1
00:00:00,000 --> 00:00:02,157
Members of the committee, thank you
for the opportunity to be here today to

2
00:00:02,240 --> 00:00:04,157
discuss the steps that must be taken to

3
00:00:04,240 --> 00:00:07,277
ensure US leadership in AI. My name is

4
00:00:07,360 --> 00:00:11,437
Alexander Wang. I'm the founder and CEO
of Scale AI. Today's hearing is personal

5
00:00:11,520 --> 00:00:15,357
for me. I grew up in Los Alamos,
where my parents were physicists at the

6
00:00:15,440 --> 00:00:19,277
National Lab, the birthplace
of the atomic bomb. They taught me that

7
00:00:19,360 --> 00:00:21,037
America's leadership in science and

8
00:00:21,120 --> 00:00:23,117
technology is vital to our national

9
00:00:23,200 --> 00:00:28,237
security and global strength. At MIT,
I learned that progress in AI depends on

10
00:00:28,320 --> 00:00:33,277
three key elements. data,
compute and algorithms. While most

11
00:00:33,360 --> 00:00:37,197
of my classmates pursued expertise
in comput and algorithms, few were focused

12
00:00:37,280 --> 00:00:39,437
on the data challenge. That inspired me

13
00:00:39,520 --> 00:00:42,957
to start scale. We deliver expert level

14
00:00:43,040 --> 00:00:47,117
data and offer technology solutions
to leading AI labs, multinational

15
00:00:47,200 --> 00:00:49,357
enterprises, and the US government and

16
00:00:49,440 --> 00:00:52,397
our allies. At scale, we keep humans at

17
00:00:52,480 --> 00:00:54,157
the center of everything we do because

18
00:00:54,240 --> 00:00:56,397
AI should always work for us, not the

19
00:00:56,480 --> 00:00:59,517
other way around. Over the past decade,

20
00:00:59,600 --> 00:01:04,077
it's become clear that the United States
faces intense global competition in

21
00:01:04,160 --> 00:01:06,157
determining how AI should evolve and who

22
00:01:06,240 --> 00:01:11,197
should lead. In 2018, the Chinese
Communist Party's AI master plan started

23
00:01:11,280 --> 00:01:15,597
taking shape. They were already
developing advanced AI capabilities and

24
00:01:15,680 --> 00:01:19,597
using that technology to surveil
and suppress their people. Fast forward to

25
00:01:19,680 --> 00:01:21,757
today, their plan is more sophisticated

26
00:01:21,840 --> 00:01:24,237
and expansive. It includes four key

27
00:01:24,320 --> 00:01:27,997
areas of focus. First, the CCP is taking

28
00:01:28,080 --> 00:01:31,357
a whole of country approach,
having recently launched their AI plus

29
00:01:31,440 --> 00:01:34,157
initiative. Second, the CCP is

30
00:01:34,240 --> 00:01:38,637
outinvesting us in data, spending
billions on AI ready data and unlocking

31
00:01:38,720 --> 00:01:43,757
vast public data sets to fuel AI systems.
Third, they're finding ways to

32
00:01:43,840 --> 00:01:47,437
catch up on compute and building out
their physical infrastructure. And

33
00:01:47,520 --> 00:01:52,237
lastly, they're developing leading AI
models and exporting them to the world.

34
00:01:52,320 --> 00:01:55,597
But we are not here today to just
talk about what China is doing, but to

35
00:01:55,680 --> 00:02:00,317
identify how the US can lead. Given
how close the competition is across all

36
00:02:00,400 --> 00:02:02,477
foundational elements, the policies this

37
00:02:02,560 --> 00:02:07,437
Congress promotes could determine
the outcome. Global AI dominance is not

38
00:02:07,520 --> 00:02:11,197
about trying to level the playing field
by mimicking China's authoritarian way

39
00:02:11,280 --> 00:02:15,597
of government and AI adoption. Instead,
the United States must chart our own

40
00:02:15,680 --> 00:02:20,317
course, one that is anchored in American
values. This is vital to our long-term

41
00:02:20,400 --> 00:02:23,517
national security.
This requires decisive action by the

42
00:02:23,600 --> 00:02:25,997
United States across four main themes.

