WEBVTT
Language: en-US

1
00:00:09.670 --> 00:00:11.890
<v Spencer Acain>Hello and welcome to the AI Spectrum
podcast.

2
00:00:12.550 --> 00:00:13.920
<v Spencer Acain>I'm your host, Spencer Kane.

3
00:00:14.310 --> 00:00:18.350
<v Spencer Acain>In this series, we explore a wide range of
AI topics from all across Siemens and how

4
00:00:18.350 --> 00:00:19.989
<v Spencer Acain>they apply to different technologies.

5
00:00:20.350 --> 00:00:24.020
<v Spencer Acain>Today I'm joined by Christopher Shutte,
senior Product portfolio Manager for

6
00:00:24.020 --> 00:00:26.459
<v Spencer Acain>robotics AI at Siemens Digital-Industries.

7
00:00:26.709 --> 00:00:27.770
<v Spencer Acain>Welcome, Christopher.

8
00:00:28.309 --> 00:00:29.100
<v Christopher Schuette>Thanks, Spencer.

9
00:00:29.270 --> 00:00:31.690
<v Christopher Schuette>Great to be here and hello to all
listeners.

10
00:00:32.229 --> 00:00:36.470
<v Spencer Acain>To start things off, uh Christopher, could
you tell me a bit more about yourself and

11
00:00:36.470 --> 00:00:37.549
<v Spencer Acain>your work at Siemens?

12
00:00:38.250 --> 00:00:38.840
<v Christopher Schuette>Yeah, for sure.

13
00:00:39.009 --> 00:00:43.770
<v Christopher Schuette>Um, so I'm working for Siemens uh quite a
while, um, starting directly after my

14
00:00:43.770 --> 00:00:44.970
<v Christopher Schuette>study of megatronics.

15
00:00:45.090 --> 00:00:47.630
<v Christopher Schuette>Um, I started in customer services.

16
00:00:47.709 --> 00:00:52.590
<v Christopher Schuette>I wanted to explore, uh, the fields out
there to how users are.

17
00:00:53.270 --> 00:00:58.830
<v Christopher Schuette>What they're doing with our products, the
Siemens products, um, and after that, also

18
00:00:58.830 --> 00:01:04.750
<v Christopher Schuette>some time in system testing, product
testing, and also training our customers in new

19
00:01:04.750 --> 00:01:09.410
<v Christopher Schuette>products, um, I turned to business
development to just follow the question.

20
00:01:10.260 --> 00:01:14.779
<v Christopher Schuette>How's the future of automation look like,
um, and can I shape it or be part of it

21
00:01:15.339 --> 00:01:22.059
<v Christopher Schuette>and, uh, one of these parts in business
development was great to be at the birth of

22
00:01:22.059 --> 00:01:27.019
<v Christopher Schuette>a new vertical that is called vertical in
logistics, uh, what is an amazing field

23
00:01:27.019 --> 00:01:33.540
<v Christopher Schuette>of, uh, new technologies, the shuttle
systems, HEVs, robotics, um, all these great

24
00:01:33.540 --> 00:01:38.980
<v Christopher Schuette>things that happened there and that
brought me to an, uh, own initiated.

25
00:01:39.819 --> 00:01:46.699
<v Christopher Schuette>Project, uh, that turned now to a complete
product line, um, where we started in to

26
00:01:46.699 --> 00:01:52.400
<v Christopher Schuette>exploring the field of AI enabled
robotics, um, and as you said, uh, I'm now the

27
00:01:52.779 --> 00:01:56.760
<v Christopher Schuette>product portfolio manager in this field of
robotics, AI products.

28
00:01:57.059 --> 00:02:03.419
<v Christopher Schuette>It focused on AI powered applications that
you can upgrade off the shelf robots

29
00:02:03.419 --> 00:02:03.660
<v Christopher Schuette>with.

30
00:02:03.900 --> 00:02:04.139
<v Christopher Schuette>Great.

31
00:02:04.220 --> 00:02:06.459
<v Spencer Acain>So it sounds like you've really been in
this from the start.

