We all believe in self-driving cars. If you've found this blog, that's pretty much a given. We believe that they'll be safer, faster, and above all more efficient. We believe that in the long run they'll revolutionize industries and lives. The economic benefits of self-driving cars are obvious to people plugged into technology. Car-based industries are ripe for "disruption." First trucking, then taxis, then commuters will discard drivers in favor of automated systems. Because, in the long run, it's better in almost every conceivable way to maintain and operate a self-driving car than a human driver.
Better, Faster, Stronger
But not everyone shares our views. Sometimes we have to convince laymen of the benefits of self-driving cars. To do so we might tally the yearly salary of a long-haul trucker, and compare it to the estimated cost of retrofitting a semi. If the cost of the retrofit divided by its lifespan is less than the trucker's salary, then it's clearly a better solution. The difference between the amortized cost of retrofitting and the trucker's salary when multiplied by the truck's lifespan is a clear indicator of the economic benefit of making the switch. At some point in the development of autonomous cars, this economic gain will become an imperative. Trucking companies who do not switch to autonomous fleets will be unable to compete, and the industry will change rapidly.
We bring up efficiency. There are already known-good heuristics for conserving energy while driving. There are optimal speeds and acceleration patterns that the average human is unaware of, or can't recognize. Electric self-driving cars will be more fuel-efficient, as they can implement these strategies reliably. They'll be better able to drive the most efficient route than a human, and better at conserving energy while accelerating and braking. Self-driving cars will, in general, consume fewer resources to move the same amount of mass the same distance because machines can optimize for these non-obvious problems far better than the average human.
In addition, there will be a drastic reduction in resources used to build cars. On average, people drive less than one hour per day, or about 4% of the time. Nevertheless, because a car is near-essential to life in many parts of the US, there are an estimated 260,000,000 passenger vehicles in the United States. Self-driving cars, used as a service rather than a possession, will eliminate the need for many US residents to own cars at all. It's incredibly optimistic to think that we only need 4% of our current national fleet. After all, driving tends to occur more during certain peak times (rush hour). But it's not unreasonable to think that we could eliminate 40% of all cars by more efficiently allocating car time, thus freeing up to 40% of the steel, plastic, and labor we collectively pour into the automotive industry.
And what about safety? To convince the uninformed we compare a human's reaction time (measured in seconds) to a machine's (measured in milliseconds). Because the machine can react faster, we reason, it is safer. It can brake or swerve or accelerate more quickly and accurately than a human. Its behavior is predictable where a human's is not. Via local networking, it can communicate more reliably with its fellows on the road than a human can. I am personally convinced that by giving control of our cars to computers, we can prevent accidents, and save human lives. And to me, that trumps any economic gain.
A traffic system can be thought of as the aggregate of all cars' actions and interactions. It isn't a mass of cars moving as an unstructured blob, it's a collection of individuals with different goals and needs interacting in a shared space. Traffic systems, when thought of as an aggregate, are functioning well when every car reaches its destination safely and quickly. We believe that when each car is working autonomously to create safety and efficiency, traffic will be more safe and more efficient. We believe this is true even with no communication between cars.
If every driver is attentive and skillful, traffic as a whole seems to move smoothly, but a few poor drivers, or good drivers making poor mistakes, can turn the entire system inside out. A single text message at the wrong time can cause delays, property damage, or death. We expect autonomous cars to drastically reduce the severity of traffic jams because each autonomous driver will be better able to zipper, follow, and take advantage of other strategies to reduce their footprint on the road. Autonomous drivers objective benefits over humans will bring benefits to the system as a whole.
Belief in the benefits of autonomous cars to society rely on this foundational assumption: if a system is an aggregate of individual pieces, and we improve each individual in the system, the system as a whole will improve as a consequence. We believe that improving each individual car will improve roadway conditions as a matter of course. Belief in the benefits of self-driving cars implies the belief that benefits trickle up from the individual to the collective when all the components are improved together.
As technologists, self-appointed guardians of the
galaxy future, we have an arsenal of arguments carefully selected to address the concerns of the uninitiated. But we have to remember that they're attached to their way of life. They're invested in the status quo. To convince them, we exhaustively attack their assumptions, break down their conclusions, and hopefully bring them around to our way of thinking. If we're successful,
Really though, isn't it a bit hypocritical to slaughter their sacred cows, and not our own? It's about time we turned some of those arguments on ourselves, and on the social structures we've chosen not to question.
