00:12:27.570 --> 00:12:28.419
<v James Loach>all this kind of stuff.

235
00:12:28.770 --> 00:12:33.070
<v James Loach>And then you want to know the history of
things, which is often extremely useful.

236
00:12:33.450 --> 00:12:37.169
<v James Loach>And yeah, that, that's the kind of thing
that's in maintenance engineers' heads.

237
00:12:37.289 --> 00:12:40.809
<v James Loach>They get to know their machines over time,
they understand their quirks, they know

238
00:12:40.809 --> 00:12:44.469
<v James Loach>what they did to them last month to fix X,
Y, Z, all of this stuff.

239
00:12:44.929 --> 00:12:46.690
<v James Loach>Yeah, it's super useful, right?

240
00:12:46.809 --> 00:12:51.159
<v James Loach>So it's the details of what is and then
over time, pushing that back over there.

241
00:12:52.150 --> 00:12:52.330
<v James Loach>It's.

242
00:12:52.989 --> 00:12:54.309
<v James Loach>narrative histories of your

243
00:12:54.309 --> 00:12:54.969
<v Speaker 2>assets.

244
00:12:55.349 --> 00:12:59.460
<v Speaker 2>I guess the, yeah, elements, and if we
focus again, you, like you've mentioned that

245
00:12:59.460 --> 00:13:03.330
<v Speaker 2>phrase, what's in the, the maintenance
person's head, I guess a lot of that.

246
00:13:03.719 --> 00:13:07.679
<v James Loach>Yeah, we, we capture this the best ways
that we can, and we're constantly thinking

247
00:13:07.679 --> 00:13:11.450
<v James Loach>of ways to improve how we do this, and
it's very tricky, right?

248
00:13:11.679 --> 00:13:16.599
<v James Loach>So traditionally in Sensi, it has been,
like you say, feedback on alerts, people

249
00:13:16.599 --> 00:13:19.919
<v James Loach>telling us oh it was this and this was
useful, this wasn't, all of that kind of

250
00:13:19.919 --> 00:13:24.140
<v James Loach>thing, and then maintenance data, so the
actual hard interventions.

251
00:13:24.559 --> 00:13:28.640
<v James Loach>On the machines, but it's very clear that
maintenance engineers have a lot more in

252
00:13:28.640 --> 00:13:29.510
<v James Loach>their head than that.

253
00:13:29.679 --> 00:13:30.039
<v James Loach>Yeah.

254
00:13:30.419 --> 00:13:34.960
<v James Loach>And a lot of that is implicit knowledge, a
feel for things and what have you.

255
00:13:35.280 --> 00:13:35.599
<v James Loach>And it.

256
00:13:36.059 --> 00:13:39.500
<v James Loach>Those things can even be things that they
find themselves difficult to express in

257
00:13:39.500 --> 00:13:39.869
<v James Loach>words.

258
00:13:40.020 --> 00:13:45.289
<v James Loach>But yeah, so it is, it is as you said, and
we're aware that this is only, as always,

259
00:13:45.299 --> 00:13:49.580
<v James Loach>quite partial and try to think of ways to
get more information without people doing

260
00:13:49.580 --> 00:13:50.330
<v James Loach>too much work.

261
00:13:50.659 --> 00:13:50.929
<v James Loach>Yeah.

262
00:13:51.260 --> 00:13:54.140
<v Speaker 2>But of course, we should point out
everything we've said, you know, in an ideal

263
00:13:54.140 --> 00:13:57.780
<v Speaker 2>world and what's practical and what's not,
but, but you can get started with,

264
00:13:58.140 --> 00:14:02.859
<v Speaker 2>there's a minimum you can get started with
and still get impressive results, right?

265
00:14:02.940 --> 00:14:03.799
<v Speaker 2>So it's probably a minimum.

266
00:14:04.400 --> 00:14:07.969
<v Speaker 2>What's ideal and then what's, when you
said about going into the physics and, and

267
00:14:07.969 --> 00:14:11.679
<v Speaker 2>that side of things, which is above and
beyond, but yeah, there is, it's, yeah,

268
00:14:11.880 --> 00:14:16.479
<v Speaker 2>there, there is a minimum starting point
you can get, get started with.

269
00:14:16.880 --> 00:14:20.789
<v James Loach>Yeah, and, and generative AI helps
massively there as well because you, you know,

270
00:14:20.799 --> 00:14:24.440
<v James Loach>have a system there which has this
knowledge of the whole world and everything in it

271
00:14:24.440 --> 00:14:25.799
<v James Loach>up to a certain level.

272
00:14:26.179 --> 00:14:30.640
<v James Loach>Sensor itself is designed to produce
interesting results out of the box, so to spot

273
00:14:31.080 --> 00:14:33.979
<v James Loach>things happening that, that might be of
interest to the user.

274
00:14:34.429 --> 00:14:38.000
<v James Loach>OK, just out of the box, yeah, I mean if
you bring those two things together, you

275
00:14:38.000 --> 00:14:42.349
<v James Loach>can immediately start giving a certain
quality of prescriptive messaging.

276
00:14:42.830 --> 00:14:46.880
<v James Loach>Even just knowing that it's a motor, the
system can understand what the time series

277
00:14:46.880 --> 00:14:50.559
<v James Loach>are, and see what's happening in them, and
it knows all about motors and all about

278
00:14:50.559 --> 00:14:55.130
<v James Loach>time series and can start to say, yeah, at
a certain level immediately.

