“Any sufficiently advanced technology is indistinguishable from magic.” - Arthur C. Clarke
Jenna Abrams is not a real person. Her Twitter feed had 70,000 followers. She inspired support from fans and angry responses from opponents. Her mostly inane, but occasionally racist blog posts were widely read and reposted, with enthusiastic comments. At one point, she managed to get a fake story picked up by mainstream news media. Her Twitter handle was on a list of accounts connected with Russian intelligence, turned over by the company to Congressional investigators.
We know what Jenna Abrams was not. She was not a living, breathing, carbon-based entity. What she was remains unclear. Was she a bot? An AI project? Was her account supported by an individual or a team? Should Abrams be described as a “she” or a “they” or an “it”? Abrams maintained a Gmail account and a blog. She replied to tweets and emails. After her Twitter account was suspended, she posted an enigmatic anti-confession on her blog, a mind-pretzel acknowledging the claims about her identity without either a formal denial or an admission.
Stories about Abrams and other accounts disclosed by Twitter and Facebook have focused on Russian meddling in our democracy. Behind that tale of espionage lies a more disturbing theme, one that will continue to trouble us after Russian political intrigue fades. We have long assumed that reality is universal and accessible to everyone. Future generations will not take reality for granted. Technology we created has defeated an evolutionary advantage we enjoyed for hundreds of thousands of years. Our brains are no longer the world’s finest computation engine.
Liberal democracy is built on assumptions developed and refined by Enlightenment Era philosophers. They concluded that reality is objective and measurable – singular and discoverable rather than varied and relational. Reality exists independent of our relationship to it or our acceptance of it. The Age of Reason was a rejection of superstition, prejudice and unquestioned tradition. By Enlightenment logic, any individual can judge reality on their own, without paternalistic intermediaries like priests or aristocrats, merely by educating themselves and learning to doubt, question and test. On the foundations of that empiricist ideal has risen a global political and economic system, dedicated to the proposition that individuals are the best judge not only of their own interests, but of the optimal direction of a civilization.
In a world governed by evolutionary dynamics, no successful innovation remains successful indefinitely. Progress bred of Enlightenment insights changed the world in ways that now challenge the relevance of those same insights. The Enlightenment is over. We don’t yet know what comes next.
This notion of an empirical, discoverable reality may still be true, but our ability to define it beyond a narrow, personal sphere is slipping away. Our first hints of trouble emerged a century ago, as discoveries in particle physics challenged our confidence in empiricism as the only basis of reality. However, what kicked the stool out from under the Enlightenment ideal was the explosion of artificial computational power over the past half-century.
Computer chips in your kitchen appliances carry more calculating power than NASA could generate in its entire array of Apollo-era mainframes. It is difficult to place a relatable context around the exponential growth of cheap, artificial computing power in our lifetimes.
We’ve grown used to a world in which our tools outperform us at most manual tasks, but cannot match us at uniquely “human” activities. That cognitive advantage is now slipping away. There are no games remaining in which a human can reliably defeat a computer. Machine learning algorithms now comb through wire reports and press releases to compose news stories for major publications. Hundreds of news reports have been produced by Heliograf, the Washington Post’s automated story-writer. Computers can compose communications that we would not recognize as artificial.
When all else fails, we tend to trust what we observe through our senses, but our computing platforms can manufacture highly realistic sensory experiences. Programmers at the University of Washington recently demonstrated an AI engine that can simulate convincing video content, dubbing a recorded statement with artificial messages. A Game of Thrones fan trained an AI engine to write the next book in the series. The output is posted here, a clumsy, though fascinating effort. More important than the quality of the robot book is the relatively minimal effort involved in producing it. It was made by a student in a Udacity (free online learning) course, using free, readily available software – the digital equivalent of kitchen chemistry.
When artificial cognitive capacity is paired with traditional robotic capabilities, the results are astonishing. Within a few years, computers will likely render human drivers too clumsy and dangerous to be trusted on the road. Truck drivers may continue to exist, but in a few years they will likely operate dozens of trucks at a time from remote control rooms. How long that oversight function will persist is hard to guess. Digital intelligence is rapidly nudging human intelligence out of its evolutionary path.
With growing computational power has come an explosion of raw data. For 244 years, the Encyclopedia Britannica maintained a printed record of human knowledge. In 2012, they finally gave up. Sometime in the 1990’s you might have read that half of all information that ever existed had been generated since 1976, or some similar factoid. With each year that passed, that data creation mid-point kept shrinking. A few years ago, you could read that 90% of all data had been created since 2010. It’s a statistic that has lost any relevance, as data growth begins to look like a singularity. By now, half of all data that ever existed on Earth has probably been generated since breakfast. No human being can even pretend to keep pace with this expansion of data. Information is now generated predominately by machines, for consumption and use by machines. Our ability to derive any value from this information depends on our capacity to interact with those machines.
