WEBVTT
Language: en-US

1
00:00:02.089 --> 00:00:06.719
<v Jonas Norinder>Welcome to another exciting episode of
Digitalization Tech Talks where we talk about

2
00:00:06.719 --> 00:00:09.949
<v Jonas Norinder>technologies that are game changers for
the process industries.

3
00:00:10.609 --> 00:00:14.329
<v Jonas Norinder>And today we are diving into the
fascinating world of generative AI.

4
00:00:15.409 --> 00:00:19.729
<v Jonas Norinder>Have you ever wondered what generative AI
is and how it can revolutionize

5
00:00:19.729 --> 00:00:21.200
<v Jonas Norinder>manufacturing processes?

6
00:00:21.530 --> 00:00:25.250
<v Jonas Norinder>If you have, you've come to the right
place because that's exactly what we're going

7
00:00:25.250 --> 00:00:29.969
<v Jonas Norinder>to be exploring today, the benefits that
generative AI offers from boosting

8
00:00:29.969 --> 00:00:35.169
<v Jonas Norinder>efficiency to sparking innovation, and
also tackle the challenges that manufacturers

9
00:00:35.169 --> 00:00:36.290
<v Jonas Norinder>might face along the way.

10
00:00:37.040 --> 00:00:41.439
<v Jonas Norinder>We'll zoom in on the process industries
and showcase real-world example of

11
00:00:41.439 --> 00:00:45.549
<v Jonas Norinder>generative AI in action, and that's the
part that I'm particularly excited about.

12
00:00:45.869 --> 00:00:50.200
<v Jonas Norinder>Plus we'll share insights on how
manufacturers can strategically adopt this cutting

13
00:00:50.200 --> 00:00:54.619
<v Jonas Norinder>edge technology, which I know is a topic
that's near and dear to our listeners.

14
00:00:55.540 --> 00:01:00.270
<v Jonas Norinder>As many of you know, we've covered AI in
the past here on this podcast, such as

15
00:01:00.270 --> 00:01:04.870
<v Jonas Norinder>episode 42 where we talked about it in the
context of predictive maintenance, but

16
00:01:04.870 --> 00:01:10.360
<v Jonas Norinder>also episode 37, where we were introduced
to an AI-based chatbot.

17
00:01:10.830 --> 00:01:15.589
<v Jonas Norinder>Today, however, we're going to get more of
an overall view of the value that AI can

18
00:01:15.589 --> 00:01:17.769
<v Jonas Norinder>bring to the table, which is really
awesome.

19
00:01:18.110 --> 00:01:22.470
<v Jonas Norinder>Now, before revealing our guest, I'd like
to have my co-host, Don, share what you're

20
00:01:22.470 --> 00:01:24.989
<v Jonas Norinder>looking forward to the most in today's
conversation.

21
00:01:26.209 --> 00:01:28.040
<v Speaker 2>Hi, Jonas, great to be here again.

22
00:01:28.139 --> 00:01:33.620
<v Speaker 2>And I think what I'm most interested in
learning about today is how and why we've

23
00:01:33.620 --> 00:01:39.819
<v Speaker 2>progressed so quickly from chat GBT just
being released less than 2 years ago to now

24
00:01:39.819 --> 00:01:41.480
<v Speaker 2>being used in the industry.

25
00:01:41.970 --> 00:01:45.819
<v Speaker 2>And as I think anybody in our audience
would agree, technology moves very quickly.

26
00:01:46.059 --> 00:01:52.430
<v Speaker 2>Uh, in my 3-decade career, uh, for an
example, I've seen, Uh, the process industries

27
00:01:52.430 --> 00:01:57.150
<v Speaker 2>has progressed from panel boards and push
buttons as the operator interfaces or

28
00:01:57.150 --> 00:02:03.190
<v Speaker 2>human-machine interfaces, to proprietary
consoles, to client-server-based PC

29
00:02:03.190 --> 00:02:05.839
<v Speaker 2>architectures, to virtualized platforms.

30
00:02:05.879 --> 00:02:09.720
<v Speaker 2>And that seems to be quite a few
significant transitions that have been taking place

31
00:02:09.720 --> 00:02:12.039
<v Speaker 2>over that reasonably short period of time.

32
00:02:12.740 --> 00:02:17.509
<v Speaker 2>But now we're moving on to AI and uh
particularly machine learning and seeing it

33
00:02:17.509 --> 00:02:21.029
<v Speaker 2>work its way into the manufacturing
environment and that's happened just in the past

34
00:02:21.029 --> 00:02:25.389
<v Speaker 2>few years and now we're already expanding
to generative AI that's a really fast

35
00:02:25.389 --> 00:02:29.630
<v Speaker 2>evolution even by typical technology
adoption time frames and that's what I want to

36
00:02:29.630 --> 00:02:30.410
<v Speaker 2>learn about today.

37
00:02:31.529 --> 00:02:35.250
<v Jonas Norinder>Yes, that's a good point, Don, and I'm
looking forward to hearing the answer to that

38
00:02:35.250 --> 00:02:35.949
<v Jonas Norinder>as well.

39
00:02:36.449 --> 00:02:38.410
<v Jonas Norinder>So with that, let's meet today's guest.

40
00:02:38.690 --> 00:02:43.289
<v Jonas Norinder>Guy Brassel is director of business
strategy with the manufacturing industry at

41
00:02:43.289 --> 00:02:43.990
<v Jonas Norinder>Microsoft.

42
00:02:44.130 --> 00:02:48.889
<v Jonas Norinder>In this role, Guy is responsible for
Microsoft's solution strategy in the domain of

43
00:02:48.889 --> 00:02:54.009
<v Jonas Norinder>digital engineering, working with many
important partners in Microsoft's partner

44
00:02:54.009 --> 00:02:56.410
<v Jonas Norinder>ecosystem, including Siemens.

45
00:02:57.029 --> 00:03:01.759
<v Jonas Norinder>The intent is to meet customers where they
are and help them reach their goals and

46
00:03:01.759 --> 00:03:06.460
<v Jonas Norinder>realize the value of cloud and Microsoft
technologies in manufacturing and mobility

47
00:03:06.460 --> 00:03:07.199
<v Jonas Norinder>industries.