32
00:02:06.610 --> 00:02:10.699
<v Spencer Acain>Um, can you tell us a bit more about
those, you know, AI powered robotics, you know,

33
00:02:10.809 --> 00:02:12.429
<v Spencer Acain>how are you upgrading robotics?

34
00:02:12.470 --> 00:02:13.399
<v Spencer Acain>What does that look like?

35
00:02:13.539 --> 00:02:16.360
<v Spencer Acain>And um you know, why is this technology
important?

36
00:02:16.699 --> 00:02:17.750
<v Christopher Schuette>Oh yeah, um.

37
00:02:19.089 --> 00:02:26.050
<v Christopher Schuette>Why it's important, so we identified that,
uh, I take some studies so make also

38
00:02:26.050 --> 00:02:32.369
<v Christopher Schuette>the word for tell us that around 50 to 60%
of work in manufacturing are still

39
00:02:32.369 --> 00:02:38.179
<v Christopher Schuette>performed by manual labor forces and are
not fully roboticized or automated and

40
00:02:38.179 --> 00:02:45.110
<v Christopher Schuette>around 50 to 40% are just in dynamic
changing environments and where traditional

41
00:02:45.110 --> 00:02:47.369
<v Christopher Schuette>automation cannot catch up it's widget.

42
00:02:47.789 --> 00:02:52.080
<v Christopher Schuette>So it struggles in performance and for
that reason humans still need to perform

43
00:02:52.080 --> 00:02:58.740
<v Christopher Schuette>repetitive robot-like work in a time where
the labor pool of

44
00:02:58.919 --> 00:03:04.199
<v Christopher Schuette>workers are shrinking, especially in
industrial countries, and the payment of that

45
00:03:04.199 --> 00:03:07.460
<v Christopher Schuette>workers are rising, so the economical
pressure is high.

46
00:03:08.320 --> 00:03:13.820
<v Christopher Schuette>That are the foundation of what we're
doing, and what we're doing is um.

47
00:03:14.649 --> 00:03:18.889
<v Christopher Schuette>So our first product is probably the one
of the first in the area of physical AI,

48
00:03:19.690 --> 00:03:26.309
<v Christopher Schuette>um, that terms try to explain or
differentiate the digital AI that is behind your

49
00:03:26.309 --> 00:03:31.570
<v Christopher Schuette>screen, um, that helps you to navigate,
that helps you to search in search

50
00:03:31.570 --> 00:03:32.169
<v Christopher Schuette>algorithm.

51
00:03:32.750 --> 00:03:38.149
<v Christopher Schuette>Uh, and also, uh, the chat GPT is for sure
something like in this area, but the

52
00:03:38.149 --> 00:03:44.029
<v Christopher Schuette>physical AI touches our real world, so
it's intact with the real world, um, and it

53
00:03:44.029 --> 00:03:48.039
<v Christopher Schuette>sees reason and act in dynamic changing
environments.

54
00:03:48.429 --> 00:03:52.809
<v Christopher Schuette>That is a complete new paradigm if this
realized as we think it will be.

55
00:03:53.190 --> 00:03:59.110
<v Christopher Schuette>And the first product that we bring up was
in the portfolio, uh PAI that enables any

56
00:03:59.110 --> 00:04:02.009
<v Christopher Schuette>robot to pick items that it has never seen
before.

57
00:04:02.630 --> 00:04:07.789
<v Christopher Schuette>Uh, without training by the user, very
easy to use, very easy to integrate in

58
00:04:07.789 --> 00:04:11.539
<v Christopher Schuette>existing machines, new machines, and that
was just our start point.

59
00:04:11.720 --> 00:04:12.110
<v Spencer Acain>Yeah.

60
00:04:12.470 --> 00:04:16.420
<v Spencer Acain>So it sounds really like you're, you're
kind of addressing that core issue like

61
00:04:16.420 --> 00:04:20.589
<v Spencer Acain>you're saying of, you know, robots are
very static in the past and the way they can

62
00:04:20.589 --> 00:04:21.029
<v Spencer Acain>pick stuff.

