00:00:00
>> Closing bell overtime for
00:00:03
eastern CNBC.
00:00:05
>> Opening eyes. Top executives
00:00:07
are in Washington, D.C. They're
00:00:09
demonstrating the technology the
00:00:10
company has to lawmakers. Of
00:00:11
course, this is the deep sea
00:00:14
threat so to speak, looms large.
00:00:15
Plus, a source tells CNBC that
00:00:17
OpenAI is looking to raise more
00:00:19
cash, this time from SoftBank.
00:00:21
Kate Rooney is in DC and she's
00:00:24
exclusively with OpenAI's chief
00:00:26
product officer. Kate.
00:00:28
>> David, good to see you. We
00:00:29
have Kevin Wheeler here, chief
00:00:30
product officer of OpenAI. It's
00:00:31
great to see you, Kevin. Thanks
00:00:32
for sitting down.
00:00:33
>> Thank you for having me.
00:00:35
>> You did just get offstage and
00:00:37
demoed some technology. I know
00:00:37
you can't speak about exactly
00:00:38
what you were walking lawmakers
00:00:40
and administration.
00:00:40
>> Folks through, but what.
00:00:42
>> Is your message to
00:00:42
Washington? Maybe give us the
00:00:44
big picture takeaway from that?
00:00:45
>> Well, we've long had a policy
00:00:47
of coming to DC and talking
00:00:49
about how AI is evolving. We
00:00:50
want to bring everybody along
00:00:52
with us. These are important
00:00:54
changes for people personally,
00:00:56
also for our nation's economy
00:00:57
and so on. So that's what we're
00:01:00
here talking about. In
00:01:01
particular, 2025 is the year
00:01:02
that agents will become a big
00:01:05
deal. And what we mean by that
00:01:06
is we're going to go from
00:01:07
ChatGPT answering questions for
00:01:10
you to ChatGPT doing things for
00:01:12
you in the real world. So we
00:01:14
launched our first agent last
00:01:16
week, a product called operator,
00:01:18
and it browses the web for you
00:01:20
so it can it can do tasks for
00:01:22
you on the web, whether that's
00:01:23
ordering groceries or filling
00:01:24
out forms. You know, the kind of
00:01:25
things that you just want to
00:01:26
have done for you so you can
00:01:28
work on more important things.
00:01:29
And then we talked about today a
00:01:31
couple of other Agentic products
00:01:32
that we'll be delivering soon.
00:01:33
>> I think that adds a lot of
00:01:35
efficiency. Maybe something the
00:01:36
fed would start looking to. I do
00:01:38
want to talk about the elephant
00:01:39
in the room, which is deep seat.
00:01:41
Can you share any more about
00:01:42
what happened here with Deep
00:01:43
Seat and the recourse for
00:01:44
OpenAI?
00:01:45
>> Look, I think there are a lot
00:01:47
of legitimate and concerning
00:01:48
questions about the origins
00:01:50
here. And at the same time,
00:01:52
there are a couple of novel
00:01:54
improvements that they've made
00:01:55
and what it what it means for me
00:01:57
is like, my takeaway is China is
00:02:00
here. We knew that this that I
00:02:02
was going to be a you know, it's
00:02:04
not just a competition between
00:02:06
US companies. It's ultimately a
00:02:08
competition between the US and
00:02:09
the PRC. This is the first, you
00:02:12
know, big kind of salvo in that.
00:02:15
And AI is going to transform our
00:02:18
personal lives. It's going to
00:02:20
transform our professional
00:02:21
lives. I want an AI that behaves
00:02:25
according to democratic values.
00:02:26
I don't want an AI that behaves
00:02:27
according to authoritarian
00:02:29
values. So we at OpenAI, our
00:02:31
mission is to make sure that US
00:02:33
wins this race.
00:02:35
>> And with those IP issues, you
00:02:36
guys have flat out accused them
00:02:38
of basically ripping off your
00:02:39
technology. I mean, that is the
00:02:40
big elephant in the room that
00:02:42
they may have stolen some
00:02:44
technology to get here. With
00:02:45
that in mind and that backdrop,
00:02:46
what are the technological
00:02:48
advances? What is the cost
00:02:49
implication for OpenAI? How does
00:02:51
it change the way you maybe
00:02:52
think about raising capital in
00:02:53
the future?
00:02:54
>> Well, this is this wouldn't
00:02:55
be the first time that China has
00:02:58
gone after us IP. But for us,
00:03:01
we're focused on building the
00:03:03
world's most intelligent models
00:03:05
and bringing them to people all
00:03:06
over the world in a way that
00:03:08
that improves people's lives.
00:03:10
And we've already, you know, the
00:03:12
we have a new model coming soon
00:03:14
that we've already talked about
00:03:15
that is head and shoulders above
00:03:18
anything that's out there today.
00:03:20
We're focused on that. We're
00:03:21
focused on using that to build
00:03:23
products like agents that help
00:03:24
people be more efficient, that
00:03:25
help businesses operate more
00:03:27
efficiently, that help the US
00:03:28
grow faster. And like, that's
00:03:31
our focus.
00:03:31
>> It doesn't mean you need to
00:03:33
spend less. To be.
