AI: the key battleground for Cold War 2.0?, by Evgeny Morozov (Le Monde diplomatique

‘The Cold War Is Over’ proclaimed a flashy 1988 marketing brochure for a strange new computer game, an unlikely hit from behind the Iron Curtain. Then came the teaser: ‘Almost…’ Perhaps they got it right, but for the wrong reasons. The brochure, now at the National Museum of American History in Washington, was the brainchild of Spectrum HoloByte, a Silicon Valley software company then owned by Robert Maxwell (father of Ghislaine) (1). Featuring a drawing of the Kremlin with some geometric figures in front, the brochure warned of the ‘Soviet challenge’, announcing that ‘just when East/West tensions are beginning to ease, the Soviets have scored a direct hit on the US.’

That hit was Tetris, the cult computer game, whose Cyrillic name Тетрис – in bold yellow letters – was proudly displayed on the brochure’s bright-red cover page (the Soviet hammer and sickle replacing the final ‘s’). Spectrum HoloByte, the game’s US distributor, leveraged all the right cold-war themes – from Russian folk music to the images of the Soviet cosmonauts – to make Tetris a hit in Reagan-era America. Even then, some in Silicon Valley knew how to cash in on the cold war.

Fast forward to 2023. Gilman Louie, then the company’s CEO, is now a key figure in what some in Washington excitedly call Cold War 2.0 – the ongoing battle between China and the US for the commanding heights of the global economy. As the battle expands to technological and even military fronts, it’s no longer about Tetris. It’s now about artificial intelligence.

‘Unlock the American heartland’

Gilman Louie’s career path is a quintessential American journey. He made his name in the early 1980s designing flight simulator games. They were so successful that the US Air Force got in touch. Eventually, one of Louie’s startups appeared on Maxwell’s radar and he snapped it up.

By the late 1990s, Louie ran In-Q-Tel, CIA’s own venture capital fund, a nonprofit entity whose most famous bet led to the technology behind Google Earth. As the Trump administration began making noises about losing the technological race to China, Louie resurfaced as a member of the National Security Commission on Artificial Intelligence, a high-profile advisory group led by Eric Schmidt, former CEO of Google.

The tech industry and military contractors don’t always see eye to eye. Many American tech companies don’t want to lose the Chinese civilian market due to its sheer size. The military contractors have no such constraints

Just a few years later, that connection with Schmidt has blossomed into a closer partnership – so close that Louie is now CEO of the Schmidt-backed America’s Frontier Fund (AFF), an In-Q-Tel-like nonprofit venture, with a mission to help Washington ‘win the 21st century global technology competition’. The fund styles itself as a panacea, promising to ‘revitalise manufacturing, create jobs, bolster local economies, and unlock the American heartland’. And it has a most impressive board, which includes the ex-CEO of IBM and Trump’s former National Security Advisor.

‘Frontier technologies’

China’s growing influence in the so-called ‘deep tech’ or ‘frontier technologies’ like artificial intelligence and quantum computing is what led to AFF’s creation. ‘Frontier tech can’t be built in a garage,’ proclaims AFF’s website, departing from Silicon Valley’s faith in the genius entrepreneur. Between the novels of Ayn Rand and government subsidies, they choose the latter.

Ironically, Gilman Louie, the man who leveraged Cold War 1.0 to hype up Tetris, is now leveraging Cold War 2.0 to hype up AI. Or perhaps vice versa. In today’s Washington, these two operations have become almost indistinguishable, and the only certainty is that all that hype will be monetised.

To work in the age of AI, the old Tetris slogan needs some polishing. ‘The Cold War Is Over. Almost…’ sounds dated. However, ‘The New Cold War Is Here. Almost…’ is a message that many in America – from tech companies to defence contractors to hawkish thinktanks – find appealing.

For all the rhetoric, some ideological shifts are unmistakable. All the recent anxiety about losing the AI race to China seems to have stirred America’s policy elites from their slumber in the magical land of free markets. They sound like a group no longer bound by the dogmas of the Washington Consensus. Some even sound like they have cast in their lot with the Beijing Consensus.

