WEBVTT

1
00:00:00.000 --> 00:00:04.904
We're filled with the use
cases of language right now,

2
00:00:04.904 --> 00:00:07.907
language AI, which is really exciting.

3
00:00:07.907 --> 00:00:10.276
But we underappreciate

4
00:00:10.276 --> 00:00:13.146
how perceptual and physical

5
00:00:13.146 --> 00:00:18.146
our living and working experiences are.

6
00:00:18.651 --> 00:00:19.919
AI is moving fast.

7
00:00:19.919 --> 00:00:21.788
ROI expectations are rising,

8
00:00:21.788 --> 00:00:23.990
and leaders everywhere
are racing to figure out

9
00:00:23.990 --> 00:00:27.160
what it actually takes to
build an AI-ready organization.

10
00:00:27.160 --> 00:00:29.829
To help us understand and
make sense of this moment,

11
00:00:29.829 --> 00:00:31.998
I'm joined by two people
at the very center

12
00:00:31.998 --> 00:00:33.867
of this conversation globally,

13
00:00:33.867 --> 00:00:36.569
Dr. Fei-Fei Li and Dylan Bolden.

14
00:00:36.569 --> 00:00:40.040
Many people call Dr.
Li the godmother of AI

15
00:00:40.040 --> 00:00:41.141
and for good reason,

16
00:00:41.141 --> 00:00:43.043
she created ImageNet,

17
00:00:43.043 --> 00:00:44.277
which is the database

18
00:00:44.277 --> 00:00:47.047
that literally taught computers to see.

19
00:00:47.047 --> 00:00:49.849
That breakthrough made possible
everything from Google Lens

20
00:00:49.849 --> 00:00:51.384
to self-driving cars.

21
00:00:51.384 --> 00:00:54.354
And today as co-founder
and CEO of World Labs,

22
00:00:54.354 --> 00:00:56.923
she's building large world models,

23
00:00:56.923 --> 00:00:59.426
AI that doesn't just process data,

24
00:00:59.426 --> 00:01:04.426
but understands and navigates
the physical world around us.

25
00:01:04.798 --> 00:01:07.267
And Dylan Bolden has spent
over two decades helping

26
00:01:07.267 --> 00:01:10.437
the world's most complex
organizations figure out

27
00:01:10.437 --> 00:01:11.738
what's next.

28
00:01:11.738 --> 00:01:14.140
He's managing director
and senior partner at BCG,

29
00:01:14.140 --> 00:01:17.644
and in that role, he leads the
firm's global thinking on AI

30
00:01:17.644 --> 00:01:19.179
and its impact on business.

31
00:01:19.179 --> 00:01:21.514
Dr. Li, Dylan, welcome.

32
00:01:21.514 --> 00:01:22.348
Thank you.

33
00:01:22.348 --> 00:01:23.483
Glad to be here.

34
00:01:23.483 --> 00:01:27.754
You used to be the chief AI
scientist at Google Cloud

35
00:01:27.754 --> 00:01:29.089
at one point.

36
00:01:29.089 --> 00:01:32.025
During that time, was there any one moment

37
00:01:32.025 --> 00:01:36.396
that made you realize
that spatial intelligence

38
00:01:36.396 --> 00:01:37.497
was the future?

39
00:01:37.497 --> 00:01:40.233
Obviously at this point we
know that, but back then.

40
00:01:41.601 --> 00:01:42.869
The short answer is yes.

41
00:01:42.869 --> 00:01:47.307
So I became Google
Chief Scientist in 2017,

42
00:01:47.307 --> 00:01:50.543
but that was already
17 years into my career

43
00:01:50.543 --> 00:01:51.811
as a AI scientist.

44
00:01:51.811 --> 00:01:55.815
It's been my life's work
thinking about vision,

45
00:01:55.815 --> 00:01:58.218
perception, robotic learning.

46
00:01:58.218 --> 00:02:00.120
At Google Cloud,

47
00:02:00.120 --> 00:02:03.857
what did change me is that I get to work

48
00:02:03.857 --> 00:02:06.860
with all the vertical industries,

49
00:02:06.860 --> 00:02:10.029
from hospitals to financial services,

50
00:02:10.029 --> 00:02:14.968
from energy to agriculture.

51
00:02:14.968 --> 00:02:19.672
And so much of those needs
were physical.For example,

52
00:02:19.672 --> 00:02:21.374
in financial services,

53
00:02:21.374 --> 00:02:25.578
ensures damage assessment

54
00:02:25.578 --> 00:02:28.414
or appraisal is very visual.

