00:00:00
Great to see everyone and I'm thrilled to be able to talk about the issue of
00:00:04
the day and that is artificial intelligence and AI in our world. Um David, Michael, I'd love you to talk
00:00:12
about what where we are right now in terms of the pursuit to be the number one uh lead AI country. How are we
00:00:20
doing, David? >> I think we're doing great. Um Maria, last year uh President Trump gave a
00:00:26
major AI policy speech. is in July and he declared that the United States had to win the AI race. Uh he he had first
00:00:33
of all declared that we were in one. Uh and I think his speech was reminiscent of when President Kenny declared that we
00:00:40
were in a space race and had to win that race. I think since then what you've
00:00:44
seen is that American companies have only innovated more. You're seeing all sorts of really incredible products
00:00:50
being released all the time. I think that um American uh uh AI models, chips, um data centers only just keep um
00:01:00
getting better and better. And so I feel very good about the American position in
00:01:04
this AI race. Certainly we have some very uh you know competent uh and formidable competitors. Um China
00:01:11
obviously has a lot of very smart people working in this area. But I do think that uh just what you see from uh
00:01:18
American companies in Silicon Valley right now is really incredible. And yet there are still so many questions about
00:01:23
all of the spending underway uh to build this out with regard to data centers. And of course the question keeps coming
00:01:29
up are we spending too much. Will we get the return on investment? How do you see
00:01:33
that? >> I I think that we will. Um I think that the reason why you're seeing this huge
00:01:39
infrastructure buildout is because the demand is ultimately there. I I know a lot of people worry and about whether
00:01:46
this could be like a dot situation. And remember where we had the whole fiber build out in the late '9s and then we
00:01:51
had a doc crash. The difference here is that uh in the late '9s and early 2000s
00:01:57
we had a a problem known as dark fiber where you had this fiber buildout and then it didn't get used. There's no such
00:02:03
thing as a dark GPU right now. Every GPU that's being put in a data center is
00:02:07
getting used. Uh and it's being used to generate tokens and that's to power the
00:02:11
this new generation of AI chat bots or coding assistants. uh and there's just
00:02:16
been some releases in the last couple months on the coding front that you know it's if you're following what develop
00:02:22
software developers are saying they're saying it's mind-blowing it's completely
00:02:25
revolutionizing their industry so demand for tokens just increases and that increases the demand for this data
00:02:32
center buildout that we're seeing so I don't think it's going to stop anytime
00:02:35
soon and just last year this infrastructure buildout added about uh 2% to the GDP growth rate and I and
00:02:42
that's what helped propel us to this you know four to 5% % growth rate and I
00:02:46
think you're going to see something similar this year. >> Well, it is certainly leading growth,
00:02:50
Michael. Um, and I'm so happy to be able to get this conversation going with both
00:02:55
of you who are really leading this. David, thank you. And Michael, thank you. Same questions for you, Michael.
00:03:00
Assess where we are right now on AI. >> I I think just a reminder for the group
00:03:03
for those who haven't been tracking as closely as we do every day. The the plan
00:03:07
really had essentially three pillars and it talked about how one, how can the US
00:03:12
continue to out innovate our competitors? two, how can we drive the infrastructure build that we need to
00:03:17
support um this this AI revolution? And three, how do we actually share with the
00:03:22
world or export our great American technology? And for each of those three pillars, there was quite a lot of
00:03:27
actions that the federal government has taken to drive that forward. Um, and I think I think we're pretty proud to say
00:03:32
that we've made, I think, pretty good progress on on all three. Um just focusing a little bit on the innovation
00:03:36
one you were talking about earlier. I think the the the the the core um insight that we've always had about how
00:03:42
you drive this innovation is you have to have a regulatory environment that allows this technology to be developed
00:03:49
and ultimately commercialized in the United States. And the US has done a great job compared to the rest of the
00:03:55
world on sort of setting that up and creating a framework that works. But we can always do better and improve it. And
00:04:01
the president in his his speech in July talked a lot about um this issue of a patchwork of state regulations and how
00:04:08
can we ensure that there aren't 50 different rules around AI and and what's
00:04:13
most what's most important about this debate which I I think a lot of people
00:04:16
sometimes don't sometimes miss is the patchwork is actually most detrimental
00:04:21
to early stage young companies and entrepreneurs. If you want to develop a new AI technology, if you want to build
00:04:27
something on top of one of our great frontier models, having to figure out how to navigate 50 different rules
00:04:32
across 50 different states creates a lot of friction and ultimately the big guys
00:04:36
are the ones that can succeed in in that environment the best. Um, so we're spending a lot of time trying to think
00:04:40
about how can you create a legislative proposal that can actually um deliver on um a sensible national framework to
00:04:47
solve to solve this regulatory issue. So, so what would you say then, Michael, are the basic frameworks that are uh
00:04:54
sort of must-have in in that kind of federal oversight? Because some states did push back in the US and say, "No,
00:05:02
no, we want to be able to control our destiny when it comes to AI." What's
00:05:06
most important when you look at that framework in terms of um a federal oversight? Yeah, I think in the
00:05:13
executive order the president signed in in December directing us to kind of work
00:05:17
through this proposal, he listed a few things that um the state should continue to be able to pursue individually on
00:05:23
their own. Um legislation or rules around child safety was on that list. Um the rules around permitting of data
00:05:29
centers and buildouts are continuing to be something that states should should look at. So there are a few things that
00:05:34
were enumerated, but that's the kind of stuff that I guess Dave and I are going
00:05:36
to be working through. I don't know if you have any thoughts on on that. Yeah,
00:05:39
I mean I I think the the basic problem that we have is that I mean frankly the states are going hog wild right now with
00:05:46
regulation. There's over,200 bills going through state legislatores right now. I
00:05:51
think it's very much a knee-jerk reaction. I know there's a lot of fears
00:05:55
and concerns about AI, but it seems like for every hypothetical concern, there's