48
00:03:07.679 --> 00:03:08.630
<v Jonas Norinder>Welcome to the show, Guy.

49
00:03:08.669 --> 00:03:10.080
<v Jonas Norinder>It's great to have you with us today.

50
00:03:11.149 --> 00:03:12.740
<v Speaker 3>Thank you for inviting me, Jonas.

51
00:03:12.949 --> 00:03:16.869
<v Speaker 3>You know, before I, um, joined Microsoft
almost 4 years ago, I spent many fruitful

52
00:03:16.869 --> 00:03:21.529
<v Speaker 3>years at Siemens, um, working with
customers in many different, uh, industries.

53
00:03:21.750 --> 00:03:24.089
<v Speaker 3>So it's a real pleasure for me to be here
with you today.

54
00:03:24.979 --> 00:03:28.940
<v Jonas Norinder>Well, it's definitely our pleasure, Guy,
and welcome back to Siemens, to the Siemens

55
00:03:28.940 --> 00:03:30.399
<v Jonas Norinder>fold, at least for a little while.

56
00:03:30.940 --> 00:03:34.660
<v Jonas Norinder>So, to get us kicked off here today, why
don't you tell, uh, first, tell us a bit

57
00:03:34.660 --> 00:03:39.179
<v Jonas Norinder>about your role at Microsoft a little bit
more and also briefly tell us about

58
00:03:39.179 --> 00:03:44.179
<v Jonas Norinder>Microsoft's involvement in generative AI,
although I hope that this is not news to

59
00:03:44.179 --> 00:03:45.240
<v Jonas Norinder>any of our listeners.

60
00:03:46.369 --> 00:03:50.729
<v Speaker 3>Sure, um, I'm responsible for the strategy
for one of our four pillars, um, in

61
00:03:50.729 --> 00:03:55.490
<v Speaker 3>manufacturing and mobility, and it's
around unlocking innovation in design and

62
00:03:55.490 --> 00:03:56.270
<v Speaker 3>engineering.

63
00:03:56.649 --> 00:04:00.410
<v Speaker 3>So it's really about ensuring that we have
the right solutions manufacturers are

64
00:04:00.410 --> 00:04:04.929
<v Speaker 3>looking for to enable them to successfully
migrate and adopt design and engineering

65
00:04:04.929 --> 00:04:09.289
<v Speaker 3>functions on the cloud and helping
manufacturers and our partners to adopt the right

66
00:04:09.289 --> 00:04:14.449
<v Speaker 3>technologies such as AI, um, into the
solutions to drive value and the right

67
00:04:14.449 --> 00:04:15.429
<v Speaker 3>business outcomes.

68
00:04:17.950 --> 00:04:18.670
<v Speaker 2>Well, thanks, Scott.

69
00:04:18.779 --> 00:04:22.070
<v Speaker 2>I appreciate that background and it's
great to have you on the podcast today.

70
00:04:22.109 --> 00:04:24.940
<v Speaker 2>And why don't we go ahead and start out
with by grounding everyone with the

71
00:04:24.940 --> 00:04:25.709
<v Speaker 2>definition.

72
00:04:26.109 --> 00:04:27.790
<v Speaker 2>What exactly is generative AI?

73
00:04:30.290 --> 00:04:36.570
<v Speaker 3>So generative AI it it's a type of AI that
is trained on data and can generate novel

74
00:04:36.570 --> 00:04:39.390
<v Speaker 3>content such as text, images, music, code.

75
00:04:39.809 --> 00:04:44.649
<v Speaker 3>You know, what's, what's incredible about
generative AI is that it, it continuously

76
00:04:44.649 --> 00:04:49.369
<v Speaker 3>learns and refines its outputs based on
user interactions.

77
00:04:50.130 --> 00:04:54.529
<v Speaker 3>Um, on the other hand, traditional AI
systems, which have been around for many years

78
00:04:54.529 --> 00:04:59.119
<v Speaker 3>after emerging back in the late 50s, like
um, you know, assembly line automation and

79
00:04:59.119 --> 00:05:04.049
<v Speaker 3>chess playing computers, you know, the,
these make decisions by operating on

80
00:05:04.049 --> 00:05:06.869
<v Speaker 3>explicit instructions and predefined
rules.

81
00:05:07.820 --> 00:05:12.540
<v Speaker 3>Then came machine learning, which is a
subset of AI that enables machines to learn

82
00:05:12.540 --> 00:05:17.700
<v Speaker 3>from existing data and improve upon that
data to make decisions or predictions.

83
00:05:18.309 --> 00:05:23.149
<v Speaker 3>Um, and then that was followed by deep
learning, um, which is a machine learning

84
00:05:23.149 --> 00:05:28.549
<v Speaker 3>technique in which, um, layers of neural
networks are used to process data and make

85
00:05:28.549 --> 00:05:29.019
<v Speaker 3>decisions.

86
00:05:29.429 --> 00:05:34.309
<v Speaker 3>Traditional AI models are not designed
really to, to create original content from

87
00:05:34.309 --> 00:05:38.790
<v Speaker 3>existing data, which, you know, is how,
um, generative AI works.

88
00:05:39.410 --> 00:05:45.890
<v Speaker 3>So generative AI is a form of AI in which
models are trained to generate new content

89
00:05:45.890 --> 00:05:48.410
<v Speaker 3>based on natural language inputs or
prompts.

90
00:05:49.079 --> 00:05:55.029
<v Speaker 3>In other words, you can describe a desired
output in normal everyday language, and

91
00:05:55.029 --> 00:06:00.239
<v Speaker 3>the model can respond by creating
appropriate text, image, code, or other types of

92
00:06:00.239 --> 00:06:01.339
<v Speaker 3>um outputs.

93
00:06:01.799 --> 00:06:07.899
<v Speaker 3>When chat GPT burst on the scene, the
result was generative AI combined with large

94
00:06:07.899 --> 00:06:13.480
<v Speaker 3>language models or LLMs, um, and the
combination of generative AI training with LLMs

95
00:06:13.480 --> 00:06:17.420
<v Speaker 3>has really resulted in AI that has the
ability to think and reason.