43
00:02:26,080 --> 00:02:28,637
Dominate, unleash, innovate, and

44
00:02:28,720 --> 00:02:33,837
promote. To dominate, we need to win
on data. The US government is one of the

45
00:02:33,920 --> 00:02:37,277
largest producers of data in the world,
but currently most of that data is

46
00:02:37,360 --> 00:02:41,197
unavailable to advance American AI
leadership. There are three immediate

47
00:02:41,280 --> 00:02:45,357
actions that would move us forward
towards data dominance. First, establish

48
00:02:45,440 --> 00:02:49,517
a national AI data reserve. This
resource should serve as a centralized

49
00:02:49,600 --> 00:02:53,037
hub for the government's AI work,
housing relevant government data and

50
00:02:53,120 --> 00:02:56,397
allowing it to be easily shared between
agencies and enabling widespread AI

51
00:02:56,480 --> 00:03:01,757
adoption. Second, make all government
data AI ready and stand up AI data

52
00:03:01,840 --> 00:03:05,597
infrastructure to enable scaled
implementation. And third, Congress

53
00:03:05,680 --> 00:03:07,277
should invest to position data dominance

54
00:03:07,360 --> 00:03:10,077
as a national priority. The next theme

55
00:03:10,160 --> 00:03:12,717
is unleash, meaning we must unleash AI

56
00:03:12,800 --> 00:03:15,197
technology and establish an agentic

57
00:03:15,280 --> 00:03:17,837
government. An agentic government is one

58
00:03:17,920 --> 00:03:20,717
that uses AI under human supervision to

59
00:03:20,800 --> 00:03:25,197
enhance its operations. For example,
federal agencies could leverage AI to

60
00:03:25,280 --> 00:03:29,037
streamline veteran healthcare paperwork,
improve improve fraud detection at the

61
00:03:29,120 --> 00:03:32,877
IRS, and boost efficiency
and information sharing across agencies.

62
00:03:32,960 --> 00:03:36,717
This will improve the lives of public
employees and the American people.

63
00:03:36,800 --> 00:03:41,357
Congress require each agency to set up
at least one flagship agent government

64
00:03:41,440 --> 00:03:45,277
program. Next, we must
maximize the ability of

65
00:03:45,360 --> 00:03:49,117
companies to innovate. I believe
the right regulatory framework is one that

66
00:03:49,200 --> 00:03:53,357
allows for innovation while still
creating proper guardrails. Congress

67
00:03:53,440 --> 00:03:57,597
should take three actions. First,
confirm a use casebased regulatory

68
00:03:57,680 --> 00:04:01,677
framework and conduct an analysis
to address regulatory gaps. Second,

69
00:04:01,760 --> 00:04:06,237
establish one single federal AI
governance standard to avoid patchwork

70
00:04:06,320 --> 00:04:10,477
legislation at the state level.
And third, implement policies that enable

71
00:04:10,560 --> 00:04:14,157
American workers to become the AI
workforce of the future. These policies

72
00:04:14,240 --> 00:04:17,677
would provide the skills necessary
to train, fine-tune, and evaluate AI

73
00:04:17,760 --> 00:04:20,557
systems. The final theme is promote,

74
00:04:20,640 --> 00:04:22,717
meaning we need need to promote US

75
00:04:22,800 --> 00:04:27,597
technology globally. Countries around
the world, what I call AI geopolitical

76
00:04:27,680 --> 00:04:31,917
swing states, will soon be forced to
choose between Western or CCP controlled

77
00:04:32,000 --> 00:04:35,997
technology. To help make sure they
choose western technology, Congress and

78
00:04:36,080 --> 00:04:39,757
the administration should empower NIST
to complete all relevant measurement

79
00:04:39,840 --> 00:04:43,517
science for AI and export it to the
world through the global network of AI

80
00:04:43,600 --> 00:04:47,677
safety institutes.
America led the industrial revolution,

81
00:04:47,760 --> 00:04:52,717
the space race, the internet a and the
internet age. AI is the next frontier

82
00:04:52,800 --> 00:04:56,877
and with your assistance, I'm confident
we will lead again. Thank you for the

83
00:04:56,960 --> 00:05:01,200
opportunity to be here today and I
look forward to your questions.