00:06:01
um multiple state bills now to try and regulate that thing before we really know how it's going to play out. And um
00:06:07
I think it would be better to I think since this technology is so new and the environment is so dynamic, I think it'd
00:06:13
be better to spend a little bit more time studying how AI is actually being used and what risks are actually
00:06:19
materializing before you overregulate the thing. But in any event, that that's
00:06:23
the what we're seeing right now at the state level. And um and and I think that
00:06:28
the president's been very consistent that it would be better to have a single
00:06:32
have one rule book, a single rule book at the federal level, lightweight federal standard. Uh I think this
00:06:37
problem is only going to get more acute over time because again you you as you have 50 different states running in 50
00:06:44
different directions, the patchwork problem only gets um more significant. So, in any event, this is something that
00:06:49
we're going to work, I think, closely together on this year, which is to see
00:06:53
if we can get enough consensus on a federal framework to enact a law. Only Congress can ultimately cramp the
00:07:00
states. We understand that. Um, and you know, as you know, it's very difficult
00:07:03
to get a bill through Congress. You need 60 votes in the Senate. So, has to be bipartisan to to a certain degree. So,
00:07:10
but we're going to try and see if we can um work to get that consensus. >> Yeah. And do you have any clarity on the
00:07:15
timing on that in terms of um support in Congress for a federal oversight or do you see push back there as well
00:07:23
depending on the state you're talking about? >> Well, there's push back in Congress to
00:07:27
the idea of preeemption without a federal standard. So, in other words, you can't replace something with
00:07:33
nothing. This is sort of the the thing that we heard uh repeatedly. But I think there is uh quite a bit of interest in
00:07:39
both the House and the Senate towards having again some sort of lightweight federal standard. But we're still in the
00:07:44
early stages of those conversations and we're going to see what we can try and
00:07:47
get done this year. >> Meanwhile, you've got some people pushing back after wanting to see the
00:07:52
innovation and growth of data centers. Now they're saying not in my backyard.
00:07:56
What about that? Is that an issue? >> Yeah, I mean we got a letter recently
00:07:59
from Bernie Sanders saying stop all data centers, all data center development. And you know if we do that we will lose
00:08:06
the AI race. I mean you do need this infrastructure. Uh other countries are building out this infrastructure.
00:08:11
China's building out I think they're um spinning up a a new uh nuclear power
00:08:16
plant or um or coal plant new energy every single week and a lot of that is going to power their data centers. So it
00:08:24
would fundamentally I think uh the United States in the AI race if we just stopped building data centers
00:08:29
altogether. At the same time, there are concerns about affordability, about um whether consumers would have to pay a
00:08:36
higher electrical rate because of data centers. Uh President Trump's been really clear that consumers should not
00:08:42
have to pay higher rates for electricity because of data centers. You saw just last week Microsoft stepped up and made
00:08:47
a pledge that it will that its data centers will not cause residential rates to increase. I think you'll likely see
00:08:55
other tech companies stepping up and making similar commitments. And in fact, when I've talked to the hyperscalers and
00:09:01
when I've talked to the AI companies, it was never their plan to draw off the
00:09:05
grid. They all are uh saw standing up their own power generation as part of their buildout. Um and what Secretary
00:09:13
Wright, the our Secretary of Energy has been doing is trying to um is is reform the regulations that actually make it
00:09:20
more difficult for these AI data centers to stand up their own power behind the meter. So that basically is is our
00:09:26
vision is let and and I should say this is President Trump's vision really since
00:09:31
the beginning of the administration is he said let the AI companies become power companies let them stand up their
00:09:37
own power generation as they built you know side by side with these new data centers and the um the result of that is
00:09:44
you know a we get this infrastructure b residential rates don't go up. Yeah,
00:09:48
because Michael, this this race has fast become it's moves from an AI race to a
00:09:54
power race. >> And and I think what we're seeing is that um we we need to share a good story
00:10:02
about how ultimately this buildout is going to be net positive for American rateayers. And I think sometimes if you
00:10:08
know and if you're in a small community and someone shows up to build the data
00:10:11
center, I mean, you have to make it clear that ultimately this something is going to actually lower your your rates
00:10:15
long term. Um and and the president put out a tweet a truth last Monday where he
00:10:19
was as as David said very clear that you know if you're going to build a data
00:10:22
center you have to pay your own way for it and um Microsoft has stepped up and our our hope is that many others will do
00:10:27
the same. >> But but some companies um because they don't have the cash right now are
00:10:33
borrowing money right to to build out the data centers and there's also a worry that the banks will be left
00:10:39
holding the bag for some of this because again the spending is is too much. your
00:10:44
thoughts on that? >> Well, I think there there is obviously that concern. I mean, you you know, I I
00:10:50
think it's it's um it's less I would say the banks are more you see Oracle making
00:10:55
a huge investment. You see, uh Blackstone making huge investments, real estate companies. Um ultimately, I think
00:11:01
these are very savvy market players, very deep companies, and they're doing this because they see an ROI there at
00:11:08
the end of the the rainbow. Um can I make one other point about just the the data center? So, um, just on
00:11:14
electricity, uh, I actually think that if we allow the data centers to stand up their own power generation, it will
00:11:21
actually bring down rates. Not only will it not increase residential rates, it'll
00:11:25
bring it down. And it'll do that in two ways. One is that the data centers can
00:11:30
can give or sell power back to the meter when they have excess. So, that will help bring down rates. Second, there's a
00:11:38
lot of fixed costs involved in power generation. It's not all variable. So when you're able to amortize those fixed
00:11:44
costs over a greater supply, you bring down the meter rate for everybody. And so there's huge economies of scale. So
00:11:51
the more scale you get in electricity, like most other things, the price comes down. That's what So it's actually a
00:11:57
good thing that the uh that we have this buildout going on because it will ultimately reduce prices for consumers.