63
00:04:21.109 --> 00:04:26.630
<v Spencer Acain>So you have this, this, um, You know, just
ability to handle whatever at any time.

64
00:04:26.660 --> 00:04:31.679
<v Spencer Acain>And then, um, but how is that from your
ability to leverage like more advanced AI

65
00:04:31.679 --> 00:04:36.760
<v Spencer Acain>now, like, how are you seeing the
technologies, that type of technology improving

66
00:04:37.239 --> 00:04:40.119
<v Spencer Acain>with, you know, with more powerful AI
accelerators available because surely this is

67
00:04:40.119 --> 00:04:44.399
<v Spencer Acain>something that's um that requires that
kind of extra power to be able to just do

68
00:04:44.399 --> 00:04:47.859
<v Spencer Acain>that humanlike tasks without without much
training, without any training.

69
00:04:48.679 --> 00:04:51.970
<v Spencer Acain>Uh, you know, like, for example, you know,
I know Siemens start to integrate like

70
00:04:51.970 --> 00:04:53.609
<v Spencer Acain>Nvidia accelerators in their PLCs.

71
00:04:53.679 --> 00:04:55.619
<v Spencer Acain>Is this something you're able to take
advantage of?

72
00:04:56.049 --> 00:04:57.649
<v Christopher Schuette>Oh yeah, definitely, Spencer.

73
00:04:57.850 --> 00:05:02.760
<v Christopher Schuette>Um, first of all, we learned very early
that we in industry want, yeah, we have to

74
00:05:02.760 --> 00:05:06.869
<v Christopher Schuette>sync in industry in system values, not in
isolated technologies.

75
00:05:08.190 --> 00:05:15.190
<v Christopher Schuette>To your question, so, um, combined PLC,
um, and it's fact that the most of the

76
00:05:15.190 --> 00:05:21.130
<v Christopher Schuette>industrial process out there are today PLC
based, um, needs to somehow merge or

77
00:05:21.130 --> 00:05:27.910
<v Christopher Schuette>integrate with what is possible for AI
today that is powered by GPUs also for an

78
00:05:27.910 --> 00:05:28.470
<v Christopher Schuette>Nvidia.

79
00:05:29.290 --> 00:05:35.100
<v Christopher Schuette>And um for that, it's very good to see
that Siemens and Vidia joining forces,

80
00:05:35.250 --> 00:05:42.209
<v Christopher Schuette>building rating systems, and that it
merging closer and closer together.

81
00:05:42.290 --> 00:05:47.829
<v Christopher Schuette>So we, we make some experience to have GPU
power technology very close to the PLC.

82
00:05:48.239 --> 00:05:52.480
<v Christopher Schuette>Um, and then we learned also to have it
separate in the PC is also worth it, and

83
00:05:52.480 --> 00:05:53.920
<v Christopher Schuette>then have a better interfaces.

84
00:05:54.279 --> 00:05:59.440
<v Christopher Schuette>But that is the moment of progress, and
I'm very happy that we are also driving the

85
00:05:59.440 --> 00:06:02.420
<v Christopher Schuette>way toward a software defined automation
concept.

86
00:06:03.070 --> 00:06:10.040
<v Christopher Schuette>And the idea behind that, it's to bring
Um, automation to a level where

87
00:06:10.040 --> 00:06:16.119
<v Christopher Schuette>it comes from a one-off specialized
machine for one customer to

88
00:06:16.119 --> 00:06:21.739
<v Christopher Schuette>amoized product that you can scale out on
the market with less engineering effort,

89
00:06:22.079 --> 00:06:24.000
<v Christopher Schuette>and you can achieve that.

90
00:06:24.899 --> 00:06:31.750
<v Christopher Schuette>By having the ITOT congregation, um, you

91
00:06:31.750 --> 00:06:37.959
<v Christopher Schuette>take the engineering methods from IT that
was built over decades

92
00:06:38.779 --> 00:06:41.470
<v Christopher Schuette>and converted to the industrial level.