96
00:06:18.179 --> 00:06:24.380
<v Speaker 3>So effectively, what we have is a new
natural user interface that's multimodal.

97
00:06:24.540 --> 00:06:30.739
<v Speaker 3>We have memory that retains important
context, recalls both our personal knowledge

98
00:06:30.739 --> 00:06:34.070
<v Speaker 3>and data across our um apps and devices.

99
00:06:34.619 --> 00:06:40.029
<v Speaker 3>We have reasoning and planning
capabilities that help us understand context and

100
00:06:40.029 --> 00:06:44.820
<v Speaker 3>complete really complex tasks whilst
reducing the cognitive load on us.

101
00:06:45.630 --> 00:06:50.170
<v Speaker 3>Generative AI like Copilot, you know, it's
really drastically changed the way we

102
00:06:50.510 --> 00:06:54.269
<v Speaker 3>interact with technology, setting a new
precedent for human innovation.

103
00:06:55.559 --> 00:06:58.480
<v Speaker 2>Well thanks guy that's uh great and seems
that it would open a world of

104
00:06:58.480 --> 00:07:03.399
<v Speaker 2>possibilities and I guess if we look at
manufacturers, um, what do you see as some

105
00:07:03.399 --> 00:07:06.920
<v Speaker 2>of the benefits and also maybe some of the
challenges that manufacturers have

106
00:07:06.920 --> 00:07:08.320
<v Speaker 2>adopting generative AI?

107
00:07:10.230 --> 00:07:14.230
<v Jonas Norinder>We're taking a brief break to make you
aware of Siemens' participation in the Global

108
00:07:14.230 --> 00:07:18.070
<v Jonas Norinder>hydrogen conference, which takes place
virtually on November 20th.

109
00:07:18.839 --> 00:07:23.320
<v Jonas Norinder>Siemens will be giving the keynote speech
at the event, which is titled Optimizing

110
00:07:23.320 --> 00:07:26.760
<v Jonas Norinder>Green Hydrogen Production with an
Operational Digital twin.

111
00:07:27.480 --> 00:07:31.839
<v Jonas Norinder>So if you're interested in hydrogen and
the digital twin, please register for the

112
00:07:31.839 --> 00:07:32.190
<v Jonas Norinder>event.

113
00:07:32.320 --> 00:07:35.500
<v Jonas Norinder>We will include a link to the event in the
show notes.

114
00:07:38.799 --> 00:07:43.059
<v Speaker 3>Manufacturers are facing unprecedented
demands to achieve greater efficiency.

115
00:07:43.239 --> 00:07:49.220
<v Speaker 3>While not a new concept, transformation
now depends on how organizations leverage

116
00:07:49.220 --> 00:07:53.899
<v Speaker 3>the data generated from their
manufacturing assets, processes, and people.

117
00:07:54.600 --> 00:08:01.559
<v Speaker 3>Today, manufacturing IOT, cloud, and AI
solutions can enable every industry to scale

118
00:08:01.559 --> 00:08:04.559
<v Speaker 3>innovation for a greater insight and
impact.

119
00:08:05.339 --> 00:08:11.720
<v Speaker 3>Within manufacturing, AI plays a pivotal
role in this change, connecting operational

120
00:08:11.859 --> 00:08:13.559
<v Speaker 3>or OT systems.

121
00:08:14.279 --> 00:08:19.869
<v Speaker 3>Information technology or IT systems and
engineering technologies, ET across the

122
00:08:19.869 --> 00:08:24.760
<v Speaker 3>entire value chain, streamlining how
products are designed and built, as well as

123
00:08:24.760 --> 00:08:27.760
<v Speaker 3>optimizing the long-term operation of
plants and facilities.

124
00:08:28.589 --> 00:08:32.309
<v Speaker 3>Many manufacturers are turning to AI to
help fill the skills gap where they're

125
00:08:32.309 --> 00:08:36.580
<v Speaker 3>struggling to hire engineers with the
right technical skills for the tools they need

126
00:08:36.580 --> 00:08:41.340
<v Speaker 3>to use, you know, gain productivity, uh,
efficiencies and help in unlocking deeper

127
00:08:41.340 --> 00:08:44.400
<v Speaker 3>insights to make better informed
decisions.

128
00:08:45.210 --> 00:08:51.159
<v Speaker 3>One of the main challenges in scaling AI
use cases is really that deficit of uh

129
00:08:51.159 --> 00:08:53.010
<v Speaker 3>talent, skills and data.

130
00:08:53.130 --> 00:08:57.090
<v Speaker 3>We see, you know, manufacturers cite this
as their one of their toughest challenges

131
00:08:57.090 --> 00:09:02.489
<v Speaker 3>in both design and engineering as well as
plant operations, and inadequate data

132
00:09:02.489 --> 00:09:06.440
<v Speaker 3>quality and governance also hamper um
these use case developments.

133
00:09:06.690 --> 00:09:11.690
<v Speaker 3>Additionally, insufficient access to
cloud-based compute power is another constraint

134
00:09:11.690 --> 00:09:13.150
<v Speaker 3>in design and engineering.

135
00:09:13.510 --> 00:09:18.669
<v Speaker 3>AI model governance is also one of the big
challenges, and this really underscores

136
00:09:18.669 --> 00:09:24.190
<v Speaker 3>the difficulties in maintaining the high
standards of data quality, consistency and

137
00:09:24.190 --> 00:09:28.679
<v Speaker 3>compliance within, you know, what is a
very complex and variable manufacturing

138
00:09:28.679 --> 00:09:29.530
<v Speaker 3>landscape.

139
00:09:29.950 --> 00:09:35.469
<v Speaker 3>Scaling AI use cases can be hindered
without solid data foundations, so that is

140
00:09:35.469 --> 00:09:41.789
<v Speaker 3>having production assets with data
prepared for utilization in current AI models.

141
00:09:42.219 --> 00:09:46.650
<v Speaker 3>To effectively scale AI data fragmentation
must be minimized.