55
00:02:28.414 --> 00:02:33.414
In transportation, whether
we're talking about driving

56
00:02:34.387 --> 00:02:38.925
or shipping or--all those are very visual

57
00:02:38.925 --> 00:02:43.196
and perceptual--or
agriculture. You know, weeds,

58
00:02:44.197 --> 00:02:48.701
understanding weeds and picking
weeds or harvesting apples

59
00:02:48.701 --> 00:02:51.871
or counting salmons of certain size.

60
00:02:51.871 --> 00:02:55.141
All of these are extremely visual spatial.

61
00:02:55.141 --> 00:02:59.479
So in our world, I know that we're filled

62
00:02:59.479 --> 00:03:03.483
with the use cases of language right now,

63
00:03:03.483 --> 00:03:06.486
language AI, which is really exciting.

64
00:03:06.486 --> 00:03:08.855
But we underappreciate

65
00:03:08.855 --> 00:03:13.855
how perceptual and physical our living

66
00:03:13.860 --> 00:03:16.963
and working experiences are.

67
00:03:16.963 --> 00:03:19.065
You know, I think about businesses

68
00:03:19.065 --> 00:03:22.802
that require physical space, retail,

69
00:03:22.802 --> 00:03:25.672
travel, industries, hospitality.

70
00:03:26.706 --> 00:03:28.474
In those environments,

71
00:03:28.474 --> 00:03:32.478
when we have AI that can perceive space,

72
00:03:32.478 --> 00:03:35.081
that can see space,

73
00:03:35.081 --> 00:03:38.751
which of those business
models breaks first?

74
00:03:38.751 --> 00:03:43.751
Think about a retailer that can better
in real time, understand based on

75
00:03:43.756 --> 00:03:47.594
how the customer experiences the store,

76
00:03:47.594 --> 00:03:49.562
where inventory needs to be.

77
00:03:49.562 --> 00:03:53.233
But I think honestly the first
place to break is in labor,

78
00:03:53.233 --> 00:03:56.236
figuring out where's the best place

79
00:03:56.236 --> 00:04:00.540
to put labor to get the most
out of it within a store

80
00:04:00.540 --> 00:04:04.244
where at certain times a day,

81
00:04:04.244 --> 00:04:07.413
do I put people in certain
parts of the operation

82
00:04:07.413 --> 00:04:08.514
with them on a restaurant?

83
00:04:08.514 --> 00:04:11.784
Do I put them next to
the front of the store?

84
00:04:11.784 --> 00:04:13.286
Do I put them in the back?

85
00:04:13.286 --> 00:04:16.089
So really good observation around

86
00:04:16.089 --> 00:04:19.726
how consumers experience the store

87
00:04:19.726 --> 00:04:21.060
I think gets you to better decisions

88
00:04:21.060 --> 00:04:22.262
around how to deploy labor.

89
00:04:22.262 --> 00:04:23.162
So that's one example.

90
00:04:23.162 --> 00:04:25.231
But I'm very interested to hear.

91
00:04:25.231 --> 00:04:26.065
Dr. Li.

92
00:04:26.065 --> 00:04:28.234
Your perspective here.

93
00:04:28.234 --> 00:04:31.337
So there's a lot of future opportunities,

94
00:04:31.337 --> 00:04:35.308
but yes, as the technology matures,

95
00:04:35.308 --> 00:04:36.409
I agree with Dylan,

96
00:04:36.409 --> 00:04:41.409
we don't really know what's
around in our physical space.

97
00:04:41.714 --> 00:04:43.850
We don't have a easy way

98
00:04:43.850 --> 00:04:46.886
to express our imagination

99
00:04:46.886 --> 00:04:49.856
into a potential physical space

100
00:04:49.856 --> 00:04:53.026
if we're talking about
storytelling or design

101
00:04:53.026 --> 00:04:56.763
or optimizing for office space

102
00:04:56.763 --> 00:05:01.367
or designing, you know,
hospitality environments

103
00:05:01.367 --> 00:05:05.071
or live events or many of these.

104
00:05:05.071 --> 00:05:09.475
So I think as this technology matures,

105
00:05:09.475 --> 00:05:12.679
as we push the frontier
of this technology,

106
00:05:12.679 --> 00:05:14.480
we will create world models

107
00:05:14.480 --> 00:05:17.817
that can reconstruct physical space,

108
00:05:17.817 --> 00:05:21.087
can understand what's
in this physical space,

109
00:05:21.087 --> 00:05:25.291
can predict the next state
of this physical space,

110
00:05:25.291 --> 00:05:30.291
can translate imagination into digital

111
00:05:30.363 --> 00:05:32.732
expression of the physical space.

112
00:05:32.732 --> 00:05:36.669
All this will empower
many business use cases.

113
00:05:36.669 --> 00:05:39.205
Thank you so much, Dylan
Bolden and Dr. Fei-Fei Li.

114
00:05:39.205 --> 00:05:40.340
Thank you.
Thank you.