Jenna Abrams is not a real person, but with thousands of followers she influenced the behavior of a great many real people, both followers and opponents. In the 2016 election, AI was a weapon leveraged by Russian intelligence against us, but our computational disadvantage persists with or without an enemy to exploit it. Cheap bots or machine learning code can be turned loose “in the wild” to create content largely without intent, creating a kind of information pollution. Microsoft experienced this nightmare last year when it debuted its adorable AI-driven social media bot, Tay. Crafted to adopt a cutesy twee pop persona and described as a bot “with no chill,” within hours of launch Tay had tweeted threats against feminists along with a series of racist posts. At one point, the engine responded to a message from a follower with this gem: “bush did 9/11 and Hitler would have done a better job than the monkey we have now. donald trump is the only hope we’ve got.”
Microsoft scrambled to delete some of Tay’s most threatening and offensive Tweets, but its human handlers couldn’t keep up. Only by shutting off the machine could they rein in their monster. Human beings craft this technology, but the intentions of its creators rapidly disappear over the horizon behind it.
Similar code is available for free. With minimal computing skills it can be harnessed into communications engines and turned loose without supervision. This kind of cheap data pollution appears to have been part of the exhaust of the Russian campaign against us in 2016, but it needn’t be accompanied by any human intent to cloud our environment. Like an abandoned waste dump, these engines of misinformation can spread confusion unattended with few limits. Bots don’t get hungry. Bots don’t sleep. Bots don’t die.
In this new data environment, shrouded in a fog of highly convincing misinformation, what happens to old-fashioned, handmade, artisanal democracy? In the age of Trump, the answer is all around us. We’ve grown used to our dependence on experts for our very survival. No one tries to perform their own cancer surgery. Few people even work on their own car anymore. But we still hold as a central tenet of our political system the notion that each of us, independently, is supposed to be able to judge reality on our own, and use it to form policy preferences. That notion is now laughably false.
An unprecedented divide opened in the 2016 election cycle between college educated and non-college educated white voters. Polite explanations usually center around a “disconnect” between more and less affluent white voters, or concerns about racial bigotry. Ask people to explain their vote for Donald Trump and an additional explanation begins to emerge. Listen to enough of these voters and you’ll hear their position supported by entirely false narratives.
Lies, spin and tribal alignments are not new to politics, but we are facing something unique – the emergence of closed, synthetic worlds. Fed by the dislocating impact of a digital existence and the complexity that accompanies it, a universe of manufactured narratives has overwhelmed our sense of a common identity and a common fate. Our system presumes that reality is accessible to everyone, a kind of free-good on which all political engagement can be premised. If on the other hand, reality is elusive, it becomes the domain of experts. Those with high levels of education and a habit of relying on expert knowledge retain a reliable handle on objective reality. That hold is more tenuous than those elites wish to acknowledge, but it remains stronger than the image of reality derived from relatively closed systems. Crazy conspiracy theories and paranoid narratives spread more slowly among the affluent and college educated than among aging or working class voters in small towns.
In 2016, the structure and content of some of those closed worlds was influenced by intentional interference from Russian intelligence. However, their efforts merely weaponized an existing weakness in our political system. A disturbing volume of the misinformation in our political system is organic. Bizarre theories, like the Pizzagate conspiracy, take hold in a data echo chamber. They are amplified by social media bots with the help of people intentionally spreading deceit.
A reader might be tempted to a dangerous conceit. One might see this crisis of reality as a challenge to the rubes, a matter to be resolved perhaps with more investment in public education, or even by moves to narrow our reliance of democracy. This would be a dangerous misunderstanding, one we seem increasingly likely to make. While it’s true that less-educated voters were marginally more susceptible to the disinformation that wrecked the 2016 election process, they are not alone. They are merely a leading indicator of an advancing problem, the first domino to fall. We are drifting into an environment in which reality becomes so difficult to discern that people lack the energy or will to pursue it.
Jenna Abrams is not a real person. Laugh if you will at those duped by the performance, but a technological environment is taking shape in which human intellects will be hopelessly outclassed by the power of our tools. No one is smart enough to beat a machine at chess. To imagine that we will be able to tell the difference between a human and an AI on Twitter is hubris.
There is no obvious response, no simple policy prescription to carry us beyond the challenges of a digital, artificial reality. Our entire philosophy of representative government is challenged by this reality and we have no ready, viable alternative. That may sound unsettling, but merely coming to terms with the scope and character of this new environment is a challenge that will take time. Unfortunately, time is one of the resources rendered scarce in this new realm. While our bodies adapt on a biological time scale, our technologies adapt with the speed of automation. A more humble assessment of our capacity for defining reality would be a helpful start. From that new launch point may come adaptations to help us thrive in an automated future.