93
00:06:41.700 --> 00:06:44.970
<v Christopher Schuette>That is what we're doing with products
like um.

94
00:06:45.399 --> 00:06:49.709
<v Christopher Schuette>The automation expansion product that is
giving you an IT like engineering.

95
00:06:50.010 --> 00:06:55.920
<v Christopher Schuette>On the other hand, to answer your question
on the GPU accelerator level, uh, we have

96
00:06:55.920 --> 00:07:01.399
<v Christopher Schuette>more and more products here that taking
that type of hardware combining better to

97
00:07:01.399 --> 00:07:07.440
<v Christopher Schuette>the PLC world and also to the level where
the PLC get visualized, um.

98
00:07:08.040 --> 00:07:13.260
<v Christopher Schuette>And you can put that on one device, but
also great, and also you can deploy that

99
00:07:13.260 --> 00:07:15.339
<v Christopher Schuette>with a concept like industrial edge.

100
00:07:15.640 --> 00:07:19.920
<v Christopher Schuette>So all these things are perfect and it's
really a system value, but I want to tell

101
00:07:19.920 --> 00:07:21.160
<v Christopher Schuette>you what we learned.

102
00:07:22.049 --> 00:07:24.190
<v Christopher Schuette>To use bigger AI models.

103
00:07:24.769 --> 00:07:31.390
<v Christopher Schuette>So our R&amp;D teams come from, Um, AI
based computer vision for post estimation at the

104
00:07:31.390 --> 00:07:32.089
<v Christopher Schuette>very beginning.

105
00:07:33.029 --> 00:07:39.679
<v Christopher Schuette>And we released just last year a second
product tier um that use also an industrial

106
00:07:39.679 --> 00:07:45.720
<v Christopher Schuette>grade foundation model for segmentation of
random items as is one part of our

107
00:07:45.720 --> 00:07:50.390
<v Christopher Schuette>product and it helps to provide our users
with more information about the unknown

108
00:07:50.390 --> 00:07:50.910
<v Christopher Schuette>item.

109
00:07:51.320 --> 00:07:56.149
<v Christopher Schuette>So at least the length, the width, the
orientation, it also part of an intelligent

110
00:07:56.149 --> 00:08:00.059
<v Christopher Schuette>suction group activation feature that
adapts to the multi-suction group are used.

111
00:08:00.660 --> 00:08:07.010
<v Christopher Schuette>To adapt to the item size position in the
room and at the end it's for the user less

112
00:08:07.010 --> 00:08:13.510
<v Christopher Schuette>to change ah and picking in a very wide
range of items just by advancing the

113
00:08:13.799 --> 00:08:15.570
<v Christopher Schuette>architecture of neural networks.

114
00:08:16.440 --> 00:08:20.600
<v Christopher Schuette>And we have another core metric that is
the end to end compute time.

115
00:08:21.290 --> 00:08:24.309
<v Christopher Schuette>And we learned with the huge foundation
model that I just named.

116
00:08:24.880 --> 00:08:31.700
<v Christopher Schuette>It took us from some milliseconds, so the
first A tier of our PI was

117
00:08:31.700 --> 00:08:38.599
<v Christopher Schuette>able to run on a, yeah, tablet PC like CPU
without the GPU, that was

118
00:08:38.599 --> 00:08:39.179
<v Christopher Schuette>OK.

119
00:08:39.799 --> 00:08:46.400
<v Christopher Schuette>The new version edition needs a GPU
because it takes with GPU over 2 seconds to

120
00:08:46.400 --> 00:08:46.780
<v Christopher Schuette>compute.

121
00:08:47.039 --> 00:08:50.619
<v Christopher Schuette>That is not good, uh, when you have a very
high performance machine.

122
00:08:51.440 --> 00:08:56.520
<v Christopher Schuette>So you need this GPU integration to reduce
the performance and to compute time

123
00:08:56.520 --> 00:08:57.039
<v Christopher Schuette>heavily.

