Technology and Tyranny
By Jonathan Jeckell
Yuval Noah Harari, author of Sapiens, asserted in the October 2018 issue of The Atlantic that “technology” naturally favors tyranny. Technology in this case seems to mean artificial intelligence specifically. Part of his argument rests on the ability of governments and the powerful to control the masses, and the rest on technology displacing people from the workforce, with both connected to the idea of transferring authority to machines.
The relationship between liberal democracy and technology is far more complex than portrayed in that article. For starters, technology created the middle class and led to many of the freedoms and participative economy we enjoy today, stemming from the Renaissance and Enlightenment through the Industrial Revolution and England’s Glorious Revolution. Artisans and skilled workers, the Burgers and city dwellers of the Middle Ages broke the stranglehold of the nobility over people in Europe and spread to other societies across the globe.
But technology itself isn’t an unalloyed good or bad thing for shared prosperity and liberty. Harari’s view is just as wrong as the widespread belief at the beginning of this decade that technology was playing a role in freedom’s inevitable march through the Arab Spring and Color Revolutions. Twitter-powered activists were seen as running circles around state security officials and organizing peaceful mass protests that toppled authoritarian regimes. But it didn’t take long for regime security agencies to adapt to this challenge.
Generally speaking, innovation and the development of new technologies, particularly the truly disruptive ones (in the Clay Christensen sense) tend to reshuffle the deck and lead to new, well-paying jobs. It’s only when technology becomes systematized to the point that the economy of scale from large investments can win that technology can begin to serve the powerful more than broad swaths of society. The agility of small businesses and the churn of creative destruction can stave off the development of stable monopolies or the unchecked accumulation of power by political leaders.
Certainly artificial intelligence—which is what Harari really means here—can concentrate a lot of power and wealth to a few people. He is also right when he notes that AI, unlike past technological shifts, could lead to a cascade of disruptions without settling into a new equilibrium. So new jobs typically created by creative destruction that outpace the loss of obsolete jobs may happen faster than we can educate people to take them. AI also allows unprecedented control of the population, as seen with China’s massive investment in face recognition surveillance systems.
While the shift of mental work to machines may be more profound than the shift from muscle to machine, we’ve experienced this many times before with the same concerns raised each time. While that sounds like an article of faith, there are other factors that contribute to the success or failure of adapting to these changes. Firstly, there is no guarantee that AI will continue to improve at the rate we are seeing today. Many researchers voiced concern that Deep Learning is running out of steam, and that adding more computing power to it will yield diminishing returns on solving new types of problems. Many researchers have delved into other areas of AI research, but there are concerns about another AI winter, where funding dries up after excessive hype results in cynicism and disappointment. Even if AI research continues unabated, there is nothing that assures us that we are closer to making general purpose AI capable of understanding and reasoning in a human-like way.
I suspect researcher will inevitably solve artificial general intelligence, but technology alone does not determine the relationship it has with liberty and free markets. Life was comparatively good in England because a balance of power between the monarchy and the nobility in England resulted in both powerful groups catering to the peasants for support. Likewise, the rise of concentrated wealth and power would either result in rival groups or the bandwagoning of smaller groups against the most powerful group. These rivalries would either benefit the rest of us by attracting support, or at least place the focus of their worst tendencies against each other.
It’s also likely there will be many jobs that are possible to automate, but aren’t attractive to automate even when it would be cheaper than paying a human worker. Robot welders have been used since the early 1970s, but skilled welders are still in short supply and are paid relatively well in many parts of the country. Even with advances in nuanced skill and mobility in robots, and the adaptability and reasoning in AI, humans are still favored for many types of work, particularly in interfacing with other humans. Even as AI begins to infiltrate into creative endeavors once thought to be the safe refuge for human skill over machines, AI and robotics will likely be used to augment human skill rather than displace it entirely.
Changes wrought by technological innovations are slippery to predict because the process is interactive, complex, and competitive. Nuclear weapons, poison gas, and machine guns were supposed to make war obsolete, but in every instance, innovations on the battlefield were countered by innovations by the other side. The internet was once thought to be bringing the entire world to better understanding and harmony, while squashing other cultures and languages, yet instead a diversity of culture and a tempest of competing ideas emerged instead, while dying languages found new life by connecting people to speak them around the world. There are often many ways to use a technology, and competitive ways to use them are not always obvious.
While AI may self-perpetuate the accumulation of power and wealth by finding new, creative uses for itself, that requires an understanding of needs, or at least large amounts of data about needs. Meanwhile the vast majority of people who are not wealthy or powerful will experience needs that will drive innovative, new uses for AI. The people closest to the needs have the best understanding and the most tacit knowledge about the problems they face.
Unlike past technological revolutions, wealth and power may provide diminishing returns in investing in AI. Deep Learning doesn’t require enormous investments and is very accessible. The tools to use Deep Learning are currently available to everyone. Google has free and approachable online courses for learning TensorFlow, and so are many others. Lobe is trying to make AI accessible to people who don’t even know how to code at all. The only expensive thing about Deep Learning is obtaining the data, which does not necessarily favor those with wealth and power if people can be mobilized to provide data relevant to problems they want solved. As Matthey Ridley pointed out in The Rational Optimist, tyranny tends to bring innovation to a halt for this very reason, while mass participation leads to an ever better understanding of needs driving new innovation.
So while its possible that artificial intelligence could cause unprecedented concentration of power and wealth, we’ve heard all of that before. What’s more likely is that social forces and institutions that began long ago will also evolve, as will the interactions with the technology itself. Technology rarely provides unilateral advantages to one side.