00:03:34
>> Clear, no. If anything like,
00:03:36
look, everything that we've seen
00:03:36
over the past ten years of
00:03:38
building these products is that
00:03:39
the more compute you apply, the
00:03:42
more intelligent the model is,
00:03:44
the more things that it can do
00:03:46
for people. So that's why we
00:03:47
were so excited. We announced
00:03:48
last week we were with President
00:03:50
Trump in the Roosevelt Room and
00:03:52
announced this project,
00:03:53
Stargate, where we're going to
00:03:55
put $500 billion towards energy
00:03:58
infrastructure chips in the US
00:04:01
across a range of states to help
00:04:04
OpenAI win. And we're very
00:04:06
excited about that. And as you
00:04:08
as AI becomes a more of a part
00:04:10
of all of our lives, it's only
00:04:12
going to be more important that
00:04:13
we have a lot of access to
00:04:15
compute, that we can build great
00:04:16
products for people.
00:04:17
>> How have conversations gone
00:04:18
with the administration around
00:04:19
this deep sea issue and then.
00:04:22
Export controls? Is that
00:04:23
something you think that would
00:04:24
help if you were semiconductors
00:04:25
were getting out of the US?
00:04:27
>> Yeah. Look, President Trump
00:04:29
wants the US to win. We want the
00:04:31
US to win. And he wants growth
00:04:33
and we want to help drive us
00:04:35
growth. So there's a lot of
00:04:36
alignment. And we're very
00:04:38
excited to work with the
00:04:38
administration. It's one of the
00:04:40
reasons we're here. You know
00:04:40
we're doing an increasing amount
00:04:41
with the government. We just
00:04:44
earlier this week we launched a
00:04:46
product called ChatGPT gov,
00:04:46
which is a version of ChatGPT
00:04:49
specifically for US government
00:04:52
agencies that respects, you
00:04:53
know, security and compliance
00:04:54
and all the things that you need
00:04:56
to do to build a great product
00:04:57
for the US government. We also
00:04:59
today announced a partnership
00:05:00
with the National Labs. So we're
00:05:01
working with Los Alamos, Sandia,
00:05:04
Lawrence Livermore. These are
00:05:07
incredible groups of scientists
00:05:09
that do fundamental physics
00:05:10
research. They do material
00:05:11
science research. They help they
00:05:13
do research that helps safeguard
00:05:15
the nuclear stockpile. And we're
00:05:16
going to bring them our most
00:05:18
intelligent models to help them
00:05:19
do their jobs and advance
00:05:20
science and advance national
00:05:21
security faster.
00:05:22
>> National security seems to be
00:05:24
such a big theme here, and
00:05:25
something officials really
00:05:26
obviously care about. What about
00:05:28
those export controls?
00:05:30
Semiconductors are such a big
00:05:31
part of that is something that's
00:05:32
something you guys have been
00:05:33
lobbying for.
00:05:34
>> Well, if you if you go and
00:05:35
look some of what even what the
00:05:38
deep Sea CEO has said is that
00:05:40
they wish they had more, they
00:05:43
wish they had more GPUs. And
00:05:45
this is why it's so important
00:05:47
for the US to invest in
00:05:49
infrastructure. We need more
00:05:50
energy. We need more data
00:05:51
centers, supercomputers, GPUs
00:05:52
because this is what drives the
00:05:54
AI revolution. And again, this
00:05:57
is us versus PRC and the US
00:05:59
needs to win.
00:05:59
>> I do want to ask you about
00:06:02
SoftBank. I've been hearing from
00:06:02
sources that you guys are in the
00:06:04
middle of raising. It might not
00:06:05
be something you can comment on,
00:06:07
but I do need to ask you, what
00:06:09
is the latest with SoftBank.
00:06:10
>> Do you plan to raise.
00:06:11
>> Up to $25 billion.
00:06:12
>> Well SoftBank is an amazing
00:06:14
partner. They were part of our
00:06:15
announcement of Stargate with
00:06:16
President Trump. It was
00:06:17
President Trump Massa Larry
00:06:18
Ellison and Sam in the Roosevelt
00:06:19
Room. So SoftBank is an
00:06:20
incredible partner. They see the
00:06:22
future the way we do. And we're
00:06:24
very excited to continue our
00:06:25
partnership there.
00:06:25
>> And last question on deep
00:06:27
tech. All of these open source
00:06:28
models, it seems, have really
00:06:29
caught people's attention. If
00:06:31
these are available to download
00:06:33
and tweak for free. What kind of
00:06:34
moat does OpenAI have from a
00:06:36
technology perspective? How are
00:06:37
you guarding against some of
00:06:39
that competition?
00:06:40
>> Well, it's two things. One,
00:06:41
we're about to launch some
00:06:42
models that are meaningfully
00:06:44
beyond the state of the art
00:06:46
today. So we're I mean, that
00:06:47
that in itself, we need to keep
00:06:49
running fast and keep our
00:06:50
research ahead of, in
00:06:51
particular, what's coming out of
00:06:52
China. And then also we don't
00:06:55
just build models, we also build
00:06:56
products. So we have ChatGPT
00:06:58
that's helping people every day
00:07:00
accomplish a huge variety of
00:07:02
tasks. And you've got we've got
00:07:04
products for businesses. And so
00:07:06
it is both increasingly
00:07:07
intelligent models an
