1
00:00:00,000 --> 00:00:01,837
at the forefront of artificial

2
00:00:01,920 --> 00:00:04,637
intelligence as I request him to address

3
00:00:04,720 --> 00:00:07,557
the esteemed gathering.

4
00:00:07,640 --> 00:00:11,047
Please welcome Mr. Dario Amodei.

5
00:00:11,130 --> 00:00:13,150
>> [applause]

6
00:00:16,880 --> 00:00:20,997
>> First, I want to thank Prime Minister
Modi for bringing us together. The

7
00:00:21,080 --> 00:00:23,597
energy and ambition in this room and

8
00:00:23,680 --> 00:00:25,797
across India are incredible. I've been

9
00:00:25,880 --> 00:00:28,357
spending the last few days meeting with

10
00:00:28,440 --> 00:00:31,117
Indian builders and enterprises and the

11
00:00:31,200 --> 00:00:33,037
energy to build together here is

12
00:00:33,120 --> 00:00:36,437
palpable unlike anywhere else. Um this

13
00:00:36,520 --> 00:00:39,197
is the fourth AI summit we've held since

14
00:00:39,280 --> 00:00:41,797
the tradition was initiated at Bletchley

15
00:00:41,880 --> 00:00:45,077
Park back in 2023, which I still

16
00:00:45,160 --> 00:00:48,157
remember. And in those 2.5 years, the

17
00:00:48,240 --> 00:00:53,077
advances in the technology have been
absolutely staggering. Along with those,

18
00:00:53,160 --> 00:00:57,037
the advances in the commercial
applications and the societal and

19
00:00:57,120 --> 00:00:59,557
ethical questions around the technology

20
00:00:59,640 --> 00:01:02,397
have only grown more urgent. My

21
00:01:02,480 --> 00:01:05,637
fundamental view is that AI has been on

22
00:01:05,720 --> 00:01:08,077
an exponential for the last for the last

23
00:01:08,160 --> 00:01:11,677
10 years. Um and as part of a sort of

24
00:01:11,760 --> 00:01:14,197
Moore's law for intelligence and that we

25
00:01:14,280 --> 00:01:16,797
are now well advanced on that curve and

26
00:01:16,880 --> 00:01:19,117
there are only a small number of years

27
00:01:19,200 --> 00:01:21,837
for AI models surpassing the cognitive

28
00:01:21,920 --> 00:01:26,677
capabilities of most humans for most
things. We're increasingly close to what

29
00:01:26,760 --> 00:01:28,437
I've called a country of geniuses in a

30
00:01:28,520 --> 00:01:31,437
data center. A set of AI agents that are

31
00:01:31,520 --> 00:01:34,037
more capable than most humans at most

32
00:01:34,120 --> 00:01:37,637
things and can coordinate at superhuman

33
00:01:37,720 --> 00:01:41,917
speed. That level of capability
is something the world has never seen

34
00:01:42,000 --> 00:01:44,477
before and brings a very wide range of

35
00:01:44,560 --> 00:01:46,957
both opportunities and concerns for

36
00:01:47,040 --> 00:01:49,557
humanity. On the positive side, we have

37
00:01:49,640 --> 00:01:52,317
the potential to cure diseases that have

38
00:01:52,400 --> 00:01:54,717
been incurable for thousands of years,

39
00:01:54,800 --> 00:01:57,437
to radically improve human health, and

40
00:01:57,520 --> 00:01:59,917
to lift billions out of poverty

41
00:02:00,000 --> 00:02:04,637
including the global south and create
a better world for everyone. On the side

42
00:02:04,720 --> 00:02:06,717
of risks, I'm concerned about the

43
00:02:06,800 --> 00:02:09,717
autonomous behavior of AI models, their

44
00:02:09,800 --> 00:02:12,557
potential for misuse by individuals and

45
00:02:12,640 --> 00:02:14,797
governments and their potential for

46
00:02:14,880 --> 00:02:17,517
economic displacement. India has an

47
00:02:17,600 --> 00:02:20,397
absolutely central role to play in these

48
00:02:20,480 --> 00:02:24,797
questions and challenges both on the
side of the opportunities and on the

49
00:02:24,880 --> 00:02:26,997
side of the risks. As a sign of our

50
00:02:27,080 --> 00:02:29,477
commitment, we just this week opened an

51
00:02:29,560 --> 00:02:32,437
office in Bengaluru and hired Irena

52
00:02:32,520 --> 00:02:37,260
Ghosh who has spent >> [applause]

53
00:02:37,920 --> 00:02:39,917
>> who has spent three decades building

54
00:02:40,000 --> 00:02:42,477
businesses in India as our managing

55
00:02:42,560 --> 00:02:45,237
director for Anthropic India. We've also

56
00:02:45,320 --> 00:02:50,437
announced partnerships with major Indian
enterprises this week including Infosys

57
00:02:50,520 --> 00:02:54,437
and others. On the opportunities,

58
00:02:54,520 --> 00:02:56,877
one dynamic that we have observed is

59
00:02:56,960 --> 00:02:59,637
that technology and practices pioneered

60
00:02:59,720 --> 00:03:01,637
in India have historically set a

61
00:03:01,720 --> 00:03:04,677
standard for the global south.