124
00:08:57.219 --> 00:08:58.099
<v Christopher Schuette>That is obvious.

125
00:08:58.520 --> 00:09:04.039
<v Christopher Schuette>I'm very great to have us so integrated
with Nvidia together, um, and it's so great

126
00:09:04.039 --> 00:09:06.989
<v Christopher Schuette>that we are getting more and more friends
here with Nvidia colleagues.

127
00:09:07.000 --> 00:09:13.440
<v Christopher Schuette>We are in regular exchange and, uh,
bringing this infrastructure in a more tangible

128
00:09:13.440 --> 00:09:15.299
<v Christopher Schuette>system value way on the market.

129
00:09:15.940 --> 00:09:19.080
<v Christopher Schuette>Um, very curious about what will happen
also in the future here.

130
00:09:19.460 --> 00:09:22.679
<v Spencer Acain>Yeah, no, I can, I can imagine, but kind
of to.

131
00:09:23.640 --> 00:09:27.520
<v Spencer Acain>To pivot from that, like, you know, you,
you have all this extra power available now

132
00:09:27.520 --> 00:09:33.840
<v Spencer Acain>in like these PM the PLCs on the factory
floor with these accelerators, but to reach

133
00:09:33.840 --> 00:09:37.960
<v Spencer Acain>the kind of goal that you're talking about
of just being able to have a robot that's

134
00:09:37.960 --> 00:09:42.520
<v Spencer Acain>picking anything and everything without
any training at the end or very little

135
00:09:42.520 --> 00:09:47.159
<v Spencer Acain>training at the end, is that, um, how much
like back end training has to go into

136
00:09:47.159 --> 00:09:48.000
<v Spencer Acain>these AI models?

137
00:09:48.080 --> 00:09:48.820
<v Spencer Acain>Is it all?

138
00:09:49.400 --> 00:09:52.570
<v Spencer Acain>Like, do you pre-train it completely and
then just ship it out and then the

139
00:09:52.570 --> 00:09:55.960
<v Spencer Acain>customers don't need to do anything else,
or do they have to train as well?

140
00:09:55.969 --> 00:09:59.809
<v Spencer Acain>Or how are you able to, how are you
managing that and that kind of training of these

141
00:09:59.809 --> 00:10:03.190
<v Spencer Acain>models to be able to handle these diverse
situations?

142
00:10:03.200 --> 00:10:07.570
<v Christopher Schuette>Unfortunately, I can give not a very clear
answer to that, but let me try it.

143
00:10:08.070 --> 00:10:14.690
<v Christopher Schuette>Um, one of the fundamental product
decisions, uh, we did, it's, if you want to give

144
00:10:14.690 --> 00:10:18.940
<v Christopher Schuette>these retraining capability in the hands
of our users, uh, which are integrators and

145
00:10:18.940 --> 00:10:25.780
<v Christopher Schuette>OEMs of, so these creators of advanced
robotics solutions, um, with all its

146
00:10:25.780 --> 00:10:31.330
<v Christopher Schuette>complexity, um, so they need to to get the
data, they need to label it, retrain it,

147
00:10:31.659 --> 00:10:32.179
<v Christopher Schuette>test it.

148
00:10:32.659 --> 00:10:37.750
<v Christopher Schuette>Um, or if we make their lives easier, as
easy as possible and keep the retraining in

149
00:10:37.750 --> 00:10:38.630
<v Christopher Schuette>our product team.

150
00:10:39.440 --> 00:10:45.119
<v Christopher Schuette>And we decided to take it in our product
team and give the OEMs integrators an out

151
00:10:45.119 --> 00:10:50.179
<v Christopher Schuette>of the box performance model and in the
case the model not reaches the performance

152
00:10:50.179 --> 00:10:55.320
<v Christopher Schuette>that we expect and theirs expect, uh, we
are able to retrain these models.

153
00:10:55.940 --> 00:11:00.960
<v Christopher Schuette>And how big is this effort to do so, just
to give you a thought.