142
00:09:46.950 --> 00:09:51.500
<v Speaker 3>Most manufacturers will need to update
their data architecture, infrastructure and

143
00:09:51.500 --> 00:09:55.950
<v Speaker 3>processes to gain the full benefits from
AI as well as other technological and

144
00:09:55.950 --> 00:09:57.210
<v Speaker 3>business objectives.

145
00:09:57.880 --> 00:10:03.159
<v Speaker 3>Prioritizing a modernization strategy that
enhances interoperability between

146
00:10:03.159 --> 00:10:08.820
<v Speaker 3>engineering and design and plant data
systems as well as between operational

147
00:10:08.820 --> 00:10:12.719
<v Speaker 3>technology, OT and information technology,
IT systems is really crucial.

148
00:10:13.070 --> 00:10:18.469
<v Speaker 3>Generative AI can help manufacturers in so
many ways, from accelerating product R&amp;D

149
00:10:18.469 --> 00:10:23.750
<v Speaker 3>and design to, You know, speeding up new
product introduction, predictive

150
00:10:23.750 --> 00:10:29.190
<v Speaker 3>maintenance and customer uh field service
to real-time use of digital twins.

151
00:10:30.450 --> 00:10:35.559
<v Jonas Norinder>Yeah, so as you, as you underscored here,
Guy, there's a lot of ways that generative

152
00:10:35.559 --> 00:10:40.210
<v Jonas Norinder>AI can help, but there are some specific
steps that must be taken to make it a

153
00:10:40.210 --> 00:10:40.409
<v Jonas Norinder>success.

154
00:10:40.460 --> 00:10:43.409
<v Jonas Norinder>So thank you for laying out those steps
for our listeners.

155
00:10:44.239 --> 00:10:48.320
<v Jonas Norinder>Now, we talked about the benefits of AI
for manufacturers in general, but I'd like

156
00:10:48.320 --> 00:10:52.140
<v Jonas Norinder>to dig a little bit deeper into the
process industries since that's the primary

157
00:10:52.140 --> 00:10:53.140
<v Jonas Norinder>focus of this podcast.

158
00:10:53.570 --> 00:10:58.520
<v Jonas Norinder>So, can you talk more specifically about
the benefits for industries such as pharma,

159
00:10:58.789 --> 00:11:02.559
<v Jonas Norinder>pharmaceuticals, steel, cement, chemicals,
and other process industries?

160
00:11:03.710 --> 00:11:08.679
<v Speaker 3>Yeah, the, the manufacturing industry
stands on the brink of a transformative era,

161
00:11:08.989 --> 00:11:12.119
<v Speaker 3>um, really propelled by generative AI.

162
00:11:12.349 --> 00:11:17.750
<v Speaker 3>AI is not just a tool for automation, it's
a catalyst for innovation, efficiency,

163
00:11:17.760 --> 00:11:18.780
<v Speaker 3>and sustainability.

164
00:11:19.919 --> 00:11:24.960
<v Speaker 3>AI can significantly enhance product and
process innovation, reduce cycle times and

165
00:11:24.960 --> 00:11:29.900
<v Speaker 3>improve maintenance and security, whilst
also helping reduce carbon emissions.

166
00:11:30.429 --> 00:11:36.109
<v Speaker 3>Generative AI allows for faster research
and development, uh, resulting in, uh,

167
00:11:36.119 --> 00:11:40.109
<v Speaker 3>shorter timelines and time to market, as
well as reducing costs.

168
00:11:40.580 --> 00:11:44.419
<v Speaker 3>Scientists and engineers can come up with
the new hypotheses faster.

169
00:11:45.539 --> 00:11:51.979
<v Speaker 3>It enables materials optimization and is
revolutionizing the way manufacturers are

170
00:11:51.979 --> 00:11:53.570
<v Speaker 3>developing new formulations.

171
00:11:54.059 --> 00:11:58.460
<v Speaker 3>Process manufacturers with all of this
data can learn from other industries such as,

172
00:11:58.549 --> 00:12:02.570
<v Speaker 3>you know, from metals used in aerospace
or, uh, even automotive.

173
00:12:03.059 --> 00:12:08.090
<v Speaker 3>A model trained on a vast library of
chemical compounds, for example, helps

174
00:12:08.090 --> 00:12:12.679
<v Speaker 3>researchers identify new molecules
exponentially faster than before.

175
00:12:13.359 --> 00:12:17.299
<v Speaker 3>Which improves product performance and
sustainability.

176
00:12:18.280 --> 00:12:22.960
<v Speaker 3>One area of materials development you can
see this impact in is EV battery material

177
00:12:22.960 --> 00:12:25.179
<v Speaker 3>science is uh I think a really good
example.

178
00:12:26.479 --> 00:12:30.479
<v Speaker 3>Predictive maintenance and supply chain
optimization are now possible, uh, resulting

179
00:12:30.479 --> 00:12:35.669
<v Speaker 3>in reduced downtime and more efficient
logistics, um, areas really that are critical

180
00:12:35.669 --> 00:12:37.729
<v Speaker 3>to, uh, process manufacturers.

181
00:12:38.469 --> 00:12:43.020
<v Speaker 3>AI has the potential to help address the
industry's growing demand for novel

182
00:12:43.510 --> 00:12:47.919
<v Speaker 3>materials and environmentally friendly
practices in a more cost-effective manner.

183
00:12:49.969 --> 00:12:54.690
<v Speaker 3>One interesting scenario I saw earlier
this year was a chemicals manufacturer who

184
00:12:54.690 --> 00:12:58.969
<v Speaker 3>was using generative AI to deconstruct
highly complex contracts.

185
00:12:59.140 --> 00:13:03.859
<v Speaker 3>So using generative AI to query and
extract key information from volumes and volumes

186
00:13:03.859 --> 00:13:08.419
<v Speaker 3>of complex, highly detailed, technical,
and, you know, very long documents.

187
00:13:09.070 --> 00:13:13.549
<v Speaker 3>Um, the model was trained on a broad
spectrum of contractual legal compliance data,

188
00:13:13.669 --> 00:13:19.510
<v Speaker 3>so, you know, when we say deconstructing,
um, generative AI knows the context, the

189
00:13:19.510 --> 00:13:23.429
<v Speaker 3>language and phrases, and the technical
jargon used in that industry, the

190
00:13:23.429 --> 00:13:26.369
<v Speaker 3>contractual language and the compliance
terminology.