62
00:03:04,760 --> 00:03:07,277
And have helped to diffuse knowledge and

63
00:03:07,360 --> 00:03:11,797
humanitarian benefits through the global
south. We're therefore partnering with

64
00:03:11,880 --> 00:03:13,397
we have been partnering with for several

65
00:03:13,480 --> 00:03:15,957
months non-profits such as the XStep

66
00:03:16,040 --> 00:03:18,957
Foundation, Pratham and Central Square

67
00:03:19,040 --> 00:03:21,357
Foundation to use our models to advance

68
00:03:21,440 --> 00:03:23,917
digital infrastructure, education,

69
00:03:24,000 --> 00:03:26,157
agricultural efficiency and health in

70
00:03:26,240 --> 00:03:28,677
the hopes of spreading AI's benefits

71
00:03:28,760 --> 00:03:33,477
across the global south starting with
India and diffusing out to the rest of

72
00:03:33,560 --> 00:03:37,037
the global south. We're also
partnering with Karya and the

73
00:03:37,120 --> 00:03:38,877
Collective Intelligence Project to build

74
00:03:38,960 --> 00:03:42,157
evaluations and metrics of our model

75
00:03:42,240 --> 00:03:44,517
Claude's performance on India's many

76
00:03:44,600 --> 00:03:47,197
regional languages on practical and

77
00:03:47,280 --> 00:03:49,837
locally relevant tasks. We'll benchmark

78
00:03:49,920 --> 00:03:52,637
like agriculture, legal tasks and

79
00:03:52,720 --> 00:03:56,237
educational content. On the risks, India

80
00:03:56,320 --> 00:04:00,757
is the world's largest democracy and can
be a partner and leader in addressing

81
00:04:00,840 --> 00:04:04,877
the global security and economic risks
of the technology. We'd like to work

82
00:04:04,960 --> 00:04:07,477
with India on testing and evaluation of

83
00:04:07,560 --> 00:04:09,957
models for safety and security risks in

84
00:04:10,040 --> 00:04:14,917
the tradition that was started
by many global and national AI security

85
00:04:15,000 --> 00:04:18,077
institutes that have been
stood up around the world.

86
00:04:18,160 --> 00:04:20,557
Even more, we see a particularly strong

87
00:04:20,640 --> 00:04:22,797
opportunity to work with India on

88
00:04:22,880 --> 00:04:27,877
studying the economic questions. As part
of the New Delhi Frontier AI Impact

89
00:04:27,960 --> 00:04:32,437
Commitments, which we're excited to join.
As part of our Anthropic Economic

90
00:04:32,520 --> 00:04:34,637
Futures Program and Anthropic Economic

91
00:04:34,720 --> 00:04:37,797
Index, we publish statistical insights

92
00:04:37,880 --> 00:04:41,397
into how AI impacts jobs in the economy.

93
00:04:41,480 --> 00:04:43,877
We're excited to increasingly share this

94
00:04:43,960 --> 00:04:46,237
information, exchange information with

95
00:04:46,320 --> 00:04:48,557
the Indian government, to share insights

96
00:04:48,640 --> 00:04:51,237
and inform evidence-based policy making,

97
00:04:51,320 --> 00:04:53,517
convene meetings with economists, labor

98
00:04:53,600 --> 00:04:56,317
leaders, and policy makers to adjust to

99
00:04:56,400 --> 00:04:58,957
to to adapt to the economic impacts of

100
00:04:59,040 --> 00:05:01,957
AI. We believe that AI will greatly grow

101
00:05:02,040 --> 00:05:06,157
the economic pie, including in India and
the global south, but that because it is

102
00:05:06,240 --> 00:05:10,317
happening so fast, it may lead to a time
of disruption. And we need to work

103
00:05:10,400 --> 00:05:14,717
together between companies and the
government to better manage that time of

104
00:05:14,800 --> 00:05:16,917
dis- disruption and bring better

105
00:05:17,000 --> 00:05:20,077
prosperity smoothly to all. Um I and

106
00:05:20,160 --> 00:05:24,357
Anthropic are very grateful to be part
of all these efforts, and I'm honored to

107
00:05:24,440 --> 00:05:29,360
be here and working on these
questions with all of you.

108
00:05:30,120 --> 00:05:32,140
>> [applause]

109
00:05:35,400 --> 00:05:42,040
>> Thank you, Mr. Dario Amodei,
for that illuminating address. Ladies and