154
00:11:01.840 --> 00:11:08.609
<v Christopher Schuette>We haven't Product that uh the the value
proposition is that it can grasp uh the

155
00:11:08.609 --> 00:11:13.049
<v Christopher Schuette>most of the items in the warehouse nearly
to every item in the warehouse, um, we are

156
00:11:13.049 --> 00:11:14.799
<v Christopher Schuette>on the way to, to get there.

157
00:11:15.210 --> 00:11:20.929
<v Christopher Schuette>It's different to these approaches where
you specifically train one item in

158
00:11:20.929 --> 00:11:26.369
<v Christopher Schuette>manufacturing probably in a new network
and it, it works reliable for that one out

159
00:11:26.369 --> 00:11:27.070
<v Christopher Schuette>for another one.

160
00:11:27.609 --> 00:11:33.059
<v Christopher Schuette>It's especially why we take these training
in our capability in a product team

161
00:11:33.059 --> 00:11:38.340
<v Christopher Schuette>because we cover item classes, not a
specific product that is fitted by CAT data,

162
00:11:38.380 --> 00:11:39.039
<v Christopher Schuette>for example.

163
00:11:40.690 --> 00:11:43.500
<v Christopher Schuette>And how big is such an effort?

164
00:11:44.080 --> 00:11:46.090
<v Christopher Schuette>It took us up to 3 years.

165
00:11:46.950 --> 00:11:52.799
<v Christopher Schuette>To come from a little baby network um for
5 degrees of reasons, uh pick post

166
00:11:52.799 --> 00:11:59.260
<v Christopher Schuette>detection and computation with a really
good industrial grade performance to

167
00:11:59.960 --> 00:12:06.419
<v Christopher Schuette>what we have today, um, a multi-section we
ready, uh, deployed in the field, uh,

168
00:12:06.429 --> 00:12:11.869
<v Christopher Schuette>neural network, the foundational model
backed, um, that really makes a job in, in

169
00:12:11.869 --> 00:12:12.179
<v Christopher Schuette>operations.

170
00:12:12.739 --> 00:12:18.119
<v Christopher Schuette>So 3 years of collecting data, firstly
from, uh, synthetic data from simulation,

171
00:12:18.500 --> 00:12:23.900
<v Christopher Schuette>because it's cheap to generate and easy to
generate, but it turns out that you need

172
00:12:23.900 --> 00:12:27.859
<v Christopher Schuette>to find other ways to make really a
reliable industrial grid.

173
00:12:28.250 --> 00:12:33.289
<v Spencer Acain>So, beyond that synthetic data you were
just talking about, you know, how are you

174
00:12:33.289 --> 00:12:36.770
<v Spencer Acain>able to get all the training data then,
um, you know, obviously, like you said, it

175
00:12:36.770 --> 00:12:41.929
<v Spencer Acain>can't, you can't just use CAD data or
synthetic data for this, uh, so, so you needed

176
00:12:41.929 --> 00:12:42.849
<v Spencer Acain>a lot of real data.

177
00:12:43.820 --> 00:12:47.659
<v Spencer Acain>To pre-train and develop your models,
where, so where are you able to get all of

178
00:12:47.659 --> 00:12:48.000
<v Spencer Acain>that?

179
00:12:49.010 --> 00:12:50.190
<v Christopher Schuette>Yeah, good question.

180
00:12:51.090 --> 00:12:55.729
<v Christopher Schuette>The underlying challenge most companies
working on physical AI are facing is the

181
00:12:55.729 --> 00:13:00.190
<v Christopher Schuette>limited availability of large scale real
world robotics data.

182
00:13:00.809 --> 00:13:05.640
<v Christopher Schuette>Models like GPT were trained on internet
scale data sets, and we don't have this

183
00:13:05.640 --> 00:13:08.369
<v Christopher Schuette>equivalent data foundation for robotic
learning.

184
00:13:08.890 --> 00:13:13.359
<v Christopher Schuette>So the data gap remains very large and
it's probably one of the biggest challenges

185
00:13:13.359 --> 00:13:16.849
<v Christopher Schuette>for model performance across the market
and its success.