279
00:14:55.599 --> 00:14:59.390
<v James Loach>Yeah, that this could be a bearing issue
or something like that, and then its

280
00:14:59.390 --> 00:15:04.109
<v James Loach>ability to do that, yeah, it improves as
it gets a better handle on the specifics of

281
00:15:04.109 --> 00:15:06.390
<v James Loach>your motor and what it's specifically
doing.

282
00:15:06.460 --> 00:15:09.140
<v James Loach>And like we're saying, it's history and
all of that.

283
00:15:09.510 --> 00:15:12.950
<v James Loach>But yeah, very important to sense that
it's a system that is designed to work

284
00:15:12.950 --> 00:15:17.669
<v James Loach>outside the box, do something useful and
valuable to you, and then try to be a

285
00:15:17.669 --> 00:15:22.789
<v James Loach>sponge for additional information,
additional context that helps to ratchet up this

286
00:15:22.789 --> 00:15:23.229
<v James Loach>performance over time.

287
00:15:24.700 --> 00:15:24.940
<v Speaker 2>Exactly.

288
00:15:25.059 --> 00:15:28.419
<v Speaker 2>Yeah, so it can build up that context over
time, of course, as well.

289
00:15:28.549 --> 00:15:29.729
<v Speaker 2>Yeah, that's what you're saying.

290
00:15:30.099 --> 00:15:34.419
<v Speaker 2>And I guess, as well, so, we talked a
little bit about the machine learning

291
00:15:34.419 --> 00:15:39.179
<v Speaker 2>foundations of Sensai, but what I thought
might be interesting to people as well,

292
00:15:39.270 --> 00:15:43.210
<v Speaker 2>because, again, Egentic AI is a hot topic,
and a lot of software platforms and

293
00:15:43.210 --> 00:15:45.780
<v Speaker 2>whatever industry are moving in that
direction.

294
00:15:45.820 --> 00:15:49.940
<v Speaker 2>But how does that sort of split, if you
wanna call it a split between the machine

295
00:15:49.940 --> 00:15:51.960
<v Speaker 2>learning side and the Agentic AI.

296
00:15:52.570 --> 00:15:57.570
<v Speaker 2>Side, how does that, I guess first of all,
why, why is a split like that and how

297
00:15:57.570 --> 00:15:58.229
<v Speaker 2>does it work?

298
00:15:58.690 --> 00:16:01.010
<v Speaker 2>How does it benefit the end user having
that split?

299
00:16:01.169 --> 00:16:02.169
<v Speaker 2>Maybe is a good way to.

300
00:16:02.570 --> 00:16:06.359
<v James Loach>Yeah, so the division you're talking there
is quite a hard division in the sensii

301
00:16:06.359 --> 00:16:11.950
<v James Loach>system that we alluded to earlier, and we
often talk about this as the lower brain

302
00:16:11.950 --> 00:16:15.169
<v James Loach>and the upper brain because in some sense,
it's mapping onto these different

303
00:16:15.169 --> 00:16:17.469
<v James Loach>functions in, in, in animal brains.

304
00:16:17.880 --> 00:16:23.280
<v James Loach>But the first part of the system is this
statistical machine learning part, and that

305
00:16:23.280 --> 00:16:27.919
<v James Loach>is designed to operate at massive scale,
maybe monitoring tens or hundreds of

306
00:16:27.919 --> 00:16:32.679
<v James Loach>thousands of time series at a customer and
doing it every time these data points

307
00:16:32.679 --> 00:16:33.280
<v James Loach>come in.

308
00:16:33.719 --> 00:16:36.940
<v James Loach>Yeah, so this is just basically a huge,
huge amount of data.

309
00:16:37.440 --> 00:16:41.919
<v James Loach>And what you are looking for in that first
part of the system is things of potential

310
00:16:41.919 --> 00:16:42.359
<v James Loach>interest.

311
00:16:42.890 --> 00:16:45.549
<v James Loach>Basically, and it would work very well.

312
00:16:45.809 --> 00:16:50.130
<v James Loach>OK, to have a, an AI agent full of data
analysis tools and all kinds of stuff

313
00:16:50.130 --> 00:16:54.849
<v James Loach>analyzing data from every machine at every
single time point, they would be

314
00:16:54.849 --> 00:16:57.070
<v James Loach>prohibitively expensive to do that.

315
00:16:57.609 --> 00:17:02.770
<v James Loach>So, the job of this first part of the
system is to find things of potential interest

316
00:17:02.770 --> 00:17:08.089
<v James Loach>in a fast, efficient, clean way, and then
basically to push high-quality triggers

317
00:17:08.089 --> 00:17:08.510
<v James Loach>forward.

318
00:17:09.219 --> 00:17:14.150
<v James Loach>To the upper brain, yeah, which is the
expensive reasoning part of the system.

319
00:17:14.589 --> 00:17:14.939
<v James Loach>OK?

320
00:17:15.189 --> 00:17:20.589
<v James Loach>And if we make that division very
efficient, we only pass through very few signals

321
00:17:20.589 --> 00:17:25.109
<v James Loach>into the agential part, and it means that
we can spend a lot more resources on the

322
00:17:25.109 --> 00:17:27.310
<v James Loach>agential part to make that very, very
good.