191
00:13:27.710 --> 00:13:31.630
<v Speaker 3>Chemicals and process manufacturers face a
myriad of challenges, you know, really

192
00:13:31.630 --> 00:13:38.210
<v Speaker 3>needing to, to balance sustainability,
regulatory compliance and um ESG priorities.

193
00:13:38.390 --> 00:13:42.179
<v Speaker 3>At the same time, they're pursuing revenue
growth in markets with greater

194
00:13:42.179 --> 00:13:46.710
<v Speaker 3>competition and slow sales growth, you
know, which really puts pressure on prices.

195
00:13:47.690 --> 00:13:52.580
<v Speaker 3>This means that companies have to innovate
to survive, and this is where I would say

196
00:13:52.580 --> 00:13:57.460
<v Speaker 3>we have some of the biggest opportunities
and benefits that arise from generative AI

197
00:13:58.059 --> 00:14:03.130
<v Speaker 3>so we can think about accelerating R&amp;D
cycles, accelerating sales cycles, improving

198
00:14:03.130 --> 00:14:08.219
<v Speaker 3>regulatory and compliance, uh,
traceability, and really thinking about meeting

199
00:14:08.219 --> 00:14:09.539
<v Speaker 3>sustainability goals.

200
00:14:11.080 --> 00:14:15.969
<v Speaker 2>So Guy, there are many benefits to be
gained that you mentioned, and they hit on a

201
00:14:15.969 --> 00:14:18.469
<v Speaker 2>lot of the industry drivers in the process
industries.

202
00:14:19.049 --> 00:14:22.770
<v Speaker 2>You also in your last answer, you covered
a couple of examples to demonstrate this

203
00:14:23.130 --> 00:14:25.049
<v Speaker 2>and interested in hearing about some
others.

204
00:14:25.130 --> 00:14:29.289
<v Speaker 2>What are other examples of manufacturers
in the process industries using generative

205
00:14:29.289 --> 00:14:29.669
<v Speaker 2>AI?

206
00:14:31.729 --> 00:14:37.090
<v Speaker 3>Wow, there are so many to choose from, but
I think, um, here's maybe a few that um

207
00:14:37.549 --> 00:14:40.750
<v Speaker 3>that personally excite me and I, I think
will resonate with our listeners.

208
00:14:40.849 --> 00:14:46.719
<v Speaker 3>So firstly, let's start with um uh product
life cycle management PLM that's

209
00:14:46.719 --> 00:14:51.229
<v Speaker 3>integrated with Microsoft Teams as a
collaboration medium producing a new generation

210
00:14:51.229 --> 00:14:57.169
<v Speaker 3>of uh apps and new ways to access PLM that
was powered by um Azure OpenAI.

211
00:14:58.020 --> 00:15:01.409
<v Speaker 3>So many issues in the plan go unreported.

212
00:15:01.789 --> 00:15:07.890
<v Speaker 3>Um, and so this is an ideal way to capture
these types of issues, um, at any stage

213
00:15:07.890 --> 00:15:08.859
<v Speaker 3>in the life cycle.

214
00:15:09.309 --> 00:15:14.859
<v Speaker 3>For many PLM users, the mobile nature of
their jobs means they spend most of their

215
00:15:14.859 --> 00:15:20.559
<v Speaker 3>time away from a computer, especially
workers on the plant floor or in service

216
00:15:20.559 --> 00:15:20.909
<v Speaker 3>operations.

217
00:15:21.419 --> 00:15:27.659
<v Speaker 3>With the PLM Teams integration, for
example, where you're connecting mobile workers

218
00:15:27.659 --> 00:15:32.940
<v Speaker 3>to the enterprise, you're providing easy
access and greater collaboration across the

219
00:15:32.940 --> 00:15:39.020
<v Speaker 3>life cycle, leveraging the latest advances
in generative AI and natural language

220
00:15:39.020 --> 00:15:45.419
<v Speaker 3>processing, these PLM apps, Uh, utilize a
natural language first

221
00:15:45.419 --> 00:15:46.179
<v Speaker 3>interface.

222
00:15:46.479 --> 00:15:51.340
<v Speaker 3>Frontline workers can connect with PLM
anywhere and simply speaking in their native

223
00:15:51.340 --> 00:15:56.380
<v Speaker 3>language, they can easily capture
equipment, product, or manufacturing issues from

224
00:15:56.380 --> 00:15:57.190
<v Speaker 3>the plant floor.

225
00:15:57.799 --> 00:16:01.280
<v Speaker 3>With speech, text, images, etc.

226
00:16:01.559 --> 00:16:06.479
<v Speaker 3>and they can close the loop with the
engineering teams who can solve those

227
00:16:06.479 --> 00:16:10.559
<v Speaker 3>particular issues, provide the fixes and
allow the engineer on the plant floor to

228
00:16:10.559 --> 00:16:13.320
<v Speaker 3>then validate the issue is fixed.

229
00:16:15.080 --> 00:16:21.080
<v Speaker 3>Another set of examples are industrial AI
powered assistants or co-pilots, which are

230
00:16:21.080 --> 00:16:25.340
<v Speaker 3>helping to boost productivity and
efficiency across the industrial life cycle.

231
00:16:25.719 --> 00:16:30.599
<v Speaker 3>These co-pilots help to guide automation
engineers through complex tasks like

232
00:16:30.599 --> 00:16:37.299
<v Speaker 3>writing and optimizing code for machines,
uh, PLCs, and really helping operators in,

233
00:16:37.320 --> 00:16:39.239
<v Speaker 3>uh, troubleshooting errors.

234
00:16:39.950 --> 00:16:44.049
<v Speaker 3>We've seen huge improvements in
productivity and software development with co-pilots

235
00:16:44.049 --> 00:16:48.109
<v Speaker 3>in the automotive industry, for example,
and these are now being transferred to

236
00:16:48.109 --> 00:16:48.909
<v Speaker 3>machine programming.

237
00:16:49.950 --> 00:16:54.090
<v Speaker 3>This is also helping to address the
shortage of skilled workers by automating

238
00:16:54.090 --> 00:16:56.630
<v Speaker 3>tedious, repetitive tasks.

