Part Three|Geopolitical Lens: AI Policy and Grand-Strategy in 2030
State of Play
AI and geopolitics have an interesting relationship: First, AI, like space exploration during the cold war, is used as a projection space for a power competition by the two main players, China and the US. Second, like dreadnoughts and nuclear weapons before, AI pledges to increase geopolitical power both in terms of influence and military power.
In his book “AI Superpowers”, Kai-Fu Lee tells the story of how official China was shocked when a US AI beat the best Asian players in Go. China also realised that AI gave it the chance to become leader in an important technology by 2030 and at the same time stabilise regime control. So President Xi made the Internet, Big Data and AI core parts of his modernisation effort, betting on its effect on the economy and security. With its closed internet, control over companies, and the buildup of AI- managed surveillance, China is developing a counter model to the vulnerable, decentralized but powerful original (western) digital revolution.
China has some advantages: With no real data privacy and the authoritarian urge to collect all possible data, China possesses a huge data reservoir for AI and Big Data to learn from. Computing power is centralised, and China is capable of organising huge state projects. Due to the close links between SOEs, private companies and the state, China can choose the best actor for each problem. It shows results: for example, according to its own head of research, Alibaba’s Siri-like AI customer phone service has overtaken Google duplex. It already manages conversations with interruptions, nonlinear meanings, and implicit intent, and is already used million times a day. On the other hand, President Xi repeats old mistakes like stuffing SOEs with old communist party cadre. The authoritarian climate cannot be favourable for researchers, who depend on criticism and a free exchange of ideas.
Daniel Fiott notes that “not only did China release its own national development plan for AI technologies in July 2017, it is developing a military doctrine designed to overmatch the US’s military technological prowess.” The West meanwhile still allows Chinese companies to buy themselves into digital firms and the supply chain, thereby probably weakening defences of critical infrastructure and giving Chinese intelligence insights and sources for data collection. Only recently did the US commerce department consider export controls for AI and other advanced technology.
One reason for this was that the United States had its own “Sputnik moment” when it realised it would be challenged on a field it has called its own until recently. The US department of defense under President Obama — the traditional state investor in high-tech — did not finalise its third offset strategy, a massive R&D investment scheme, and President Trumps administration has not picked up on it until recently. It seems the current leadership is skeptical of novel technological solutions to warfighting. Nevertheless, the new administration has not stopped running research and the buildup of the Defense Innovation Unit Experimental (DIUx) which is meant to finance silicon valley start-ups of interest to the military. Secretary of Defense Mattis also made it clear that his strategy was focused towards China, which gave the AI strategies more of a direction. On the whole-of- government level, the US is streamlining its national security effort in AI by creating a “National Security Commission on Artificial Intelligence” uniting congress and agency heads.
US AI policy looks well on its way. However, it seems many in the US still trust Silicon Valley with its talent and financing to lead the way. Maybe that will be enough, but one needs to bear in mand that former technological advances like personal or mobile computing or the internet were based on state-funded inventions. There also is a political component that should not be underestimated: The US was the economic winner of automation and social media, but also lost many jobs because of it (see the previous chapter). Therefore, President Trump and many politicians in both parties might not see helping the Californian IT industry as their priority.
The European Union focuses on two goals of its own: catching up on transforming its good research into viable products and services, and trying to figure out where the strengths of its AI approach are (experts mentioned business-to-business AI, robotics, industrial AI). Many member states and the European Institutions (e.g. the European Parliament’s 2017 Delvaux Report on Robotics) came up with AI strategies. Last year, France and Germany set up the decentralised Joint European Disruptive Initiative (JEDI). Two large groups of European scientists founded CERN-like scientific projects calledCLAIRE and ELLIS to structure the catching up effort of European business and research entities in AI. In terms of financing military research, the EU also makes up leeway. Recently it madecommitments to use 5% of the European Defence Fund for disruptive technologies, and some of the 34 PESCO projects have an AI component.
On the warfighting level the “ethics gap” — the weakening of western military power due to ethical considerations — might look bigger than it really is, argues RAND. Still, as stated in the ESPAS ideaspaper on Warfare, “western democracies are more at ease with the AI-augmentation of humans, the teaming of unmanned systems with manned ones (e.g. formations of drones and planes, where the pilot decides for all), and AI in an advisory instead of a command role. The US and most European militaries has already decided on having an AI-policy of a ‘human in the loop’ in all operations that involve harming someone, but as always, the devil is in the detail.”