323
00:17:27.699 --> 00:17:28.089
<v James Loach>OK?

324
00:17:28.510 --> 00:17:31.119
<v James Loach>Yeah, so there's this, it's basically a
filter.

325
00:17:31.579 --> 00:17:35.770
<v James Loach>OK, to make sure that you only call the
expensive agent when it's worth its effort

326
00:17:35.770 --> 00:17:39.160
<v James Loach>to do that and you get a return, return on
the resources.

327
00:17:39.579 --> 00:17:40.150
<v James Loach>OK?

328
00:17:40.630 --> 00:17:44.250
<v James Loach>Yeah, and the, the benefit to use it here
is purely one of one of efficiency.

329
00:17:44.910 --> 00:17:50.069
<v James Loach>OK, that the intelligence, which is
expensive, gets directed precisely to where it's

330
00:17:50.069 --> 00:17:54.390
<v James Loach>needed and not wasted on lower value,
low-value stuff, right?

331
00:17:54.689 --> 00:17:59.650
<v James Loach>And the output of the intelligence
insensi, the agentic part, is the thing the user

332
00:17:59.650 --> 00:18:00.609
<v James Loach>sees, right?

333
00:18:00.689 --> 00:18:03.030
<v James Loach>It's composing and structuring and writing
though.

334
00:18:03.650 --> 00:18:05.030
<v James Loach>Yeah, writing the output.

335
00:18:05.569 --> 00:18:08.109
<v James Loach>OK, and we obviously want that to be as
good as possible.

336
00:18:08.849 --> 00:18:12.410
<v Speaker 2>Yeah, so it's interesting we're talking
about this lower half and upper half of the

337
00:18:12.410 --> 00:18:16.770
<v Speaker 2>brain, and I was wondering, like both now
and into the future, whether is it the

338
00:18:16.770 --> 00:18:21.949
<v Speaker 2>50/50, if we talk about percentage terms,
like a 50%, both are equally.

339
00:18:22.260 --> 00:18:27.369
<v Speaker 2>Important in what Sensi produces as an
output, and is that the case now, or, and

340
00:18:27.369 --> 00:18:32.180
<v Speaker 2>will that change in the future as well as
we go to lead more into the agentic side

341
00:18:32.180 --> 00:18:32.780
<v Speaker 2>of things?

342
00:18:32.979 --> 00:18:35.459
<v Speaker 2>I don't know if it's, yeah, uh, what you
think about that.

343
00:18:35.699 --> 00:18:37.520
<v James Loach>Yeah, I know it's an interesting question.

344
00:18:37.780 --> 00:18:42.260
<v James Loach>I think with Sensor systems like sensor,
you have to have both parts of the system.

345
00:18:42.339 --> 00:18:46.619
<v James Loach>So, in some way they are, they're both
necessary for the thing to work, right?

346
00:18:46.699 --> 00:18:48.939
<v James Loach>So both 100% valuable.

347
00:18:49.150 --> 00:18:49.339
<v James Loach>Yeah.

348
00:18:49.459 --> 00:18:52.260
<v James Loach>And then the tension there is, of course,
this thing you Sensai wants to be

349
00:18:52.260 --> 00:18:57.020
<v James Loach>extremely broad, looking across many loads
of data on, on, on lots of different

350
00:18:57.020 --> 00:18:59.140
<v James Loach>assets and then also be maximally deep.

351
00:18:59.300 --> 00:19:02.939
<v James Loach>So for any particular situation of
interest, go as, as deep into that, right?

352
00:19:03.300 --> 00:19:03.660
<v James Loach>So.

353
00:19:03.969 --> 00:19:08.050
<v James Loach>If you try to be maximally broad and
maximally deep, it becomes extremely

354
00:19:08.050 --> 00:19:09.489
<v James Loach>inefficient, right?

355
00:19:09.660 --> 00:19:12.869
<v James Loach>And ultimately unnecessarily expensive.

356
00:19:13.410 --> 00:19:17.390
<v James Loach>And the division of sensi gets you through
this problem by essentially doing a

357
00:19:17.390 --> 00:19:22.260
<v James Loach>light, lighter scan of this space,
identifying only what's needed.

358
00:19:22.430 --> 00:19:26.770
<v James Loach>You, you, you then go through to, to the
second bit, OK?

359
00:19:27.040 --> 00:19:32.319
<v James Loach>And that is very important for systems
like Sensei, which basically acting as agents

360
00:19:32.319 --> 00:19:33.160
<v James Loach>for you, right?

361
00:19:33.400 --> 00:19:36.760
<v James Loach>So Sensite as a whole is something that
you task with looking at data from all of

362
00:19:36.760 --> 00:19:40.660
<v James Loach>your assets all day every day and coming
to you when it's got an issue.

363
00:19:41.119 --> 00:19:43.550
<v James Loach>So there you have this problem to solve.

364
00:19:43.839 --> 00:19:47.959
<v James Loach>Another class of application is ones where
you're starting the conversation with the

365
00:19:47.959 --> 00:19:48.910
<v James Loach>thing, right?

366
00:19:49.130 --> 00:19:53.760
<v James Loach>And in situations like that, you can live
entirely in the expensive agentic space

367
00:19:53.760 --> 00:19:56.900
<v James Loach>and you obviously don't need the earlier
filtering thing.