00:12:05
But we do have to make sure that these new data centers aren't just plugging
00:12:09
into the grid and using, they have to be contributing back. >> And I think what a great policy change
00:12:14
has made under this administration, the the B administration had as a matter of policy had made it such that you
00:12:19
couldn't do this behind the meter energy generation. Um, and if you wanted to
00:12:23
bring your own power, you couldn't. You had to be part of the larger grid. So I
00:12:26
think um that rule has has changed uh uh by by Secretary Wright and by FK to kind
00:12:31
of allow this to happen. And ultimately I agree with David. I think once you have sort of greater scale in in the
00:12:36
power generation, you'll be contributing back into the grid in a way that that
00:12:39
benefits rateayers. >> Let's go back to the uses and how AI is changing our lives. You you mentioned
00:12:46
earlier um all of the uses and and and the impact the AI is having. What do you see as the most important use and where
00:12:54
AI is being deployed and implemented best right now? >> Well, it's interesting. I think it
00:12:59
there's been an evolution. So I think we started with you know AI chat bots like
00:13:04
chatgpt and in a sense that was kind of like better web search. Um it was really
00:13:09
great for research asking it questions and give you answers to anything. Then we saw um we saw models add chain of
00:13:16
thought and they could start to do you know deeper reasoning. Then we saw coding assistance. This is really and
00:13:22
and I think over the past few months there's been a real breakthrough. If you
00:13:26
talk to people, software developers, it really seems like there's been, you know, a major shift in in just
00:13:31
improvement in the quality of the coding assistants. And I think where that's
00:13:35
going next is um tools for knowledge workers. So the same types of assistants that have been outputting code can now
00:13:44
output any type of format. So whether it's like Excel models, PowerPoints, websites, you name it, knowledge workers
00:13:52
are now going to be able to generate all these different types of things the same
00:13:56
way that coders have been gen that software developers have been using AI generate code. I think that's one of the
00:14:01
big things you're going to see in 2026 is again just this um this productivity
00:14:07
boom for knowledge workers. So I think that's like one of the things you're
00:14:10
seeing on the ground. And then separately there's there's a bunch of things happening in industry verticals.
00:14:16
So different industries being impacted by AI. So in healthcare I think there's
00:14:21
a tremendous opportunity uh to improve um or to to reduce sort of administrative bureaucracy to uh to
00:14:30
improve this um processing of paperwork that happens. also to use uh AI and medical and scientific research to help
00:14:37
find new uh cures. You're already seeing users tell all sorts of stories about uh
00:14:42
diagnosis. They've been able to put in their medical records into chat GBT or
00:14:47
you know other chat engine or chat bots and get like remarkable results. They've
00:14:52
been able to you know finally figure out what was you know what what was wrong with them and they've been able to take
00:14:57
that to a doctor. You have doctors using it too. So medical I think is a really interesting area but there's a whole
00:15:03
bunch of these um examples of different different industries are now being impacted.
00:15:08
>> The the one area I think a lot about is is AI for science and and back to to to
00:15:13
David's initial point about the progress we've seen these frontier models. I
00:15:16
think the very early ones sort of started with just general knowledge and you have to go back and understand like
00:15:21
why and the question was what was the data available for those model builders to start training their models and for
00:15:27
the early ones you could just scrape the internet and just kind of cram everything into a model and train it and
00:15:32
and that's where you kind of had this this first phase of of large language models and the second one was coding and
00:15:38
if you think about how do you get a really good coding model you again you have to trade it you have to train it on
00:15:42
existing code and that's again something that is you know relative atively easier
00:15:47
to to acquire than other types of data and you saw great progress and jumps in in the coding models. I think the the
00:15:52
third big sort of shift that hasn't really been touched on yet which the government itself is trying to do a good
00:15:58
uh push on is the AI for science question and why it's so challenging for scientific discovery to like tie in with
00:16:05
the way that LMS are are traditionally trained is that the science data is extraordinarily fragmented and it's not
00:16:12
done in a way or formatted in a way that um can easily be applied to a large language model sort of like training run
00:16:18
and if you think about scientific discovery it's spread out across so many different disciplines. You have
00:16:22
chemistry data, you have math data, you have material science data and all of that is is all types of different
00:16:28
formats. And our effort in administration um we launched something called the Genesis mission which are is
00:16:34
our attempt to sort of make these big bold leaps in AI for scientific discovery. and our national labs at the
00:16:41
Department of Energy are have been doing incredible research over the last you know 50 60 years and all of that has is
00:16:48
sitting and is ready to be used to be trained for for for these models. So my hope is that over the next year we're
00:16:53
going to see a lot more work in this in scientific discovery to be able to actually accelerate how quickly we can
00:16:58
choose which experiments to run, run those experiments, go back and figure out what we did wrong and run them
00:17:03
again. And and this ties in with lots of interesting ideas that people have around some of these AI labs where you
00:17:09
essentially have you can put in the the the thesis or the hypothesis and ultimately these labs can do lab
00:17:14
experiment itself and move forward. So that's kind of the dream that I have that that ultimately we as a country can
00:17:19
can almost double our our R&D output over the next 10 years because of AI. >> So so what kind of breakthroughs um
00:17:26
would you expect or would you like to see? Yeah, I I think there um the ones that I think can make a big impact are
00:17:34
uh first the the the the experimentation and training runs around fusion um are extraordinarily computationheavy
00:17:43
and they themselves if we can if we can have a a a faster feedback loop on how we do these these simulations for fusion