Main Trends
According to the Belfer Center, the “impact of technologies such as autonomous robots on war and international relations could rival that of nuclear weapons”. Deciding whether to field and use such weapons will be a hard choice. Improvements in AI might “shake up the balance of international power by making it easier for smaller nations and organisations to threaten big powers like the US.” It is fairly cheap and quick to couple AIs with physical or cyber weapons. On the other hand, only a few states seem to be on the forefront of AI, and most are big (except maybe Israel).
Again quoting the ESPAS Warfare paper, “the West has to be aware that preferring AI in a supportive role to purely AI-controlled weapons might be a price it pays in combat losses at some point. It would be a price that the West might be willing to pay for keeping to its ideals. This is also not new: Soviet weaponry cared less about the survivability of their crews, so long as they were efficient, while Western countries prioritised safety. This sometimes put the West in weaker positions, but it was deemed worth it.”
Offensive cyber capabilities will increase rapidly due to AI. Earlier this year, technology giant IBMunveiled a proof-of-concept called DeepLocker, the first-ever malware based on AI. It is more or less invisible, can be employed with unprecedented accuracy and provides high anonymity to attackers.
While there will be an overall shift away from hierarchies towards networks, AI and Big Data will make both more powerful. Centralised military command and control will gain in oversight in rapidly changing environments due to AI. Meanwhile, AI will also facilitate decentral communication and learning.
In addition, AI supported analyst networks, using all available sensors to their advantage will makecamouflage and deception much more difficult. “Already today, it seems that intelligence services struggle to build fake identities due to the missing data trail, military units give away their positions when soldiers post photos online, and collaborative or AI-guided commercial satellite image analysismakes it difficult to hide any larger piece of equipment in the field.”
Key Uncertainties
China is becoming more powerful in all aspects. In terms of military power, China has the capabilities and the will to catch up. However, to surpass “the West” China would need to undertake its full transformation into a interventionist power. In addition, no one is sure in how far China’s current strengths — centralised control over priorities, investment and the private sector, a long-term plan and a focus on big ticket technical solutions such as AI and space technology — are useful in innovating its military in the future. Military strength also depends on economic strength, and China’s economy seems less stable in the recent years. Finally, what if the fixation of China (just like Russia) to beat the US is making them invest inefficiently both in military and economic research?
Will China overtake the US in its AI capability? In the short term no, quoting Eric Schmidt, CEO of Google: “US will maintain its lead over the People’s Republic of China for the next five years”. Business leaders are split (yes, no argument) on this, so are academics. In the long run, after these five years, Schmidt expects China to “pass us [the US] extremely quickly.” But how would an economic downturn or political instability affect AI policy?
What is the next magic formula for AI innovation? Nobody knows how good companies really are in the battle over future technological change. It might be that Silicon Valley with its “let’s try it” approach, its connections to money and the best and brightest, has only been good for a certain period. Just like US government-led innovation has been working best from the Manhattan project to NASA’s moon landing.
It seems that Europe follows a broad approach, highlighting a human-centred focus, high ethical standards and broad investment. It is not clear what the rewards or penalties for such a broad strategy are.
Possible Disruptions
China has already isolated its internet from the rest of the world. Meanwhile its AI and social media services face resistance in other countries due to data issues and fear of Chinese mischief. With western products being replaced in China, and Chinese digital products not selling in advanced economies, will there be two incompatible digital worlds and markets in the future?
What if, in a turn around, China becomes a champion of data privacy? Which could the causes for such a change be? Can authoritarian states be trusted to guard data rights?
What if another country makes a dash for AI leadership? India lags behind, but AI promises to make the country more transparent and efficient, something many strive for. Japan has a good research base and sees a way to replace its ageing workforce in AI.
What is the chance that we see AIs fighting each other in the real world or the cyber dimension? The goal of war is to make an enemy submits to your will. This means AIs and robots will not be the ultimate targets, but humans and the things they cherish. However, AIs are vulnerable to hacking: governments, terrorists, activists and businesses might try to introduce intentional bias into AIs, leading to vulnerabilities.
Written by Leopold Schmertzing
See follow-up articles for detailed analysis