368
00:19:57.489 --> 00:20:00.209
<v James Loach>OK, but applications like Sensor, you do
need both.

369
00:20:00.449 --> 00:20:03.589
<v James Loach>How this changes over time will be very
interesting.

370
00:20:03.689 --> 00:20:05.729
<v James Loach>And of course we do think about this,
right?

371
00:20:06.010 --> 00:20:11.729
<v James Loach>We do wonder as, as modern intelligence
increases and costs drop.

372
00:20:12.270 --> 00:20:17.719
<v James Loach>At what point it would be feasible to be,
to have the intelligence go deeper and

373
00:20:17.719 --> 00:20:19.910
<v James Loach>deeper into the earlier parts of the
system.

374
00:20:20.280 --> 00:20:25.079
<v James Loach>So, at some point, it will make sense to
start reflecting some of the statistic and

375
00:20:25.079 --> 00:20:28.400
<v James Loach>machine learning things in the lower part
of the brain with a little bit more

376
00:20:28.400 --> 00:20:29.270
<v James Loach>intelligence.

377
00:20:29.560 --> 00:20:34.180
<v James Loach>OK, and we imagine winding this up as it
becomes, as it makes sense resource-wise.

378
00:20:34.680 --> 00:20:35.000
<v James Loach>Yeah.

379
00:20:35.680 --> 00:20:38.939
<v James Loach>But you know it's quite how this will,
will play out, but certainly.

380
00:20:39.729 --> 00:20:45.979
<v James Loach>That that movement is a thing in sensor
and would be a thing in similar systems is

381
00:20:45.979 --> 00:20:50.140
<v James Loach>that intelligence gets cheaper, you use it
for more stuff, right, and push it more

382
00:20:50.140 --> 00:20:51.439
<v James Loach>into the foundations.

383
00:20:52.020 --> 00:20:54.439
<v Speaker 2>And I, I guess that brings me to a more
broader question.

384
00:20:54.699 --> 00:20:58.290
<v Speaker 2>So when we think about AI, you know,
building systems and software themselves and

385
00:20:58.290 --> 00:21:00.140
<v Speaker 2>whether you're seeing.

386
00:21:00.670 --> 00:21:04.319
<v Speaker 2>So, I guess there's a lot of legacy
software that's now, the past few years, in

387
00:21:04.319 --> 00:21:09.000
<v Speaker 2>fact, have been a couple of 2 or 3 years,
been adding AI and generative AI

388
00:21:09.000 --> 00:21:10.800
<v Speaker 2>functionality into their apps.

389
00:21:10.839 --> 00:21:15.280
<v Speaker 2>But I'm wondering whether you're seeing a
shift more towards, you know, new systems

390
00:21:15.280 --> 00:21:19.680
<v Speaker 2>are just being built, you know, AI native
right from the start, rather than layering

391
00:21:19.680 --> 00:21:20.430
<v Speaker 2>AI on top.

392
00:21:20.520 --> 00:21:23.040
<v Speaker 2>I just wondered on a more broad view.

393
00:21:23.319 --> 00:21:25.920
<v Speaker 2>Could be industrial AI, but it could be
beyond that, of course.

394
00:21:26.239 --> 00:21:26.359
<v Speaker 2>Yeah,

395
00:21:26.520 --> 00:21:31.109
<v James Loach>again, this is interesting and Difficult
question.

396
00:21:31.369 --> 00:21:37.219
<v James Loach>In terms of AI being added onto things
that can be added on in different ways, some

397
00:21:37.219 --> 00:21:42.569
<v James Loach>of which make a lot of sense, and other
things which are a bit more tokenistic, like

398
00:21:42.569 --> 00:21:46.489
<v James Loach>the copilots being added to every system
we, we encounter, right?

399
00:21:46.609 --> 00:21:47.849
<v James Loach>They vary a lot in value.

400
00:21:48.209 --> 00:21:52.410
<v James Loach>One interesting thing about since I've
probably mentioned in these podcasts before

401
00:21:52.410 --> 00:21:56.810
<v James Loach>is that we were quite lucky in the sense
that we built the lower brain that I was

402
00:21:56.810 --> 00:21:58.869
<v James Loach>just describing before.

403
00:21:59.479 --> 00:22:04.219
<v James Loach>Generative AI happened and then we had the
opportunity to build out the upper brain.

404
00:22:04.569 --> 00:22:08.089
<v James Loach>So you can look at sensesight and you can
think we had an existing system and we

405
00:22:08.089 --> 00:22:12.719
<v James Loach>added, but that's Lucky in the sense that
if you look at the, the system that gives

406
00:22:12.719 --> 00:22:15.750
<v James Loach>you, it is a system we would have designed
anyway, right?

407
00:22:16.040 --> 00:22:20.800
<v James Loach>If we were starting from scratch because
for the reasons that I, I just described.

408
00:22:21.239 --> 00:22:27.599
<v James Loach>So I'm just making this point that AI can
be added on in ways that make deep sense

409
00:22:27.599 --> 00:22:30.140
<v James Loach>and in ways which are, are more
superficial.

410
00:22:30.599 --> 00:22:35.119
<v James Loach>OK, then in terms of AI for systems, I
think the answer is probably no.

411
00:22:35.199 --> 00:22:39.260
<v James Loach>I'm not, I don't see a lot of super
imaginative AI systems out there.