239
00:16:56.789 --> 00:17:02.349
<v Speaker 3>I think uh a lot of automation code is
pretty standard code that is repeatedly used

240
00:17:02.349 --> 00:17:07.150
<v Speaker 3>but needs to be programmed from scratch
each time, which results in a lot of wasted

241
00:17:07.150 --> 00:17:07.609
<v Speaker 3>time.

242
00:17:08.140 --> 00:17:13.670
<v Speaker 3>Another great example of this outcome is
generative engineering using generative AI

243
00:17:13.670 --> 00:17:14.349
<v Speaker 3>in CAD.

244
00:17:14.968 --> 00:17:19.399
<v Speaker 3>We saw earlier this year this type of AI
driven innovation that can be applied to

245
00:17:19.399 --> 00:17:24.168
<v Speaker 3>the creation of electrical connectors
based on the inputs of the needs described

246
00:17:24.168 --> 00:17:26.529
<v Speaker 3>through the copilot or natural language
interface.

247
00:17:27.449 --> 00:17:32.099
<v Speaker 3>And I think one of the areas of AI
adoption that really excites me the most is in

248
00:17:32.099 --> 00:17:38.219
<v Speaker 3>the simulation and analysis space where
this is, you know, super important for

249
00:17:38.219 --> 00:17:39.369
<v Speaker 3>process manufacturers.

250
00:17:39.859 --> 00:17:44.540
<v Speaker 3>Um, within the simulation application
software, physics informed neural nets are

251
00:17:44.540 --> 00:17:46.619
<v Speaker 3>trained on physics simulations.

252
00:17:47.160 --> 00:17:51.300
<v Speaker 3>This is typically called AI physics, you
know, examples of this include

253
00:17:51.300 --> 00:17:54.380
<v Speaker 3>accelerating, uh, simulation workflows.

254
00:17:54.839 --> 00:18:00.020
<v Speaker 3>Building proprietary physics inference
models and modeling complex physics

255
00:18:00.020 --> 00:18:00.660
<v Speaker 3>phenomenon.

256
00:18:01.510 --> 00:18:03.609
<v Speaker 3>You can also think of molecular models
here.

257
00:18:03.709 --> 00:18:09.140
<v Speaker 3>On engineering data, LLMs are fine-tuned
to engineering data, allowing engineers to

258
00:18:09.140 --> 00:18:14.910
<v Speaker 3>query knowledge bases or company data,
certification or regulatory data.

259
00:18:15.430 --> 00:18:19.910
<v Speaker 3>And then machine learning or deep
learning, you know, is also being combined to

260
00:18:19.910 --> 00:18:24.910
<v Speaker 3>ensure efficient usage of high performance
compute and AI infrastructure when

261
00:18:24.910 --> 00:18:30.270
<v Speaker 3>running these simulations with dynamic
hardware, um, performance and forecasting.

262
00:18:31.160 --> 00:18:36.890
<v Speaker 3>AI in simulation is, it's not a single
application of any one AI technology, but

263
00:18:36.890 --> 00:18:43.329
<v Speaker 3>it's a, it's a confluence of AI
technologies that, you know, deliver value that is

264
00:18:43.329 --> 00:18:44.910
<v Speaker 3>larger than the sum of the whole.

265
00:18:45.880 --> 00:18:51.250
<v Speaker 3>And then one last example I wanted to uh
to also bring up here is one that I see

266
00:18:51.250 --> 00:18:54.349
<v Speaker 3>huge potential in, especially for process
manufacturers.

267
00:18:54.930 --> 00:19:00.329
<v Speaker 3>Much of the technical data across process
um industries is hard copy, printed or it

268
00:19:00.329 --> 00:19:02.229
<v Speaker 3>exists in PDF documents.

269
00:19:02.810 --> 00:19:06.489
<v Speaker 3>I've met with a number of manufacturers
across process industries who are looking at

270
00:19:06.489 --> 00:19:07.290
<v Speaker 3>this challenge.

271
00:19:07.709 --> 00:19:12.270
<v Speaker 3>And it's one that generative AI can be a
solution to, you know, imagine being able

272
00:19:12.270 --> 00:19:18.430
<v Speaker 3>to take legacy piping and instrumentation
diagrams, or commonly known as PNIDs in

273
00:19:18.430 --> 00:19:23.709
<v Speaker 3>the process industries, which represent
the design, modification and maintenance of

274
00:19:23.709 --> 00:19:30.109
<v Speaker 3>an engineering process and use generative
AI to interrogate, validate these, uh,

275
00:19:30.130 --> 00:19:33.689
<v Speaker 3>technical schematics or automate, uh, the
creation.

276
00:19:34.000 --> 00:19:40.150
<v Speaker 3>Of PNID schematics or even drive
interaction with a digital twin through a co-pilot

277
00:19:40.150 --> 00:19:46.520
<v Speaker 3>natural language interface based on a PNID
schematic, the value of digitalization

278
00:19:46.520 --> 00:19:49.849
<v Speaker 3>and generative AI just really cannot be
understated.

279
00:19:50.630 --> 00:19:55.469
<v Speaker 2>Well, guys, a lot of good examples and,
and use cases that we come across every day

280
00:19:55.469 --> 00:19:57.369
<v Speaker 2>in, in the process industries.

281
00:19:58.109 --> 00:20:01.310
<v Speaker 2>I'm guessing that some in our audience are
going to hear this and they're gonna want

282
00:20:01.310 --> 00:20:02.729
<v Speaker 2>to go ahead and get started.

283
00:20:03.310 --> 00:20:07.430
<v Speaker 2>Can you talk about some initial steps that
the manufacturers in the process

284
00:20:07.430 --> 00:20:11.160
<v Speaker 2>industries can take when they're adopting
generative of AI to get off to a good

285
00:20:11.160 --> 00:20:11.569
<v Speaker 2>start?

286
00:20:13.689 --> 00:20:18.489
<v Speaker 3>As, as manufacturers in all industries
worldwide progress in their digital

287
00:20:18.489 --> 00:20:24.310
<v Speaker 3>transformation journeys, um, including
process manufacturers, you know, data serves

288
00:20:24.729 --> 00:20:26.349
<v Speaker 3>as the crucial enabler.