A recent essay in Foreign Affairs (2), the publication of choice of the US foreign policy establishment, reveals not only a new-found enthusiasm for a strong, AI-boosting state, but a re-examination of policy errors of the past. Co-written by Eric Schmidt, the essay chides Washington’s earlier fascination with globalisation for distracting America from ‘strategic considerations’ and also attacks the venture capital industry’s short-termism.

It celebrates instead ‘grants, government-backed loans and purchase commitments’ as the right vehicle for accomplishing Washington’s long-term tech objectives. Obviously, such grants would probably be disbursed through entities like AFF, which, unlike conventional VCs, would know how to spend it with long-term interest in mind.

At times, the essay is just a few steps away from demanding a robust industrial policy. But it never quite gets there, saying that it remains a ‘fraught label’. As things stand, the revised Washington Consensus expects more public handouts to the private sector – and uses the fear of the US losing the next cold war as the main rationale.

Such arguments are usually framed to appeal to both sides of the aisle, making it imperative to complement geopolitical considerations with economic ones. And so it is this time, with the push for AI framed as a way to make America great abroad and at home, in the latter case by turbocharging new AI-enabled industries.

Thus, while some have mistaken the new American consensus for the emergence of ‘post-neoliberalism’, it has all the trappings of the cold war-era ‘military Keynesianism’ whereby greater military spending was supposed to help beat the Soviet Union and ensure American economic prosperity.

The Pentagon stands back

Yet, erasing the last three decades of neoliberal statecraft is proving difficult. Apparently, one can’t simply go back to the cold-war days, with nearly unrestrained funding flowing to a handful of military contractors. Today, one has to play lean and entrepreneurial, and generals don’t dream of relaunching themselves as Silicon Valley startups. The Pentagon has even been reluctant to operate its own In-Q-Tel-like venture fund, walking away from the money the US Congress has allocated it (Fortune.com, 17 March 2023). This might explain why AFF had to be created as a private venture.

Still, the links between the Pentagon and Silicon Valley have undeniably grown stronger. The US Defense Department has even recently created the new post of Chief Digital and AI officer – a position filled by Craig Martell, former head of machine learning at the ride-sharing platform Lyft. And, for all the ethical concerns from their employees, America’s tech companies continue making inroads into the military’s procurement budgets.

Thus, while Alphabet may have shelved its plans to work on the Pentagon’s controversial Project Maven – a computer vision project that prompted protests among its engineers – it went ahead and formed Google Public Service, an innocent-sounding unit which does offer cloud services to the military.

It’s hardly alone. Silicon Valley’s cloud and machine learning capabilities remain crucial to the Pentagon’s vision of building a system to integrate data coming from terrestrial and air sensors – these are no longer limited to radar – from across all branches of the armed forces. The idea is to crunch them with artificial intelligence, so as to mount an effective joint response. To that end, in late 2022 the Pentagon awarded four tech giants – Microsoft, Google, Oracle and Amazon – a $9bn contract to develop the infrastructure for this bold effort (defensescoop.com, 7 December 2022).

But, unlike in the old cold-war days, it’s not at all clear how much of this money would trickle down, Keynesian-style, to ordinary Americans. When it comes to AI, the labour costs accrue either to the star engineers – and we are talking hundreds, not millions of them – or to the numerous low-paid contractors who toil to help train the AI models. Most of them are not even in America, with OpenAI relying on contractors in Kenya to prevent ChatGPT, the popular AI chatbot, from showing obscene images and text.

As for cloud computing, it’s also not clear how its expansion would help. Building data centres is expensive and their positive effects on the economy have yet to be seen. They do tend to drive up the cost of real estate. And the environmental costs of both AI and cloud computing are not trivial. So the multiplier effect of pouring all this money into military AI might be a mirage.

So perhaps this Cold War 2.0 would not feature a return to ‘military Keynesianism’. Short of AI leading to the much-awaited Singularity, merely dumping more money in the tech industry won’t suddenly bring Keynes to life. Perhaps we are more likely to witness the weird new regime of ‘military neoliberalism’ in which ever-greater government spending on AI- and cloud-related matters would widen inequality and enrich the tech giants’ shareholders.