00:17:51
we can move the timelines in for fusion. So that could be a big a big step. Material science is also a very a very
00:17:57
big area where you want to be able to test all types of of different molecules and how interact with each other. This
00:18:02
is important for all the big things we're trying to do in space. Whether it's our lunar base or getting to Mars
00:18:07
or bringing nuclear energy to space, having advanced material science is important. And the third is one that
00:18:12
everyone always cares about is is healthcare and and therapeutics. How can you more quickly be able to identify the
00:18:18
the best molecules to solve a particular particular health challenge? and how do
00:18:22
you more quickly iterate to a point where you can move to a to a clinical trial >> and on a everyday level I mean you also
00:18:29
have the auto sector I think as a big beneficiary here I think that's one area
00:18:33
that seems to be spending a lot on this as well you agree with that >> well I mean with like self-driving or
00:18:41
>> I mean self-driving for sure is going to be huge it feels like we've hit some
00:18:45
sort of new inflection point there where the quality's gotten to the point where
00:18:49
you're starting to see robo taxis now Whimo and Tesla. Um >> what what about an AI assistant? I mean,
00:18:56
is that going to be something that is sort of common place? I someone said to me the other day that oh, in China we're
00:19:02
doing things so much differently because you're using AI for research as as you
00:19:08
said, but we're using it as I have my AI assistant and I'm um you know, they're
00:19:14
paying my bills and cleaning my house and buying my wife a birthday present and and doing everything for me. I I
00:19:20
think so. I think that'll happen probably this year. So the the product that just came out recently that
00:19:26
everyone's kind of going crazy over is the latest iteration of flawed code uh
00:19:31
which is uh powered by uh anthropics uh opus 4.5 model which seems to be a real breakthrough in in in coding and so
00:19:40
again this is you know the software developers are really imp impressed with it but in inside of cloud code they had
00:19:46
they introduced a new tab called co-work again you can as a non coder uh or as someone who is looking for um
00:19:56
to create output other than code, you can now use it to uh to basically create all sorts of other kinds of outputs.
00:20:02
Like I mentioned, you can do uh spreadsheets or powerpoints, things like that. And you can have it, you can point
00:20:09
it to your file drive and it can look at the work you've already done. So if
00:20:14
there's a particular type of format for a PowerPoint you like, you just point it
00:20:18
to the work you've already done and say I want to do you know a new um you know
00:20:23
presentation but using this style but on this topic and it'll actually emulate
00:20:29
you know your style and and the work your format the work you've already done. And um people are very impressed
00:20:36
with this and you can also point it at your email and have it analyze your email pull things out of it. So it right
00:20:42
now it's very taskbased. You you the user have to prompt it for each task. But you can see there the beginning of a
00:20:49
personal digital assistant where you connect it to your file drive to your email to all of your data sources and it
00:20:57
can start to do tasks for you and again it understands the format and the style that you like to produce work in. So, it
00:21:05
feels to me like we just need one more layer of abstraction on top of a tool like that and you'll have your own
00:21:13
personal digital assistant >> and um you know there'll be like a voice
00:21:17
interface. You ever seen the movie Her you know with uh walking Phoenix and um I think Scarlett Johansson is just the
00:21:25
voice but uh you know he's telling her what what to do through an earpiece. I
00:21:29
mean we're very close to something like that. I mean, I'm not saying that, you
00:21:32
know, the AI is going to become sensient or whatever, but um but no, we're like I
00:21:36
think in 2026, you could see that that these types of of tools again started as coding assistants, but now they become
00:21:44
personal digital assistants. That could definitely happen this year. >> Michael, what what don't people
00:21:49
understand about AI? What what do you think is most important for us to understand about the innovation underway
00:21:54
right now with science and and AI? I I think some people I I think it's easy to
00:22:00
underestimate the the long-term impact this is going to have across so many industries and and domains. Um I think
00:22:07
very much, you know, it's easy to to to quickly think about AI as a as just a
00:22:12
sophisticated chatbot because that's what most people interact with every day
00:22:16
and and that's what they they they touch and feel. Um, but I think that to me I
00:22:20
think the long-term impacts and not to keep harping on the science, I think there is a there's a a real fundamental
00:22:25
shift happening in the velocity and pace that we can test and uh and evaluate and
00:22:31
execute scientific discovery and endeavors. And I think I think that's going to have huge repercussions for the
00:22:36
way that we as a country innovate broadly speaking in the years ahead >> which is why we're watching what China
00:22:41
is doing. Let's talk a bit about China and where it is relative to the United
00:22:45
States. Are we winning? Is it about chips? What's the race specifically really about?
00:22:50
>> Well, I I I think that in general we're ahead of China. There's different layers
00:22:55
of the stack. So, you've got the the the models, then you've got the chips, and
00:23:01
you know, then you've got the chipm equipment, you know. So, you go down the
00:23:04
stack. I would say that the deeper in in the stack that you go, the greater the American advantage. Um I think on models
00:23:10
most people would say that we're our models are maybe 6 months ahead or so plus or minus of the Chinese models. You
00:23:18
look at chips maybe 2 years ahead. You go to the semiconductor manufacturing equipment it could be like 5 years. So
00:23:25
the US does have sign significant advantages there. There's only maybe a couple of areas where I think China has
00:23:33
has an advantage. Um one is on energy production. And if you look at the their grid, their grid has roughly doubled in
00:23:41
the last 10 years, whereas ours has only grown by about 2 to 3%. Energy production in the US has been a
00:23:49
relatively sleepy industry before AI came along. And a lot of that had to do with regulations and the antipathy of
00:23:57
the previous administration towards energy production. Obviously, President Trump had a very different view on this.
