| Early
into Donald Trump’s first term, Eric Schmidt, then-executive chair of
Alphabet, helped change America’s lens on the world — though we did not
know it at the time. I first heard Schmidt’s pitch that the US and China
were in a technological fight to the death at the Halifax International
Security Forum in late 2017. The
gist was that China’s Xi Jinping had just released his bold “Made in
China” strategy. He had also laid out three national goals on artificial
intelligence — to catch up with America by 2020, to make major AI
breakthroughs by 2025 and to dominate global AI by 2030. To Schmidt,
Xi’s plan was a wake-up call. America was in a technological race for
global supremacy that China could win. Here he is delivering that same warning at the Center for a New American Security in November 2017. I
was reminded of Schmidt’s jeremiad this week when China’s DeepSeek
delivered what was instantly dubbed a “Sputnik moment” with the release
of its stunningly good language learning model. Not only had China made
an apparent breakthrough in its targeted year but had done so at a
fraction of the cost of its US rivals. Whether this has been overstated
by markets and technologists because of the DeepSeek shock value, I am
unqualified to say. Every consensus is prone to overcorrection. I do
feel confident to point out that DeepSeek’s announcement shook both
Silicon Valley and Washington’s defence industrial complex (given how
intermeshed the two are nowadays I risk repeating myself). My aim here
is not to evaluate open source versus proprietary LLMs, or to project
where the US-China race goes from here. I don’t have enough knowledge.
It is to point out how remarkably effective Schmidt has been. Remember
that 2017 was the year of peak Silicon Valley notoriety. Companies like
Facebook, as Meta was then known, and Google were branded as the new
“big tobacco”. There was talk of rewriting monopoly law to break them
up. Mainstream Democrats were embracing an EU-style privacy law even
stronger than the one being drafted in California. Big tech behemoths
were suddenly viewed through the frame of an overhaul of competition
law. Then Schmidt repurposed “competition” to mean US-China competition.
I am not saying he was motivated simply by the desire to help his
Silicon Valley peers. I think he was — and remains — sincere in making
the national security case for big tech. He set up a think-tank, the
Special Competitive Studies Project, which churns out briefings on
US-China competition. Either
way, 2017 was the year that Washington reimagined Silicon Valley’s bad
boys as shields against China. Far from regulating big tech, Washington
resolved to treat the west coast titans as weapons in democracy’s
arsenal. That has been one of the propellants behind the outsized market
growth of the Magnificent Seven big tech companies in recent years. The
view was that China and the US are in a race to see which could attain
artificial general intelligence (AGI) first. The country that prevailed
in AI would also triumph in the geopolitical battle. Until Monday, the
consensus was that America was in the lead. Now we are not quite so
sure. As Ryan Grim and Waqas Ahmed observed in this smart essay,
DeepSeek has also single-handedly revived the case for breaking up the
monopolies. Joe Biden’s competition tsar, Lina Khan, was unable to make a
big trustbusting breakthrough during her four years in that job.
DeepSeek just belatedly restated her claim: “Khan warned that enabling
protectionism for tech monopolies wouldn’t just hurt all of us, it would
hurt them too,” they wrote. I
would add that the escalating political antics of Elon Musk, Peter
Thiel and others has made it far easier to depict big tech leaders as
the new robber barons. Once upon a time, people such as Mark Zuckerberg
and Schmidt told us that social media would usher in a new global
community and dissolve social barriers. They have since flipped 180
degrees. Here are the three goals of the Schmidt think-tanks’s recent defence briefing for Donald Trump’s administration. Forge the world’s most dominant fighting force Seize and hold the commanding ground in AI and traditional warfare Resurrect
America’s industrial might by building a 21st century arsenal of
democracy to secure our technological and military edge.
I am turning this week to Henry Farrell, a DC-based scholar, whose Substack Programmable Mutter
is essential reading on these and related questions. Henry, I know you
are sceptical of the idea that the US and China are in a race to the AGI
finishing line. You see the AI story as a less dramatic process of
diffused learning and innovation. I suspect you are in a minority in
Washington which has a bias for Manichean battles. My
question to you is two-fold: how big a surprise is it that it was
autocratic China that produced the start-up shock — and also that its
tool is open source? Second, are the Metas and Alphabets and OpenAIs of
today wasting tens of billions of dollars? |