Return of Google’s ex-chief

No wonder some of them are so keen to restart the cold war. In fact, it’s Google’s former boss Eric Schmidt who’s done more than anyone to shape this new consensus (3). Worth roughly $20bn, Schmidt has been a fixture in Washington’s policy circles ever since campaigning for Barack Obama in the 2008 election. Between 2016 and 2020 he chaired the Pentagon’s Defense Innovation Advisory Board – for whom he visited some hundred US military bases around the world. He then made a quick transition to chair of the AI commission and has now resurfaced on a government commission on emerging biotechnology.

Schmidt’s many ventures to push his cold war narrative make it hard to keep track. There’s Innovation Endeavors, Schmidt’s venture capital fund, which lavishly funds startups that specialise in military AI. The highest-profile of them, Rebellion, has been recently described as an ‘early-stage WeWork and Theranos’ – a reference to the overblown promises made by its founders (Vox.com, 14 December 2022).

While Schmidt chaired the two government entities – the Pentagon’s board and the AI commission – he and his business partners invested more than $2bn in AI-related startups. Given that the commission recommended pouring more government money into such startups, one can only wonder what Schmidt’s advocacy is all about.

Elizabeth Warren, the firebrand Democratic senator, has even pressed the Pentagon for more details on Schmidt’s involvement with the US administration, suggesting that the Defense Department may have ‘failed to protect public interest’ in granting Schmidt so much influence. Now that he’s joined the new biotech commission, raising some eyebrows, his investments in this sector – through yet another venture capital fund – have attracted more criticism (CNBC.com, 13 December 2022).

The case of Schmidt Futures

Then there’s Schmidt Futures, his philanthropic foundation, which on closer examination turns out to be a for-profit company (4). It was recently in the news for funding the salaries of staffers working in the US government (5) – including in positions related to AI policy and tech regulation. Eric Schmidt (and indirectly Schmidt Futures) even played a role in helping Lyft’s Craig Martell to become the Pentagon’s AI chief.

How could a company possibly be paying the salaries of administration officials? Well, there’s a loophole: certain nonprofits are allowed to do that, and, as nonprofits, they can even receive money from companies. In this case, the nonprofit that receives money from Schmidt Futures (but not only them) is the Federation of American Scientists, a venerated cold-war thinktank that traces its origins to the Manhattan Project. Conveniently, its current chairman is that same Gilman Louie of Tetris fame (6).

Schmidt’s most effective cold war publicity gamble has been to recruit to his cause Henry Kissinger, not known for shunning billionaires. Perhaps, as a result of Schmidt’s influence, the 99-year-old Kissinger opining on AI sounds much like a 19-year-old describing their first LSD trip. ‘I think the technology companies have led the way into a new period of human consciousness,’ Kissinger said in a recent interview, adding that it is akin to what ‘the Enlightenment generations did when they moved from religion to reason’. By this logic, Eric Schmidt must be the new Voltaire.

Schmidt, Kissinger and a third collaborator have even penned a book-length manifesto (7) about this new era, in which they warn that AI warfare may lead to ‘inherently destabilising’ situations that are like those ‘created by nuclear weapons’. They ask, ‘Will terrorists engineer AI attacks? Will they be able to (falsely) attribute them to states or other actors?’ But they offer no answers. At times, it seems just a rehashing of the old talking points around the inevitable ‘cyber 9/11’, the earlier rallying cry of defence contractors eager to pillage public budgets.

After all this fearmongering, they arrive at a logical conclusion. What the world needs is a ‘concept of arms control for AI’. That’s all they say. The book remains vague on detail, opting for open-ended philosophical questions over analysis.

Schmidt is so keen to leverage whatever is left of Kissinger’s reputation that in 2021 he even launched a dedicated AI thinktank – Special Competitive Studies Project (SCSP) – modelled on a similar initiative that Kissinger himself ran at the height of the cold war in the late 1950s. At the time, Kissinger was far from advocating arms control of any kind. Rather, he was one of the most vocal proponents of the view that a limited nuclear conflict with the Soviets was all but inevitable, and probably a good thing for America.

Kissinger’s first cold-war initiative was itself a power-grabbing vanity project of another billionaire, Nelson Rockefeller. Its most high-profile report, released in 1958, called for continued increases in defence spending of roughly 1% a year and the overall ratcheting-up of America’s military preparedness. Its stance on arms control was unequivocal: ‘The illusion of security brought about by a spurious agreement to disarm would be a poor substitute for vigilance based on strength’ (8).