00:24:03
I think he was preient on this issue. you go back 10 years and he was talking about we got a drill baby drill and um
00:24:10
and I think he understood that energy growth was the precondition for economic growth and it's definitely the pre
00:24:18
precondition for this uh AI infrastructure growth. So this is an area where again we have to basically
00:24:24
expand our energy production um and I and and and so I think that is an area where we need to catch up. The other
00:24:32
area where I would say, you know, I I don't know if I would call this an advantage exactly, but if you but you
00:24:40
could argue that China has the edge in what is what's being um called AI optimism. So there was a a
00:24:50
polling done by Stanford across countries and they asked the citizens of all these different countries uh do you
00:24:57
feel that the benefits of AI will be more beneficial or more harmful? And if if you thought that it that overall be
00:25:05
more beneficial than harmful they call that AI optimism. Well in China AI optimism was 83%. So 83% of the
00:25:13
population feels that it's being more beneficial than harmful. That number in
00:25:16
the United States is only 39%. So for some reason people in China are more optimistic about AI than in the
00:25:25
United States and you generally you generally see this that uh Asian countries are very high on AI optimism
00:25:32
in the western countries are lower and I think it's a interesting or open question about why this is. I think
00:25:39
there's a few possible explanations for it. I I think that um first of all, the
00:25:43
the media tends to focus on the doom and gloom stories with with AI, >> the fear,
00:25:48
>> the fears. Um and we can talk about some of those fears and uh and and how, you
00:25:53
know, whether we think they're they're real. Um but I think the media has a lot
00:25:57
to do with it. I think that the the way that Hollywood has portrayed AI over the
00:26:03
decades, you know, with whether it's the Terminator or 2001, uh has you has portrayed this dystopian view of the
00:26:10
future. And I think that plays into fuel's thinking. And then frankly, I would say that part of the the fault
00:26:16
lies with our tech leaders who haven't necessarily done a great job describing
00:26:21
the benefits of AI. In fact, when they're talking about, you know, AI eliminating 50% of knowledge workers,
00:26:28
that doesn't sound like a, you know, very utopian scenario. That sounds dystopian to most people. And so I do
00:26:34
think that unintentionally some of our tech leaders have played into this um AI pessimism. And the reason why I think
00:26:42
this could be a disadvantage for the United States is because again it's feeding into this regulatory frenzy
00:26:48
we're seeing again 1,200 bills at the state level. And right now I think you
00:26:53
know we are winning this AI race. We're ahead in all the key dimensions chips
00:26:57
models and so on. But we could shoot ourselves in the foot, you know, if we end up overregulating this thing to to
00:27:03
death, we could actually cost ourselves this AI race. So, I do worry about this question of AI optimism,
00:27:11
>> right? It's a great point. And how what would happen if the US is not number one
00:27:14
in this, Michael? >> Yeah, I I I think we we need to be and that's why we put put the plan out. I
00:27:19
think you know when I think about the the China question and about the the sort of larger question of how do we win
00:27:24
the AI race what always what I always like to think about is this question of adoption and I think sometimes there's
00:27:30
this overemphasis on the leaderboard it's like which frontier model is number
00:27:34
one on some sort sort of metric and the reality is we're neck and neck and as
00:27:39
David said we're probably had you know six to 12 months on our frontier models
00:27:42
but I think what we have seen over over time and over history is that um you don't necessarily need to have the very
00:27:48
best model or very best piece of technology in the world for it to perforate globally. And a lot of us who
00:27:54
were part of the first Trump administration saw this very firsthand with the telecom wars of that era of
00:28:00
what Huawei was able to do globally. And at the time when when Huawei first started their their sort of global
00:28:05
export push um they certainly were not the very best technology in the world. They were current they were certainly
00:28:12
you know you know subpar compared to to Ericen and Nokia. that they were good enough and they were subsidized enough
00:28:18
such that they became sort of the default telecom um system for a lot of the world and we've learned a lot of
00:28:23
lessons from that and we take that very seriously. When it comes to AI, we know there's ambition for the Chinese to
00:28:29
export their models and have them be the models that are powering all these different use cases across across the
00:28:36
global south and across the rest of the world. Um, that's why the president launched something called the American
00:28:40
AI export program and our mission and I think we're in a very lucky position
00:28:43
here compared to what we're dealing with with Huawei is as David said, we are
00:28:48
dominant in almost every part of the stack. We have the very best models. We have the various applications. We have
00:28:52
the very best chips. So, we are in a position of power now and is up to us as a country to share that technology with
00:28:58
the world with all of our partners and allies. make sure that any developer anywhere in the world that wants to
00:29:03
build a new application using AI is using is fine-tuning an American model on top of an American chip. And that
00:29:09
isn't that isn't a a hard reality to see. That is something that I think we
00:29:12
can very easily do just because we have the very best tech. That's a program
00:29:15
that um we launched last late last year and we're doing a big push this year to
00:29:19
get that get that out the door. >> It's an important point that you make in
00:29:22
terms of exporting AI to the rest of the world. Is it true that China is telling
00:29:27
its companies don't use American chips, don't use American AI right now?