186
00:13:17.409 --> 00:13:19.950
<v Christopher Schuette>So it's well recognized across the
technology field.

187
00:13:20.250 --> 00:13:24.570
<v Christopher Schuette>For applied AI robotic system in
industrial settings, continuous improvement depends

188
00:13:24.570 --> 00:13:31.309
<v Christopher Schuette>heavily on operational real world data,
particularly on the edge cases.

189
00:13:31.849 --> 00:13:37.669
<v Christopher Schuette>Uh, this includes vision data along with
contextual signals such as system states,

190
00:13:37.729 --> 00:13:40.150
<v Christopher Schuette>also from the PLC and outcomes.

191
00:13:41.049 --> 00:13:45.690
<v Christopher Schuette>For example, perception foundation models
like segmentation models, foundational

192
00:13:45.690 --> 00:13:52.010
<v Christopher Schuette>ones are trained on large graded labeled
data sets, often still involving

193
00:13:52.010 --> 00:13:54.650
<v Christopher Schuette>significant manual annotation efforts.

194
00:13:55.840 --> 00:14:01.760
<v Christopher Schuette>Yeah, looking ahead, vision action models
promise to generalize manipulation skills

195
00:14:01.760 --> 00:14:07.000
<v Christopher Schuette>across tasks and environments and to
achieve this, virtual training data sets are

196
00:14:07.000 --> 00:14:07.599
<v Christopher Schuette>required.

197
00:14:09.059 --> 00:14:15.320
<v Christopher Schuette>Today, a combination of visual observation
with trajectory, level, motion, context,

198
00:14:16.140 --> 00:14:21.099
<v Christopher Schuette>um, this allows models to understand the
scene versus grade and adapt directly the

199
00:14:21.099 --> 00:14:22.409
<v Christopher Schuette>robot motion to it.

200
00:14:22.820 --> 00:14:25.599
<v Christopher Schuette>Um, the technology field is evolving
quickly.

201
00:14:26.330 --> 00:14:31.080
<v Christopher Schuette>Within a year, entire foundation model
architectures can change several times, um,

202
00:14:31.140 --> 00:14:37.580
<v Christopher Schuette>driven also by new research insights and
this significantly advanced the maturity of

203
00:14:37.580 --> 00:14:38.359
<v Christopher Schuette>the technology.

204
00:14:39.039 --> 00:14:45.900
<v Christopher Schuette>Um, one constant will probably stay, uh,
infrastructure to feed the need for large

205
00:14:45.900 --> 00:14:48.859
<v Christopher Schuette>scale data collection and continuous moral
improvement.

206
00:14:49.210 --> 00:14:54.179
<v Christopher Schuette>And it's this vision behind the mostly
automated data flywheel concept.

207
00:14:54.880 --> 00:15:01.400
<v Christopher Schuette>In the foreseeable future, real world data
stay a must, not optional, and to

208
00:15:01.400 --> 00:15:04.700
<v Christopher Schuette>achieve internet great performance.

209
00:15:05.659 --> 00:15:10.169
<v Christopher Schuette>That is why we actively building the
capability within our product team.

210
00:15:10.669 --> 00:15:14.229
<v Spencer Acain>Yeah, I mean, obviously, building that
kind of infrastructure and designing that

211
00:15:14.229 --> 00:15:18.510
<v Spencer Acain>isn't something you want to have everybody
doing on their own all the time, but that

212
00:15:18.510 --> 00:15:20.210
<v Spencer Acain>is all the time we have for this episode.

213
00:15:20.429 --> 00:15:24.750
<v Spencer Acain>So once again, I have been your host,
Spencer Kane, joined by Christopher Shita on

214
00:15:24.750 --> 00:15:25.969
<v Spencer Acain>the AI Spectrum podcast.

215
00:15:26.469 --> 00:15:30.390
<v Spencer Acain>Tune in again next time as we continue
diving into what it takes to bring AI

216
00:15:30.390 --> 00:15:32.070
<v Spencer Acain>automation into the factory.