412
00:22:39.829 --> 00:22:42.989
<v James Loach>I think maybe this is a little bit of a
disappointment because you imagine you have

413
00:22:42.989 --> 00:22:47.989
<v James Loach>this massively new and different
technology that we can all reinvent what apps are

414
00:22:47.989 --> 00:22:52.489
<v James Loach>and do everything different with this kind
of thing as a, as a foundation.

415
00:22:52.869 --> 00:22:58.150
<v James Loach>So you've seen a whole class of back and
forth chat kind of applications that we're

416
00:22:58.150 --> 00:23:02.109
<v James Loach>all familiar with, uh, but with some with
absolutely massive value like the agentic

417
00:23:02.109 --> 00:23:03.290
<v James Loach>coding tools and things.

418
00:23:03.800 --> 00:23:05.800
<v James Loach>So we could say that's 11 category.

419
00:23:06.280 --> 00:23:09.880
<v James Loach>But I don't know how many more there are
than that really, right?

420
00:23:10.339 --> 00:23:10.989
<v James Loach>It, but it's odd.

421
00:23:11.109 --> 00:23:12.520
<v James Loach>It's like we do talk about this a lot.

422
00:23:12.599 --> 00:23:14.969
<v James Loach>We feel that there's something there,
right?

423
00:23:15.119 --> 00:23:17.790
<v James Loach>That these things can be imagined,
reimagined or done a lot better.

424
00:23:17.839 --> 00:23:22.020
<v James Loach>It's close and you can't like a word on
the tip of the tongue kind of feeling.

425
00:23:22.439 --> 00:23:28.199
<v James Loach>But yeah, like a lot, not, not much I
think of huge interest is being, is being

426
00:23:28.199 --> 00:23:29.369
<v James Loach>revealed there at the moment.

427
00:23:29.819 --> 00:23:30.099
<v James Loach>Yeah.

428
00:23:30.540 --> 00:23:35.459
<v James Loach>The 11 thing specific to Senssite I will
mention, but is in this category that we

429
00:23:35.459 --> 00:23:41.099
<v James Loach>ponder about is the ability of the AI to
control the user experience and the user

430
00:23:41.099 --> 00:23:42.010
<v James Loach>interface.

431
00:23:42.500 --> 00:23:48.060
<v James Loach>So that's to me one area where I feel
there will be a bunch of creativity, but you

432
00:23:48.060 --> 00:23:52.459
<v James Loach>see the very beginnings of this in Sensei
because the agent has a language that it's

433
00:23:52.459 --> 00:23:55.619
<v James Loach>taught that it can write to determine.

434
00:23:55.930 --> 00:23:57.170
<v James Loach>What the user sees.

435
00:23:57.489 --> 00:24:01.270
<v James Loach>So in terms of charts and other different
UI elements and things like this, it can,

436
00:24:01.410 --> 00:24:05.109
<v James Loach>it kind of can lay this out and has a
certain amount of flexibility and things.

437
00:24:05.569 --> 00:24:09.310
<v James Loach>But the, the models are capable of much
more than that in terms of laying out web

438
00:24:09.310 --> 00:24:10.959
<v James Loach>pages in applications.

439
00:24:11.449 --> 00:24:14.650
<v James Loach>And I guess we, we wonder whether there's
something interesting that can be done

440
00:24:14.650 --> 00:24:17.790
<v James Loach>there to make that massively more custom
and dynamic.

441
00:24:18.410 --> 00:24:21.650
<v Speaker 2>Uh, so, so I guess to, you know, summarize
it, so you think it.

442
00:24:22.229 --> 00:24:27.469
<v Speaker 2>There could be more opportunity in the,
the front end of these platforms to utilize

443
00:24:27.469 --> 00:24:28.719
<v Speaker 2>AI more.

444
00:24:29.000 --> 00:24:32.319
<v Speaker 2>I know we talked about a lot of coding,
but maybe is that leaning more towards sort

445
00:24:32.319 --> 00:24:35.040
<v Speaker 2>of backend work, maybe not so much
front-end work?

446
00:24:35.079 --> 00:24:36.719
<v Speaker 2>Is that maybe a way to summarize?

447
00:24:36.900 --> 00:24:37.140
<v Speaker 2>That, that,

448
00:24:37.239 --> 00:24:40.199
<v James Loach>that's just one example, but I think that
the thing is this, isn't it?

449
00:24:40.449 --> 00:24:41.939
<v James Loach>Maybe I to be very unimaginative here.

450
00:24:42.119 --> 00:24:44.900
<v James Loach>There are certain form factors that
interacting with compu.

451
00:24:45.430 --> 00:24:47.109
<v James Loach>Computable modalities, right?

452
00:24:47.349 --> 00:24:51.219
<v James Loach>And those are mostly the, the, the things
that are, that, that are.

453
00:24:51.430 --> 00:24:53.989
<v James Loach>And then if you're thinking about these
kind of applications, specifically

454
00:24:53.989 --> 00:24:56.709
<v James Loach>industrial areaer stuff, what, what
options are there, right?

455
00:24:56.790 --> 00:24:59.589
<v James Loach>There's charts, and there's text, and
there's images, and there's audio and there's

456
00:24:59.589 --> 00:25:00.849
<v James Loach>these different things.

457
00:25:01.189 --> 00:25:04.469
<v James Loach>And generally that gives you a lot of
flexibility in how you mesh these things

458
00:25:04.469 --> 00:25:04.979
<v James Loach>together.