00:29:33
>> It it it seems so. Um I mean China is developing its own models. Obviously
00:29:37
about a year ago you had the Deepseek moment where you you had a powerful model released by Deep Seek and I think
00:29:43
that kind of put Chinese uh AI on the map in a way. I think people in the west didn't realize you in a way how good
00:29:51
China was at producing models and there was a little bit of complacency uh towards our relative position. People
00:29:59
weren't really talking about the global competition two years ago. It wasn't
00:30:03
really discussed at all. Uh I remember when you know the Biden administration created this you know 100page Biden
00:30:09
executive order regulating AI. No one was talking about whether this might slow whether all this regulation would
00:30:15
slow us down. Visa v China wasn't even part of the conversation. Then Deepseek
00:30:18
launched and I think we did realize we're in a global competition and we have to win and that's why we have to
00:30:23
actually be quite careful about how we regulate this and not make sure we're
00:30:26
not overregulating it. But I think you know China definitely wants to compete. Um there have been some stories recently
00:30:35
I think uh Bloomberg and Reuters reported that they actually are not allowing Nvidia chips into their country
00:30:41
and the reason for that we think is that they want to indigenize chip production.
00:30:46
They want to stand up Huawei as their national champion and effectively they're creating a market subsidy for
00:30:52
Huawei by keeping out the competition. So, they're protecting their market to
00:30:56
stand up Huawei. And I think their plan would be to have Huawei dominate chips in China first and then use that to
00:31:04
scale up and then try to take over the rest of the world. Chip production is a scale up business. So, you know, if they
00:31:11
can dominate the Chinese market first, that gives them a powerful platform to then proliferate to the rest of the
00:31:16
world. >> So, so where are we in that, Michael? I mean, first you all came up with the AI
00:31:21
action plan, then came up with another plan in terms of exporting AI to the rest of the world. What can you tell us
00:31:27
in terms of where we are in that? >> Yeah, so the the progress is is moving
00:31:30
on that. We um we closed a request for information from the commerce department late last year, which went out to
00:31:36
industry and said, "Hey, if we want to export the American AI stack, what should we be thinking about? How should
00:31:40
we be designing these packages that we share with the world?" Commerce is now
00:31:43
ingesting that that information. There'll be a request for proposals that comes out very shortly. And that's where
00:31:48
we actually want companies to come together to form consortia and say like look this is what a package looks like.
00:31:54
And I think what um you know what what people need to sort what I always try to remind people is that the the the the
00:32:01
buyers of AI around the world um vary quite dramatically in their level of sophistication. So in the US, if you're
00:32:09
a very sort of, you know, if you're a Fortune50 company and you want to deploy
00:32:13
AI, you have a pretty sophisticated sort of CIO or CTO shop, you are thinking very carefully about like which cloud
00:32:19
you want to buy, which potential model you want to use, do you want to fine-tune it on your own data? Do you
00:32:23
want to build your own application? You know, what application you go and see? You can like test various things. You
00:32:27
like go to all these third parties and evaluate which is best. And it's a very
00:32:31
sort of complicated mix of how you end up creating something that's optimum for
00:32:34
your particular company. for a lot of countries around the world that are aspiring to to use AI for their people
00:32:40
or to support the services whether it be health care or um you know tax collection or whatever it may be um you
00:32:46
know they don't have a a you know billion dollar IT budget you know they're just trying to figure out what
00:32:52
is a tool that I can use in my country to deliver the benefits of AI to my people so we think very carefully around
00:32:58
how can we craft solutions which you know turn keys could be one way to put it or how do you how do you provide a
00:33:04
solution that can easily be deployed in a country and what's often you know what
00:33:09
often sort of gets caught up in this debate is this question of you know how many chips is the US going to be sending
00:33:14
around the world and and what I always try to remind people is that you know outside of the US China and maybe a few
00:33:20
other countries most countries around the world do not have the capital or the aspiration to do largecale training runs
00:33:27
or development of their own frontier models there are very few countries around the world that are going to build
00:33:32
sort of colossus style training centers most countries around the world need smaller data centers that just have
00:33:37
inference related chips that can drive and and and do the you know do the inference on on the particular um runs
00:33:45
that the government wants to have. So I think what we're working very hard to do
00:33:49
is is is create sort of these these these turnkey manageablysized AI solutions that then we can partner with
00:33:57
a lot of our export finance organizations like Development Finance Corporation or the Export Import Bank to
00:34:03
make the export of that particular stack much more appealing and commercially viable in countries that are not
00:34:09
extraordinarily deeped. Um so we're going to be in India next month for the India um AI impact summit. Um, this is
00:34:16
sort of the largest global gathering for for AI folks. Um, and we're going to be
00:34:20
sharing a lot more on on the progress of this uh of this program there. >> You want to weigh in?
