Reasons why machine learning technology could lead to an unstable world. Rapid and continuous progress in machine learning will lead to a new geopolitics: artificial intelligence nationalism (AI nationalism). Machine learning is a ubiquitous strategy that will affect all industries and parts of society. The changes associated with machine learning in the economy and the military will lead to national and international destabilization that will force authorities to take action. Ambitious governments have already begun to view machine learning as a key "breakaway" technology of the 21st century, and the race is on.
Artificial intelligence policy will become a critical area of general government policy. An arms race among the leading nations will accelerate. We will see an increase in protectionist measures to protect the nation's favorite companies, with blocked takeovers of these companies by foreign firms and talent attraction.
This arms race will potentially accelerate the pace of machine intelligence and shorten the time it takes for so-called "strong artificial intelligence. And while there will be many common aspects to this techno-nationalist plan, there will also be policies specific to certain states. Nationalism is a dangerous path, especially if the international order and norms change.
Progress in machine learning.
Over the past few years, we have seen remarkable advances in machine learning research and the commercialization of these technologies. Here are a few examples:
Image recognition is beginning to achieve the inherent accuracy of human perception in particularly complex tasks - such as skin cancer classification.
There is significant progress in the use of neural networks for machine translation at Baidu, Google, Microsoft and others. Microsoft's system translates news from Mandarin (a dialect of Chinese) into English at the level of an ordinary human translator.
In March 2016, DeepMind created AlphaGo, the first computer program to beat a world champion in the game of go. This is a significant achievement, as machine learning researchers have worked for decades to create a system capable of defeating a professional player. AlphaGo "learned" from 30 million moves made by human experts.
Eighteen months later, DeepMind released AlphaZero. Unlike AlphaGo, it did not use the moves made by humans during learning. Instead, it learned by playing against itself. Not only was AlphaZero able to beat AlphaGo's predecessor, but it also beat best-in-class chess and segi (Japanese chess) computers using a technique known as "transient learning." Leading machine learning researchers have consistently noted the "uncanny" importance of simpler algorithms that did not use human-generated data at all, but ended up being more competent and suitable for transfer intelligence.
There is a significant gap between AlphaZero's achievements and "strong artificial intelligence," although there is a sense that this is just another little bit of a step in its direction.
In addition to the research proper, there are incredible advances in the use of machine intelligence in large marketplaces, from search engines (Baidu), ad targeting (Facebook), and store automation (Amazon) to many emerging sectors - self-driving vehicles, new drug creation, and biosecurity and robotics. CB Insights has done a good overview of all the markets in which startup firms are using machine learning.
This rapid progress has led leading practitioners in artificial intelligence to think seriously about how it will affect society. Even at Google, a leading machine learning company, management seems to be moving away from techno-utopian positions and publicly acknowledging the risks accompanying machine learning research's accelerated development and commercialization.
Three manifestations of instability
So why does it matter to nation-states? There are three main areas where the accelerated progress of machine learning could destabilize the international order:
- Commercial applications of machine learning will create powerful new businesses and destroy millions of jobs. The country that invests most effectively in these industries will become the strongest economically.
- Machine learning will make possible new ways of waging war, both powerful cyberattacks or defense and various forms of autonomous or semi-autonomous weapons, such as the long-range anti-ship missile that Lockheed Martin is developing. At the very least, the country that invests first and most aggressively in this industry can gain an absolute military advantage.
- Finally, the artificial intelligence of more general specialization would provide a fundamental acceleration of scientific and technological research. In my view, this could be the most significant source of instability. Imagine, for example, a state whose leadership in artificial intelligence enabled it to be the first to create a fusion reactor to generate electricity. Again, this could provide the country with radical technological superiority at its most extreme.
Machine learning is a unique all-encompassing technology capable of influencing almost any area of national policy. Human intelligence has given shape to everything we now see around us, so our ability to create machines with greater and greater brilliance can have the same impact.
We can find certain historical parallels to help us make sense of possible developments. Nuclear technology has dual military and peaceful applications (atomic weapons, radiotherapy, power generation), as does oil (whose use has spread from lighting to heating and then to a wide range of military and industrial applications). Both of these technologies have a huge impact on geopolitics. Because of this, states quickly became players in the first roles and remain so today (e.g., the 6,800 nuclear warheads of the United States or the 695 million barrels of oil in the strategic oil reserve).
Ambitious governments have already begun to view machine learning as the most important technological "breakthrough" in the 21st century, and the race is on. This race will in some ways resemble the nuclear arms race of the last century and the geopolitical tensions and alliances between states and multinational companies that arise around oil. Economic, military and technological superiority has always been a powerful incentive for countries.