SCSP’s new ‘offset strategy’

For all the ‘AI arms control’ speculations of the Schmidt-Kissinger book, the policies advocated by SCSP point in the opposite direction. Of those, one stands out in particular: SCSP’s efforts to shape the next ‘offset strategy’ – a push SCSP markets under the catchy name of Offset-X. During the cold war, ‘offset strategy’ referred to the Pentagon’s efforts to leverage the latest technology, from battlefield nuclear weapons to airborne sensors, in order make up for its numerical inferiority with regard to Soviet tanks, planes and troops. From the mid-1940s, there have been three such ‘offset strategies’, all of them relying on different technologies and assumptions.

SCSP’s Offset-X advocacy assumes that, should a war between China and the US break out, the People’s Liberation Army (PLA) will go after US networks – and the US must be prepared. Thus, according to a recent SCSP report (9), ‘the outcome of a potential war with the PLA will increasingly come down to superiority and resilience of sensors, networks, software, interfaces between humans and machines, logistics and – especially – the systems that connect or empower them all together.’ This doesn’t exactly sound like arms control, does it?

To the uninitiated, it may sound awe-inspiring, but anyone familiar with Pentagon thinking over the past decade would probably yawn. Much of this was already part of Offset Three, which ran between 2014 and 2018, spearheaded by, among others, then Deputy Secretary of Defense Robert Work, who’s now resurfaced on SCSP’s board of advisors.

The primary audience for SCSP reports is not military types, but the general public. It’s they who need to be persuaded to increase the Pentagon’s AI budgets. And to do that, one must make a convincing argument that China is not only winning the AI war, but that its victory would crush the US militarily. The second assumption remains the stuff of science fiction. But is China winning that AI war?

US culture wars

To judge by its inability to offer an alternative to America’s ChatGPT, Beijing is still a long way off winning the AI race. Baidu’s launch of its ChatGPT competitor, ERNIE Bot, went so badly its stock price tanked.

To some extent, Silicon Valley’s leadership in the so-called Large Language Models – the type of deep learning AI behind ChatGPT – is a consequence of America’s cultural hegemony. The reason OpenAI, the company behind ChatGPT, is so far ahead of the competition is, in part, because its model is trained on vast holdings of English texts, abundant online. You won’t find as much Mandarin content.

Those who have been worrying about cultural imperialism have a great cause for concern now: ChatGPT may well emerge as the default provider of answers to the world’s questions. However, with ChatGPT, one is likely to be treated only to the blandest and most politically correct answers. We might all become prisoners of America’s culture wars.

Viewed outside of the narrow confines of ChatGPT-like language models, one may suspect China’s technology push is still going strong. A recent report by a prominent Australian thinktank suggests that it leads in 37 out of 44 critical technologies (10), a list that spans defence, space, robotics, energy, the environment, biotechnology, AI, advanced materials and key quantum technology areas.

The problem with most of such studies is that they focus – almost excessively – on research indicators tied to the relative performance of academic institutions, the number of publications and the degrees of individual researchers. Though this can be a useful proxy for dominance in a particular field, all these research papers count for nothing without the capacity to implement their findings.

And it’s here that US attempts to break China’s rise are bearing fruit – from efforts to cripple Huawei’s dominance in 5G to policies aimed at preventing Beijing from achieving self-sufficiency in the manufacture of advanced chips.

The tech industry and military contractors don’t always see eye to eye. Many American tech companies don’t want to lose the Chinese civilian market due to its sheer size. They definitely don’t want a full-blown Cold War 2.0, which would adversely affect their own sales there.

The military contractors, who don’t have many civilian contracts to begin with and who cannot work with the Chinese military without losing their Pentagon ties, have no such constraints. They want Cold War 2.0 – and they want it now. And some probably wouldn’t mind a hot war either.

Uneasy compromise

The Biden administration’s policy towards China’s tech ascendancy – slow but profitable strangulation – reflects the uneasy compromise between the two sides. Washington tightens the noose around Beijing’s neck by cajoling allies like the Netherlands, South Korea and Japan to stop selling their own critical technologies to China. As usual, it is also using cold war-era legal instruments – like the so-called Foreign Direct Product Rule – which can limit what foreign companies can ship to China, as long as their products are made using American technology.