00:34:25
>> Well, I just just to build on that. I think people sometimes ask, you know, h
00:34:30
how do you how will you know if if you've won the AI race, you know, with with with China with with other
00:34:35
countries and I think there's a very simple answer to that which is market share. You know, if 5 years we look
00:34:41
around the world and we see that it's American chips and models are being used
00:34:46
everywhere, well that means we won. But if in 5 years we look around the world and it's Huawei chips and Deepseek
00:34:51
models, then that would be very bad, right? That would be a bad sign. That means that we lost. So I do think that
00:34:56
the proliferation or diffusion of American technology is really critical to winning this AI race. We know from
00:35:03
Silicon Valley that the companies that end up becoming huge are the ones that create ecosystems. It's the, you know,
00:35:10
you you you as a technology company, you want to have the most apps in your app store. You want to have the most
00:35:15
developers writing on top of your API. You want to be a platform company. And so in all these technology races,
00:35:24
biggest ecosystem wins. And we want to have the so that's basically why I think
00:35:28
this program is so important is we want to create the biggest ecosystem. Now this is not only about benefiting the US
00:35:35
because in order to have a successful ecosystem you have to create value for your partners and that's really
00:35:41
important like Michael's saying not every country is going to be on the cutting edge of developing its own chips
00:35:46
or developing its own frontier models but they can use these tools to derive value to apply them to their businesses
00:35:54
to their economies to extract value and be part of this technological revolution. So I think that you know we
00:36:00
have to think in this with this partner mindset and I do think that this this type of mindset is actually very common
00:36:08
to Silicon Valley. Like I mentioned I think every great technology company thinks in terms of how do we get the
00:36:13
most people on top of our tech stack but it is a form of thinking that's pretty
00:36:17
alien to the bureaucracy in Washington which has much more of a command and control type of mindset. Y
00:36:24
>> and when President Trump came into office, just give a couple examples of
00:36:28
this, the regulations that were sitting on our desk that had just been handed down by our predecessors, again, we had
00:36:34
this 100page Biden executive order on AI that was all this new regulation and there was a 200page
00:36:42
uh was called the Biden diffusion rule, which was 200 pages of regulations uh on the export of semiconductors. So
00:36:50
we were turning the the AI industry models and chips into a highly regulated industry. That was that was basically
00:36:57
the direction that Washington was going in. And the first thing President Trump did his first week in office was rescend
00:37:04
all of those unust regulations which I think was absolutely critical. You know the thing that really makes Silicon
00:37:10
Valley special is this concept of permissionless innovation. you know, since um Hulin and Packard started 85
00:37:17
years ago, started building Silicon Valley, it's it the idea has always been
00:37:22
that just a couple of founders kind of a great idea start their company, they get
00:37:26
some angel investors to write, you know, a check for, you know, seed capital. Those investors think they're probably
00:37:33
going to lose their money, but they figure there's a shot. and you know and it's so it could be the two guys in
00:37:39
their garage or it could be the college dropout in the dorm room and they don't
00:37:43
need to go to Washington to get permission for their idea right it's permissionless innovation that's what's
00:37:48
has made Silicon Valley the crown jewel of the world it's why so many of the I
00:37:52
think heads of state who are here are always asking how do we create our own Silicon Valley that was not the
00:37:58
direction we were on when President Trump came into office the new 300 pages of regulations concerning AI the Biden
00:38:06
administration left us with would have changed this um environment of permissionless innovation to an
00:38:12
environment of you have to go to Washington to get approval for your idea and I think that President Trump really
00:38:18
corrected that and since then we've been implementing you know his AI action plan
00:38:23
uh which is all about you know pro innovation pro infrastructure pro energy and pro export so it's been I think a
00:38:29
total change and I think just in the past year you've seen the results of that
00:38:33
>> and I think what one thing to to add there Um part of the the international
00:38:38
uh agenda that we have on AI is one obviously let's let's do the export but
00:38:42
the other piece is trying to share with all of our partners and allies how you can actually create a regatory
00:38:48
environment that allows this technology to succeed and here we are in Europe and
00:38:52
I think many of us that sort of have you know tried to work with technology companies in Europe have have hit sort
00:38:58
of a lot of roadblocks and a lot of stumbles and no matter you know the drug reporting came out and and he can say
00:39:04
that there's a lot of issues But things don't ever seem to seem to really
00:39:07
change. And I think all of that that the the the way that our regulatory structure is is designed in the US and
00:39:15
the way that the entrepreneurial spirit thrives in the US is something that we try to share with countries all around
00:39:20
the world. And I think the the the general um knee-jerk reaction for most policy makers around the world is one
00:39:28
that moves to a corner that is obsessed with the precautionary principle. this concept that every time something new
00:39:35
comes out, the role of the policy maker is to sort of like sit in a room and whiteboard everything that could go
00:39:41
wrong and then design regulations to make sure those wrong things, these hypothetical wrong things don't happen.
00:39:48
When in reality, what we do in the US, what we try to do is sit in a room and whiteboard what rules we can create to
00:39:53
actually unlock innovation. What are the ones we should remove to allow more innovation to happen? And I think that
00:39:59
mindset is something that we constantly try to share at all these international fora. The US has you know there has been
00:40:05
an AB test on what regulatory structure works and what succeeds. You know we've
00:40:10
seen how the how how Europe has approached this in the last 20 years and we've seen what the US has done. So I
00:40:15
think the the recipe is kind of obvious but but sometimes we have to just keep repeating it to to our counterparts. And
00:40:20
I love the Draghy report because it was so clearly uh identifying companies that
00:40:26
are in Europe that you know like no Novo Nordus is like 350 billion or $400 billion company and in America we've had
00:40:34
companies of trillion dollar companies Nvidia hitting$5 trillion. So so what is the path to innovation?