459
00:25:05.229 --> 00:25:11.180
<v James Loach>But again, I think there's not been a, a
blooming of 1000 different new interaction

460
00:25:11.180 --> 00:25:14.369
<v James Loach>paradigms that, that are interesting and
exciting.

461
00:25:15.160 --> 00:25:18.640
<v James Loach>It seems to me that there are, there are a
few, and yet this sort of more, more

462
00:25:18.640 --> 00:25:22.130
<v James Loach>dynamic front ends of visual
visualizations is one.

463
00:25:22.599 --> 00:25:26.599
<v James Loach>But anyway, I think it's just an area
where it's not been quite as exciting as it's

464
00:25:26.599 --> 00:25:28.069
<v James Loach>first seemed like it would be.

465
00:25:28.640 --> 00:25:28.939
<v James Loach>No,

466
00:25:29.359 --> 00:25:34.119
<v Speaker 2>no, I bet, yeah, I'm sure it's, it's the
lag before the next sort of big.

467
00:25:34.599 --> 00:25:35.949
<v Speaker 2>Acceleration and innovation.

468
00:25:36.000 --> 00:25:40.119
<v Speaker 2>I guess on that note, as I pre-warned you
before we started recording, we do have a

469
00:25:40.119 --> 00:25:41.349
<v Speaker 2>lightning round coming up.

470
00:25:41.400 --> 00:25:45.010
<v Speaker 2>But I guess before I'm gonna ask that
dreaded question, which I think, particularly

471
00:25:45.010 --> 00:25:48.859
<v Speaker 2>for AI, it's a, it's almost impossible
question to answer.

472
00:25:48.949 --> 00:25:52.560
<v Speaker 2>But with, with you being so embedded,
obviously, in the topic, James, I'd be

473
00:25:52.560 --> 00:25:56.400
<v Speaker 2>interested to get your thoughts on where
you think things are gonna evolve over the

474
00:25:56.400 --> 00:25:59.800
<v Speaker 2>next, you can talk about the next 3
months, if you want the next year or the next 5

475
00:25:59.800 --> 00:26:01.160
<v Speaker 2>years, I'll let you decide.

476
00:26:01.239 --> 00:26:04.359
<v Speaker 2>But just how things are gonna evolve in
this space.

477
00:26:04.790 --> 00:26:08.949
<v James Loach>Yeah, no, it's, it's a very interesting
and difficult question, and I can maybe,

478
00:26:09.069 --> 00:26:12.750
<v James Loach>lot, lots I could say, but I just pick up
some things related to, to what we've

479
00:26:13.099 --> 00:26:13.390
<v James Loach>talked about.

480
00:26:13.469 --> 00:26:18.959
<v James Loach>You imagine that the, these intelligs are
going to continue to, to increase in power

481
00:26:19.189 --> 00:26:23.310
<v James Loach>and so the ability within these systems to
do the inference and reasoning will, yes,

482
00:26:23.510 --> 00:26:25.489
<v James Loach>become, will tend to perfect.

483
00:26:26.020 --> 00:26:30.640
<v James Loach>And then you're held back essentially by
all the non-intelligence things, right,

484
00:26:30.709 --> 00:26:35.479
<v James Loach>which is all the human side of stuff, the
organizational things, fatigue, different

485
00:26:35.479 --> 00:26:40.500
<v James Loach>kinds of data to get access to different
kinds of context, all of that stuff.

486
00:26:40.739 --> 00:26:44.380
<v James Loach>And I think you probably see this, maybe
we've seen it already, but we see it more

487
00:26:44.380 --> 00:26:49.219
<v James Loach>broadly, that, yeah, basically all the
non-intelligent stuff, all the human stuff,

488
00:26:49.339 --> 00:26:51.750
<v James Loach>the existing stuff becomes the bottleneck,
right?

489
00:26:51.930 --> 00:26:54.619
<v James Loach>Anything with intelligence just collapses,
right?

490
00:26:55.060 --> 00:27:00.449
<v James Loach>And then another aspect of this which
again it's not particularly original to say,

491
00:27:00.510 --> 00:27:03.900
<v James Loach>but will be quite impactful is the agentic
side.

492
00:27:04.310 --> 00:27:10.069
<v James Loach>So rather than just reasoning over
information in some sense, the ability to start

493
00:27:10.069 --> 00:27:15.270
<v James Loach>going out and doing things in the outside
world, and there are many systems around

494
00:27:15.270 --> 00:27:20.650
<v James Loach>in the consumer space and what have you
that allow these kinds of interactions, but

495
00:27:20.819 --> 00:27:25.949
<v James Loach>I think apart from coding, You've not seen
a big takeoff because there are just some

496
00:27:25.949 --> 00:27:32.719
<v James Loach>sort of trust issues and again a harness
sort of issues and Guess quality at

497
00:27:32.719 --> 00:27:33.920
<v James Loach>some point, right?

498
00:27:34.020 --> 00:27:37.760
<v James Loach>But eventually, lots of these barriers
will be crossed and you'll get a lot more

499
00:27:37.760 --> 00:27:38.680
<v James Loach>interesting stuff on the side.