This is not to say that Chinese companies have been completely cut off from the US hardware they need to continue their work on AI self-sufficiency. Instead of buying chips, they are now directed to rent them, sometimes at exorbitant prices, which has allowed some US tech giants to profit from Washington’s crackdown on Beijing.

The idea is to make AI development costly – but not prohibitively costly – thus profiting from Beijing’s desire for autonomy. And since Biden’s policies also slow down the pace at which Beijing achieves its goals, they give the US time to sort out its own AI-related problems (which by and large are to do with having too many eggs in Taiwan’s chip basket).

Keeping ‘China reliant on foreign chips’

At least no one in Washington is hiding the fact that maintaining and monetising China’s dependence – the one-time complaint of dependency theorists like Andre Gunder Frank or Ruy Mauro Marini – is the explicit objective.

Thus, another recent article in Foreign Affairs (11) – predictably, by someone in the Eric Schmidt universe – holds that ‘instead of implementing broad bands, US policymakers should work closely with allies to maintain China’s dependence on foreign chips’ and that ‘ensuring that Washington maintains the upper hand as the AI revolution progresses’ requires keeping ‘China reliant on foreign chips’.

Beijing is not taking any of this lying down. Recently, it insisted that Japan should think twice about joining Washington’s campaign to limit China’s access to advanced chips, all while starting a cybersecurity probe into Micron, an important US chipmaker.

What isn’t yet clear is whether Beijing would be able to steer an international coalition of some kind to support its agenda. Washington has not been acting alone to counter China, tapping into or even spearheading international initiatives such as Global Partnership for Artificial Intelligence and AI Partnership for Defense. Recently AFF – the Schmidt fund run by Gilman Louie – announced a joint fund with India, Japan and Australia under the auspices of the Quadrilateral Security Dialogue, a joint military initiative of the four countries aimed at restraining China.

Most of these efforts are taking place under the banner of promoting democracy and world peace, even if getting there requires boosting defence budgets and letting technology companies and their shareholders get even richer.

Europe is mostly absent from these efforts, for the obvious reason that on military matters it follows the US lead. And when changes occur, they are usually smallscale. It was recently announced that the Netherlands will host the investment management arm of the new €1bn innovation fund announced by NATO. That’s peanuts.

While the war in Ukraine has boosted European defence budgets, it will probably be American companies like Peter Thiel’s Palantir that will get most of the new AI-related funds. At this point, it’s Europe’s privacy laws – and not active public policy – that prevents US giants from advancing even further and faster. And it’s not just the recent case of Italy banning ChatGPT. A recent court ruling in Germany has found that the police use of Palantir-supplied data analysis software to prevent crime before it happens is unconstitutional. How long these privacy defences might last is anyone’s guess.

Judging by recent high-profile speeches by European Commission officials, they find Washington’s rhetoric of Cold War 2.0 quite convincing. This is likely to have an adverse impact on EU-China relations, while pushing the EU further into the arms of US tech giants. A much smarter strategy for the EU would have been to play the two sides against each other, something that Brussels has tried to do in the past, at least on some issues.

Absence of a cold-war enemy

In her 2014 book on how America’s national security state (rather than Silicon Valley) has been the true engine of the country’s technological leadership (12), the political scientist Linda Weiss noted that the absence of a cold-war enemy has undermined the Pentagon’s ability to create ground-breaking innovations. She even asked ‘why China has not yet metamorphosed into a rival that spurs innovation like the Soviet Union and Japan’.

It turns out this was just a matter of time. Then, Weiss argued that if the US was serious about preserving its technological leadership, it would need to get over its obsession with what she called ‘financialism’, put Wall Street’s interests aside and focus on rebuilding its manufacturing industry.

Financialism, of course, never ended. What we got instead is something much weirder. While some re-shoring does, in fact, take place, it’s anyone’s guess whether America will really reinvent itself as the world’s primary chip manufacturer.

Surprisingly, it’s not the downfall of Wall Street but the rise of Silicon Valley, with its desire to capitalise on the AI hype, that may have awakened the US from its slumber, while turning China into the strategic enemy that the Soviet Union once was.