289
00:20:26.729 --> 00:20:31.719
<v Speaker 3>To harness the full potential of AI, it is
imperative to establish and maintain data

290
00:20:31.719 --> 00:20:37.329
<v Speaker 3>quality, integration and governance
throughout the entire value chain, extending

291
00:20:37.329 --> 00:20:41.790
<v Speaker 3>from design and production all the way to
distribution and field service.

292
00:20:42.420 --> 00:20:47.959
<v Speaker 3>Bridging the divide between OT and IT
systems which frequently function in isolation

293
00:20:47.959 --> 00:20:54.459
<v Speaker 3>and adhere to, you know, disparate, uh,
protocols and standards really is essential.

294
00:20:54.780 --> 00:20:59.329
<v Speaker 3>As manufacturers continue to tap into the
transformative power of AI, uh, there are

295
00:20:59.329 --> 00:21:04.020
<v Speaker 3>probably, you know, several considerations
I would recommend to take into account.

296
00:21:04.790 --> 00:21:06.609
<v Speaker 3>So first, start with the business case.

297
00:21:07.530 --> 00:21:13.650
<v Speaker 3>Companies should begin by identifying the
business problem or challenge that

298
00:21:13.660 --> 00:21:15.650
<v Speaker 3>technology, including AI could address.

299
00:21:15.770 --> 00:21:19.319
<v Speaker 3>Only then should technology solutions be
considered.

300
00:21:19.770 --> 00:21:24.810
<v Speaker 3>There are so many great ideas, but to get
real impact starts with the actual

301
00:21:24.810 --> 00:21:25.609
<v Speaker 3>business need.

302
00:21:26.800 --> 00:21:29.800
<v Speaker 3>Second, define policy and governance.

303
00:21:29.920 --> 00:21:36.719
<v Speaker 3>As as AI plays a larger role in our daily
lives, it's more important than ever that

304
00:21:36.719 --> 00:21:42.719
<v Speaker 3>AI systems are built to provide a helpful,
safe, and trustworthy experience for

305
00:21:42.719 --> 00:21:42.939
<v Speaker 3>everyone.

306
00:21:43.829 --> 00:21:49.150
<v Speaker 3>This is why Microsoft develops and deploys
technology using responsible AI

307
00:21:49.150 --> 00:21:54.790
<v Speaker 3>practices, and responsible AI intends to
keep people and their goals at the center

308
00:21:54.790 --> 00:22:01.469
<v Speaker 3>of the design process and considers the
benefits and potential harms that AI systems

309
00:22:01.469 --> 00:22:03.109
<v Speaker 3>can have on society.

310
00:22:03.680 --> 00:22:09.920
<v Speaker 3>Um, and it's really around 6 key
principles of, you know, fairness, reliability and

311
00:22:09.920 --> 00:22:16.349
<v Speaker 3>safety, privacy, security, inclusiveness,
transparency, and finally accountability.

312
00:22:17.430 --> 00:22:22.989
<v Speaker 3>Our work is, is guided by these core set
of six responsible AI principles, and we

313
00:22:22.989 --> 00:22:28.109
<v Speaker 3>take a cross-company approach through
cutting edge research, best of breed

314
00:22:28.109 --> 00:22:32.589
<v Speaker 3>engineering systems, and excellence in
policy and governance, really forming the

315
00:22:32.589 --> 00:22:38.630
<v Speaker 3>basis of the, uh, the responsible AI
standard, which was, you know, this was

316
00:22:38.630 --> 00:22:40.390
<v Speaker 3>introduced back in 2019.

317
00:22:41.640 --> 00:22:45.670
<v Speaker 3>Thirdly, embrace structural flexibility.

318
00:22:45.930 --> 00:22:50.810
<v Speaker 3>In this evolving landscape, use case
development extends beyond the realm of AI

319
00:22:50.810 --> 00:22:51.369
<v Speaker 3>experts.

320
00:22:51.449 --> 00:22:57.410
<v Speaker 3>It has become a multidisciplinary endeavor
as internal expertise is harnessed

321
00:22:57.410 --> 00:23:00.689
<v Speaker 3>alongside data science and engineering
teams in.

322
00:23:01.349 --> 00:23:07.589
<v Speaker 3>You know, collaboration with uh with
business, um, product owners and um IT

323
00:23:07.589 --> 00:23:08.290
<v Speaker 3>professionals.

324
00:23:09.959 --> 00:23:13.750
<v Speaker 3>Fourth Prepare your data foundation.

325
00:23:14.020 --> 00:23:17.739
<v Speaker 3>You know, AI thrives in a mature data
environment.

326
00:23:17.930 --> 00:23:23.300
<v Speaker 3>To achieve this, manufacturers can
strategically adopt modular updates and employ,

327
00:23:23.739 --> 00:23:28.699
<v Speaker 3>you know, data virtualization techniques
to incrementally enhance their systems and

328
00:23:28.699 --> 00:23:32.359
<v Speaker 3>align their data architecture with AI
capabilities.

329
00:23:33.189 --> 00:23:38.400
<v Speaker 3>It's important to assess your
organisation's data collection, storage and processing

330
00:23:38.400 --> 00:23:45.319
<v Speaker 3>capabilities, addressing any identified
weaknesses before deploying AI use

331
00:23:45.319 --> 00:23:45.859
<v Speaker 3>cases.

332
00:23:46.910 --> 00:23:52.380
<v Speaker 3>While AI models necessitate high quality
data, perfection is not really a

333
00:23:52.380 --> 00:23:55.469
<v Speaker 3>prerequisite for moving use cases into
production.

334
00:23:56.479 --> 00:24:02.390
<v Speaker 3>And the last consideration is, you know,
think about leveraging AI for skill

335
00:24:02.390 --> 00:24:02.910
<v Speaker 3>development.

336
00:24:03.119 --> 00:24:07.719
<v Speaker 3>It's natural for manufacturers to be
concerned about the lack of AI skills and

337
00:24:07.719 --> 00:24:14.079
<v Speaker 3>talent, but they should recognize, uh,
that AI can aid in cultivating these

338
00:24:14.079 --> 00:24:15.819
<v Speaker 3>skills within their workforce.