500
00:27:39.239 --> 00:27:43.199
<v James Loach>And we haven't talked about that, but
there, there is a situation eventually where

501
00:27:43.199 --> 00:27:47.880
<v James Loach>Sensii is with its lower brain spotting
something of potential interest, thinking

502
00:27:47.880 --> 00:27:49.719
<v James Loach>through what that might be.

503
00:27:50.180 --> 00:27:55.910
<v James Loach>And itself by controlling whatever device
is doing it, robot kind of thing, is to go

504
00:27:55.910 --> 00:28:00.910
<v James Loach>out there and be actually doing the
inspections and getting the information it needs

505
00:28:01.109 --> 00:28:05.670
<v James Loach>ultimately to make a decision about
whether to stop the line needs to be stopped or

506
00:28:05.670 --> 00:28:09.290
<v James Loach>when and what parts need to be bought and
things like that.

507
00:28:09.750 --> 00:28:15.550
<v James Loach>So everyone believes that latter bit I
talked about, the agentic robotic automated

508
00:28:15.550 --> 00:28:17.709
<v James Loach>part will, will come to pass.

509
00:28:18.449 --> 00:28:20.650
<v James Loach>But I don't know quite the, the time scale
there, yeah.

510
00:28:21.170 --> 00:28:21.449
<v James Loach>No,

511
00:28:21.770 --> 00:28:22.969
<v Speaker 2>no, yeah, exactly.

512
00:28:23.089 --> 00:28:25.489
<v Speaker 2>And, yeah, and the, yeah, time scale's one
thing.

513
00:28:25.689 --> 00:28:28.890
<v Speaker 2>Adoption, and, yeah, and we haven't really
talked about trust as well.

514
00:28:28.969 --> 00:28:31.780
<v Speaker 2>It's actually one of the first questions
in the lightning rounds.

515
00:28:31.969 --> 00:28:36.109
<v Speaker 2>It's actually gonna be quite interesting
around that, cos obviously trust is key

516
00:28:36.109 --> 00:28:40.449
<v Speaker 2>now, like, trusting the output of a
system, but for what you're talking about, which

517
00:28:40.449 --> 00:28:41.530
<v Speaker 2>could be, yeah.

518
00:28:42.089 --> 00:28:45.560
<v Speaker 2>I say it's difficult to put a date on it,
but where you'll talk about tasks being

519
00:28:45.560 --> 00:28:51.189
<v Speaker 2>automatically completed, right, that,
that's a new level of trust required, right?

520
00:28:52.369 --> 00:28:53.239
<v Speaker 2>Beyond where we are now.

521
00:29:04.939 --> 00:29:08.140
<v Speaker 2>So, yeah, so let's, let's move to the, the
lightning round now.

522
00:29:08.219 --> 00:29:09.880
<v Speaker 2>So, like I said, 3 questions.

523
00:29:10.579 --> 00:29:14.719
<v Speaker 2>Past guests have taken different
approaches to this, so, they've either said the one

524
00:29:14.719 --> 00:29:18.979
<v Speaker 2>word answer, which a lightning round
suggests, or they've got into a little bit

525
00:29:18.979 --> 00:29:23.619
<v Speaker 2>more, provided a bit more context, that
word again, provided more context about why

526
00:29:23.619 --> 00:29:24.180
<v Speaker 2>they've answered it.

527
00:29:24.250 --> 00:29:25.380
<v Speaker 2>So, we'll see how we go.

528
00:29:25.459 --> 00:29:27.520
<v Speaker 2>There's no real rules around it, but.

529
00:29:27.930 --> 00:29:34.000
<v Speaker 2>So the first question, so for industrial
AI, which matters most, is it the model?

530
00:29:34.290 --> 00:29:36.410
<v Speaker 2>Is it context, or is it trust?

531
00:29:36.819 --> 00:29:39.579
<v James Loach>I, I had to say, I, I would say it's the
model, right?

532
00:29:40.959 --> 00:29:41.099
<v James Loach>I thought you

533
00:29:41.099 --> 00:29:41.229
<v Speaker 2>might

534
00:29:41.810 --> 00:29:45.300
<v James Loach>what everything else builds upon, but of
course the, the real answer is they all

535
00:29:45.300 --> 00:29:45.699
<v James Loach>matter.

536
00:29:46.180 --> 00:29:46.900
<v James Loach>Yeah, yeah, I know,

537
00:29:46.979 --> 00:29:50.599
<v Speaker 2>I know that was a tough one, but yeah, I
did suspect model might be your, but that,

538
00:29:50.619 --> 00:29:50.780
<v Speaker 2>that's

539
00:29:50.780 --> 00:29:51.810
<v James Loach>the foundation, yeah,

540
00:29:52.630 --> 00:29:52.859
<v Speaker 2>exactly.

541
00:29:53.260 --> 00:29:56.420
<v Speaker 2>And then, so that we've just talked about
this now, but the future of maintenance.

542
00:29:56.569 --> 00:29:59.130
<v Speaker 2>So is it predictive, prescriptive, or
autonomous?

543
00:29:59.219 --> 00:30:00.640
<v Speaker 2>And again, I think I know your answer.

544
00:30:01.680 --> 00:30:03.119
<v James Loach>Yeah, these are all odd things.

545
00:30:03.300 --> 00:30:05.250
<v James Loach>I, yeah, maybe we could say a lot about
this.

546
00:30:05.760 --> 00:30:06.489
<v James Loach>But anyway, what, what is that?