00:40:43
Well, I I I think part of it is and I I I think this is the difference between maybe the American mindset and the
00:40:50
European mindset towards this is that ultimately the innovation in the United States comes from the private sector. It
00:40:57
comes from the entrepreneurs, the founders, the innovators, the geniuses with an idea. And I think that the
00:41:04
government sees its role, at least when it's thinking properly about this, as
00:41:08
being an enabler and as just setting the rules of the road. um and maybe putting
00:41:13
in some guard rails, but basically it's letting the entrepreneurs cook and that's how you get innovation. And now I
00:41:21
don't want to bash our European host too much, but you know the when when the uh
00:41:27
when the EU talks about AI leadership, they're talking about the regulators and
00:41:32
they think their value ad is well, we're going to we're going to show the whole
00:41:35
world the regulatory model for AI. So, it's kind of a bad case of u main character syndrome where uh you know
00:41:42
where like the regulators think they're the main characters in this. No, look,
00:41:46
the regulators are the supporting players. The main characters always have to be the entrepreneurs. It's got to be
00:41:52
the innovators. That's how you unlock innovation. When the when you start to
00:41:57
see yourself, I mean, the the regulators and the policy makers as the as the main
00:42:01
characters, that's not a great recipe for innovation. And I think just just a
00:42:05
minor point on the on the AI stuff in Europe that you know the EUAI act which has been so detrimental to to the AI
00:42:12
ecosystem here here in Europe was passed before chat GPT was even invented and that shows the challenge here. You're
00:42:20
you're you're believing that you can solve some kind of problem or some you're solving something but the end of
00:42:25
the day innovation is moving so much more quickly and ultimately that that rule makes no sense now in a world of of
00:42:30
frontier models large language models and they have to sort of edit it. So, let me push back before we go and ask
00:42:36
you to identify any risks or threats or downside risks in all of this. What should we be worried about if anything
00:42:45
with regard to AI usage? >> Well, I I think there are Orwellian scenarios uh of AI that I think we
00:42:54
should be concerned about. And again, I I tend to think that those scenarios were described by George Orwell, not by,
00:43:01
you know, James Cameron and the Terminator. And specifically, it's misuse of AI by government. I do think
00:43:07
that AI could be used as a tool to um surveil, to censor, to even potentially brainwash the population. This is why
00:43:17
the administration has taken such a firm stance against what it's called woke AI,
00:43:21
which I almost think that that name maybe trivializes the magnitude of the problem we're talking about. We're
00:43:27
talking about AI having a political bias built into it. Um, and it the bias can be so subtle that people don't even
00:43:34
necessarily notice over time, but it has a huge impact on what people are allowed
00:43:39
to learn and think and know and what you know children learn. And so I think it's
00:43:44
very important that we try to make sure that AI was politically unbiased. Um there just in this regard, one of the
00:43:52
things that we were so concerned about with that Biden executive order on AI that we were sended in the first week is
00:43:58
that it had 20 pages of language on DEI and it was promoting this idea that AI models need to build in a DEI layer.
00:44:08
Well, you know, this is how you ended up with, you know, the the the black George
00:44:12
Washington uh you know, story where that the first version of of uh Gemini came out and it was, you know, it was
00:44:19
basically rewriting history to serve a current political agenda of DEI. And um you know that that was in a way that
00:44:27
that that case of bias was so ludicrous that everyone kind of laughed at it. But
00:44:32
it gives you a sense of what could happen if you start to build the the bias into AI and you know that same you
00:44:41
know so-called trust and safety apparatus that was starting to be built into social media sites as a way to
00:44:48
censor and deplatform and shadowban. You could see that being built into AI models as a way to control uh the the
00:44:57
public discourse in in a very serious way. And I think that, you know, President Trump again just put a total
00:45:04
halt to that, you know, rescended that. But it was also we also um President Trump signed an executive order saying
00:45:11
that the federal government would not procure politically biased AI. So look, on a first amendment basis, if an AI
00:45:18
company wants its AI to be biased in some direction, they probably have a first amendment right to do that. But we
00:45:26
have as the federal government have the discretion not to buy that software and we've said that we won't. So, I feel
00:45:33
very good that during President Trump's uh term in office for the next three
00:45:37
years, this idea of Orwellian AI is not going to be a problem. But I do worry that at some point in the future, if you
00:45:44
had a different regime in Washington, you know, if the federal government started to pressure AI companies to
00:45:52
build in this political bias, that would be a very serious threat, I think, to to
00:45:57
our freedoms. It's a it's a great point to make. Before we wrap up real quick on
00:46:00
jobs, can either of you explain what Elon Musk is is saying about the impact of AI said we're not going to need to
00:46:07
work. You know, the the AI AI is going to do it all. I I just I'm trying to understand what he's saying that it's
00:46:14
we're going to go on holiday. Um jobs are going away and AI is going to do everything. Well, Elvon's a friend of
00:46:21
mine and um I'll I'll uh I I'll disagree with him slightly on this, but um but
00:46:27
but let me just the his comment about the the job loss obviously is what gets all the
00:46:33
headlines, but at the same time he's saying that he's also saying that in
00:46:37
this future there's going to be so much abundance that everyone's going to have
00:46:40
what they want and there's not going to be any money. So people people leave out
00:46:46
that part of the story and they just report Elon says everyone's going to lose their jobs. No, we're talking about
00:46:50
a radically different future. It could be the future. It's kind of described in
00:46:53
Star Trek, you know, where like there is no money because we have everything. Look, I I think that, you know, Elon is
00:47:01
directionally correct about the future. I think we are heading toward toward uh towards a world of much greater
00:47:06
abundance, rising living standards for everybody, greater productivity. I think that will lead to rising wages. I don't
00:47:13
think it's going to put everyone out of work. I don't think that's going to
00:47:15
happen. Uh but again, the timelines matter a lot. And you know, getting to a world with no money is not something
00:47:22
that's going to happen in the next 5 years. >> And of course, Michael, this is helping
00:47:25
us um in terms of longevity and living longer, right? In terms of the impact on science.
00:47:32
>> Totally. I think generally the the abundance story extends itself well into
00:47:36
into, you know, health care and everywhere else that and and just quality of life. So, good things ahead.
00:47:42
I think >> we'll leave it there. Michael Katzios and David Saxs, thanks so much.
00:47:46
>> Thank you.