339
00:24:17.400 --> 00:24:21.150
<v Jonas Norinder>That was some really great advice, Guy,
and I like how you covered, you know,

340
00:24:21.319 --> 00:24:24.920
<v Jonas Norinder>different things, different aspects of it,
especially around the people development,

341
00:24:24.959 --> 00:24:26.959
<v Jonas Norinder>the skill development, the data
foundation.

342
00:24:27.030 --> 00:24:31.839
<v Jonas Norinder>It seems like there is some foundational
things that need to be in order before you

343
00:24:31.839 --> 00:24:34.040
<v Jonas Norinder>want to start going down that AI path.

344
00:24:34.109 --> 00:24:38.119
<v Jonas Norinder>But also, as you mentioned, it does not
have to be perfect to start using it.

345
00:24:38.239 --> 00:24:41.439
<v Jonas Norinder>So I thought those were some really,
really good points and great conversation

346
00:24:41.439 --> 00:24:41.819
<v Jonas Norinder>today.

347
00:24:42.239 --> 00:24:46.760
<v Jonas Norinder>That was really insightful and I'm sure
our listeners felt the same way.

348
00:24:47.400 --> 00:24:51.359
<v Jonas Norinder>As we're nearing the end of our
conversation today, do you have a few final key

349
00:24:51.359 --> 00:24:55.400
<v Jonas Norinder>takeaways related to generative AI that
you'd like to leave our listeners with?

350
00:24:55.439 --> 00:24:57.150
<v Jonas Norinder>You've covered a lot today.

351
00:24:57.520 --> 00:25:02.359
<v Jonas Norinder>Um, so maybe just highlight a couple of
those really key points that you want people

352
00:25:02.359 --> 00:25:03.459
<v Jonas Norinder>to, to walk away with.

353
00:25:05.339 --> 00:25:10.189
<v Speaker 3>Thanks Jonas, uh, sure, um, I, I'll wrap
with three takeaways.

354
00:25:10.510 --> 00:25:17.040
<v Speaker 3>First, AI and generative AI is being made
more readily accessible with cloud-based

355
00:25:17.040 --> 00:25:21.829
<v Speaker 3>services, that makes it easier for
developers of all skill levels to build various

356
00:25:21.829 --> 00:25:23.119
<v Speaker 3>AI applications.

357
00:25:24.180 --> 00:25:29.439
<v Speaker 3>Uh, Second, yeah, you can readily infuse
your solutions with AI services.

358
00:25:29.569 --> 00:25:36.420
<v Speaker 3>Think about speech, language, and
translation, vision, documentation, content

359
00:25:36.420 --> 00:25:40.599
<v Speaker 3>safety, you know, there's, there's so many
more, uh, options that are available.

360
00:25:40.780 --> 00:25:41.939
<v Speaker 3>And then finally, third.

361
00:25:42.500 --> 00:25:48.619
<v Speaker 3>Model catalogs are making an extensive
number of large and small models available to

362
00:25:48.619 --> 00:25:53.060
<v Speaker 3>you so you can select from a diverse set
of models with different capabilities

363
00:25:53.060 --> 00:25:54.939
<v Speaker 3>really to suit your scenarios.

364
00:25:56.819 --> 00:25:57.270
<v Jonas Norinder>Awesome.

365
00:25:57.349 --> 00:26:01.390
<v Jonas Norinder>I'd like the way that you just distilled
this very long conversation into those

366
00:26:01.390 --> 00:26:02.489
<v Jonas Norinder>three takeaways.

367
00:26:02.569 --> 00:26:06.869
<v Jonas Norinder>I want to thank you for being our guest on
the Digitalization TED Talks podcast

368
00:26:06.869 --> 00:26:07.180
<v Jonas Norinder>today.

369
00:26:07.469 --> 00:26:11.750
<v Jonas Norinder>We really appreciate it that you took your
time to be here and we enjoyed having you

370
00:26:11.750 --> 00:26:12.030
<v Jonas Norinder>here.

371
00:26:14.030 --> 00:26:17.329
<v Speaker 3>My pleasure and thank you again Jonas for
er for having me.

372
00:26:18.150 --> 00:26:20.670
<v Jonas Norinder>With that, we are at the end of this
episode.

373
00:26:20.869 --> 00:26:25.510
<v Jonas Norinder>If you have any further questions about
generative AI as it relates to the process

374
00:26:25.510 --> 00:26:29.579
<v Jonas Norinder>industries and other areas, please feel
free to reach out to Guy directly.

375
00:26:29.910 --> 00:26:34.550
<v Jonas Norinder>We did include his email address in the
show notes, so you can go look there to find

376
00:26:34.550 --> 00:26:35.770
<v Jonas Norinder>it and send him an email.

377
00:26:36.310 --> 00:26:41.550
<v Jonas Norinder>And as usual, Don and I are always open to
any comments you may have, any thoughts

378
00:26:41.550 --> 00:26:46.869
<v Jonas Norinder>and feedback on the show or on the topics,
and you can of course also find our

379
00:26:46.869 --> 00:26:48.910
<v Jonas Norinder>contact information in the show notes.

380
00:26:48.989 --> 00:26:53.589
<v Jonas Norinder>We'd love to hear from you, especially if
you have any great ideas for guests on the

381
00:26:53.589 --> 00:26:54.849
<v Jonas Norinder>show or topics.

382
00:26:55.770 --> 00:26:59.319
<v Jonas Norinder>If you liked what you heard today, please
help spread the word by rating the show

383
00:26:59.319 --> 00:27:04.819
<v Jonas Norinder>and if you're not already a subscriber and
would like to be notified of new episodes

384
00:27:04.819 --> 00:27:07.479
<v Jonas Norinder>as they are released, please subscribe to
the series.

385
00:27:07.979 --> 00:27:12.540
<v Jonas Norinder>So thank you very much for listening to
another episode of Digitalisation Tech

386
00:27:12.540 --> 00:27:12.819
<v Speaker 4>Talk.