547
00:30:06.540 --> 00:30:08.800
<v James Loach>So ultimately the future is autonomous.

548
00:30:09.579 --> 00:30:10.040
<v James Loach>OK.

549
00:30:10.459 --> 00:30:14.780
<v James Loach>The other things that you talked about in
our world, they're mixed together, I

550
00:30:14.780 --> 00:30:15.119
<v James Loach>think.

551
00:30:15.140 --> 00:30:17.719
<v James Loach>We mainly think of things in terms of.

552
00:30:18.229 --> 00:30:21.910
<v James Loach>Attention management, spotting interesting
things, providing information for people

553
00:30:21.910 --> 00:30:27.380
<v James Loach>to make decisions, and those decisions can
be today and very quick and reactive, or

554
00:30:27.380 --> 00:30:29.709
<v James Loach>they can just automatically be long term.

555
00:30:30.310 --> 00:30:34.239
<v James Loach>So there's a sense in which I think that
division, predictive and prescriptive there

556
00:30:34.239 --> 00:30:38.310
<v James Loach>is a little bit soft in certain ways.

557
00:30:38.719 --> 00:30:39.800
<v James Loach>Yeah, anyway, I'll stop.

558
00:30:39.869 --> 00:30:41.750
<v James Loach>It's not a lightning round style answer,
is it?

559
00:30:42.119 --> 00:30:42.430
<v James Loach>But.

560
00:30:42.780 --> 00:30:46.280
<v Speaker 2>No, no, but don't worry, don't worry, many
before you have chosen a different route.

561
00:30:46.359 --> 00:30:50.189
<v Speaker 2>I'm gonna, I should change it to something
else than lightning round, but, um, no,

562
00:30:50.199 --> 00:30:50.680
<v Speaker 2>it's all good.

563
00:30:50.869 --> 00:30:52.000
<v Speaker 2>Oh, yeah, so the final question.

564
00:30:52.119 --> 00:30:54.060
<v Speaker 2>So one, we're gonna talk about industrial.

565
00:30:54.199 --> 00:30:59.810
<v Speaker 2>So one industrial AI capability that's
overhyped and one that's underrated.

566
00:30:59.920 --> 00:31:02.199
<v James Loach>I don't know if I can answer that question
right.

567
00:31:02.599 --> 00:31:06.880
<v James Loach>Maybe I'm not connected enough with the
hype about in industrial AI things to, to

568
00:31:06.880 --> 00:31:08.260
<v James Loach>have a good sense of that.

569
00:31:08.729 --> 00:31:12.140
<v James Loach>Yeah, maybe the classic answer would be,
it's, it's the same thing for both, right,

570
00:31:12.250 --> 00:31:16.569
<v James Loach>which is agents and then the, the
difference there is on different applications and

571
00:31:16.569 --> 00:31:17.810
<v James Loach>different timescales.

572
00:31:18.089 --> 00:31:23.329
<v James Loach>It, yeah, maybe for certain applications,
it's somehow massively overhyped for what

573
00:31:23.329 --> 00:31:23.920
<v James Loach>it is.

574
00:31:24.390 --> 00:31:27.640
<v James Loach>And then, yeah, other applications
underappreciate how powerful.

575
00:31:28.089 --> 00:31:29.489
<v James Loach>So maybe that's a cop-out answer,

576
00:31:29.650 --> 00:31:29.949
<v Speaker 2>but.

577
00:31:30.369 --> 00:31:30.729
<v Speaker 2>No, it's good.

578
00:31:30.810 --> 00:31:33.910
<v Speaker 2>It's whatever comes into your mind, first
of all, and you haven't been shown these

579
00:31:33.910 --> 00:31:36.530
<v Speaker 2>or shared these before, so I appreciate
that as well.

580
00:31:36.609 --> 00:31:37.189
<v Speaker 2>So that's fine.

581
00:31:37.410 --> 00:31:37.729
<v Speaker 2>That's fine.

582
00:31:37.849 --> 00:31:38.369
<v Speaker 2>That's all good.

583
00:31:38.829 --> 00:31:41.800
<v Speaker 2>OK, so that brings us to the end, I
believe.

584
00:31:41.949 --> 00:31:46.839
<v Speaker 2>Thank you, James, for joining us today and
sharing your, yeah, yeah, sharing your

585
00:31:46.839 --> 00:31:49.079
<v Speaker 2>views on where we are today and into the
future.

586
00:31:49.140 --> 00:31:50.699
<v Speaker 2>It's been great having you on the podcast.

587
00:31:51.359 --> 00:31:52.869
<v James Loach>Yeah, no, it's good, it's nice to talk to
you.

588
00:31:53.079 --> 00:31:53.339
<v James Loach>Yeah.

589
00:31:53.719 --> 00:31:57.640
<v Speaker 2>And I'd like to thank everyone out there,
wherever you are, for listening to this

590
00:31:57.640 --> 00:31:59.189
<v Speaker 2>episode and we'll see you on the next one.

591
00:31:59.439 --> 00:32:00.290
<v Speaker 2>Thank you very much.

592
00:32:01.569 --> 00:32:05.050
<v Speaker 3>Please like and subscribe on all major
podcast channels.

593
00:32:05.130 --> 00:32:10.050
<v Speaker 3>And to find out more about Sensi
predictive maintenance, go to Siemens.com.
