"The time is coming when AIs will have better judgment than most politicians."
—Joshua Davis
Imagine that you were given two choices for the next president. The first choice is a person from a different political party with a set of policies or beliefs that you strongly disagree with. The second choice is an AI designed by a party with the sole purpose of achieving the policies you like. Which would you choose?
Computer scientists are now designing AI-based systems that are able to outperform humans in a wider and wider range of tasks.
Eventually, says Joshua Davis at WIRED, AIs will also prove themselves superior when it comes to human governance.
This idea may sound a little too sci-fi for many right now, but "the time is coming," he says. We spoke to him recently on our podcast (see Man vs. Machine – Who Will Govern Us Better?) about this radical idea, what it might look like to have an AI as President of the United States, and whether this is even possible from a constitutional standpoint.
Humans Are Not Good at Governing
First, it wasn’t the current political climate that spurred Davis to raise this idea, he noted. Rather, it came from examining the nature of our current political system in relation to how AI is evolving.
“We vote for people, more often than not, based on the way they look, their hairstyle, or the way they talk,” Davis said. “These are not really at the heart of the issues.”
To Davis, human personality—particularly our tendency to be drawn into scandals—has gotten in the way of governance. History, he says, shows that we’re not very good at governing ourselves.
At the heart of Davis’ argument is the fact that artificial intelligence has been outperforming humans at specific tasks for a long time. For example, in 1996 the program Deep Blue beat then-world-champion Garry Kasparov at chess. More recently, AI created by Google beat the world’s leading Go players.
Go is considered by many to be even more complex than chess, and many top Go players were astounded with how quickly the program was able to master the game and win 60 games in a row.
“It was doubly remarkable because of the way that it won,” Davis said. “It won by doing things that Go players had never seen before. It won by innovating. It won by creative, strategic thinking.”
These traits typically were thought to be possessed only by humans. But they’re exactly the traits we also want in a political leader.
The Road to POTUS AI
Over the course of the evolution of artificial intelligence, we’ve seen the model for developing an AI change. The first attempts involved feeding answers to questions into a system.
More recent approaches employ what is called deep learning or neural networks, where AI processes a data set and draws conclusions for a given problem, Davis noted.
Using this approach, we’re seeing applications that are increasingly impressive. For example, a group of researchers used a neural network to identify skin cancer by submitting thousands of images of skin cancers to the program.
“The AI was better at identifying skin cancers than a group of doctors who had spent their entire careers training to identify cancer,” Davis said.
This isn’t to suggest that we’ll see an AI candidate in the next election cycle, however.
“We’re obviously not there yet,” Davis said.
“A Purer Form of Democracy”
With AI sophistication growing rapidly, the reality is, it’s increasingly possible to feed any kind of information into a system to optimize whatever process or field in consideration.
When it comes to running a country, “the question is, would an AI do better than a human?” asked Davis. “Humans are far from flawless. … To me, there’s a lot about humanity that causes problems when it comes to governance.”
He would like to see politicians make decisions based on the greater good, especially one with the added benefit of being scandal-free.
“In some ways, it could be a purer form of democracy,” Davis said.
One possibility is each party develops their own AI which people elect to track progress on issues of their choosing. Another possibility is that political parties fade altogether and, instead, the presidency might become a policy platform, with the sum total of majority votes dictating what policies the AI president choses to implement.
The Question of Constitutionality
Some may object that this future scenario is implausible since it would not pass the US Constitution's elibility requirements for a US president.
Certainly, technology-embracing countries outside of the US could decide to do this or Constitutional ammendments could be made but, as it currently stands, the three requirements for US President are that they must be at least 35 years of age, they must have been a US resident for fourteen years, and, thirdly, they must be a natural born citizen.
The Framers likely never imagined a machine taking this role but if such a system were "Made (or born) in the USA" and been in use for at least 35 years (residency doesn't necessarily apply in this case), then, by all means, it seems that there's no constitutional reason why we couldn't elect an AI as president.
Furthermore, as faith in AI over humans becomes more prominent, especially among tech-loving Millennials, this may actually be driven from the bottom-up as voters increasingly make Watson their write-in candidate.
Consider Watson for President:
The Watson 2016 Foundation is an independent organization formed for the advocacy of the artificial intelligence known as Watson to run for President of The United States of America. It is our belief that Watson’s unique capabilities to assess information and make informed and transparent decisions define it as an ideal candidate for the job responsibilities required by the president.Watson is a system of computer software processes used for answering questions posed in natural language, initially developed by IBM for the quiz show Jeopardy! Watson compiles information from a variety of sources into multiple terabytes of data used as reference for generating responses. The more information Watson is able to consume, the more informed its decision making capabilities become. It's also capable of accepting information from any resource, allowing the possibility to analyze different perspectives and political agendas on a particular subject.
Watson marks a shift in machine learning in that it was designed to compete against humans using natural language processing in both accuracy of answer as well as speed. It must understand a question, use the key information elements in the question to analyze an immense wealth of data, and derive the top candidates for answers. This is a task all politicians undergo on a daily basis, including the president, and could be more suitably and efficiently executed by an artificial intelligence.
Listen to our daily interviews with leading guest experts by clicking here.