Specifically, markets. Markets closely mirror traffic systems. It's not a perfect metaphor, but there are broad similarities. Each individual driver, or investor, or entrepreneur, or consumer makes decisions based on what they know, and what they want. These decisions don't consider the benefit to the system as a whole, but rather only how the individual can work in that common system to maximize personal benefit. Intelligent individuals account for the unpredictable environments created by others via defensive driving, or portfolio diversification. Markets and traffic systems are each a collection of autonomous self-interested agents, interacting in a shared space to accomplish individual goals.
Like traffic jams generally indicate some poor driving decisions, the health of a market is a broad indicator of the economic health and skill of the individuals participating. Good decisions by large numbers individuals conspire to create a smoothly functioning system, while a small number of people making bad decisions regarding say, real estate derivatives, can create negative events that affect all participants in the global economy.
When each actor makes intelligent decisions to maximize its own outcome, the collective benefits, and each individual benefits. If every person in the world seeks her own economic improvement skillfully and attentively, there will be plenty of resources to go around. Rand's "rational egoism" starts to look pretty attractive at this point. It's lovely to think that logical selfishness will save the world. After all, humans are good at being selfish. The only sticking point is "rational." Unfortunately for all of us, humans are incapable of being rational.
We have cultural and evolutionary hacks deeply ingrained in our brain. We're vulnerable to dozens of cognitive biases. They manifest in well known ways, e.g. the gambler's fallacy, the effort heuristic, the sunk-cost fallacy. There are a bevy of well-documented ways in which the human brain innately does not make the rational choice. Perhaps these biases were evolved to help us survive in the harsh world of the deep past – loss-aversion is likely to help an organism survive when food is scarce – they actively inhibit our ability to function in a market. We see value where there is none, and are often unable to recognize a paradigm shift as it surrounds us.
There are any number of examples of poor economic decisions being made by large groups as a result of cognitive biases. Bubbles flow in the global stream, vanishing and forming again around tulips, beanie babies, the internet. Millions of people are caught up in the rush, unable to step back and rationally evaluate their own actions. Vast amounts of resources are wasted on things that ultimately benefit no one. Cognitive biases trickle up from the individual to the collective when all components are flawed together.
Furthermore, our decisions are strongly affected by the state of our body. We evaluate costs and benefits drastically differently when hungry, as our brains prioritize calories over everything else. Rage, depression, and greed create regret. We get consumed by lust, or anxiety, and hurt people around us. Until we understand the processes that make up our own brain, we can never escape the chemical bounds of our intelligence.
Slow it down
Fortunately, as humans, we have a (fairly) effective defense against cognitive bias. We can slow down our thinking. We can learn to step back from a situation, and analyze it dispassionately. Subject it to rigorous logic, and reduce external problems to objective and actionable descriptions. Formal methods, applied via technology, have already improved the lives of billions. We enjoy a level of safety from disease and violence unprecedented in human history. This makes humans unique in this world: the ability to identify our own biases and root them out of our works.
Unfortunately, this defense is not accessible to most of the world. And it won't be any time soon. Most of the world lives in poverty. It's difficult, if not impossible, to learn to apply the scientific method when you don't know if you'll be able to eat tomorrow. Nobody learns mindfulness or rationalism while living in the slums of Mumbai, or in inner-city LA. These people rarely have the option to acquire these skills. They don't have the education, or the tools, or the leisure time, or the energy. All that they have is devoted to surviving another day. And they are forced to participate in markets to do so.
There are still dropouts, burnouts, and failures-to-launch among the affluent and educated, but many fewer. Some people natural talents and inclinations lean in other directions. They pursue beauty, or love, or hedonism over economic value. We call them artists or failures-to-launch, and we make jokes about English majors working at McDonalds. But this strongly implies even if we eliminated poverty, and provided free education to everyone, and figured out a way to permanently correct racial biases, and magically fixed every other systemic issue, many people still wouldn't be able to effectively participate in markets.
The vast majority of players in today's markets can't play well. The best of us barely beat the market. This strongly implies that the markets themselves are poorly allocating resources. It's visible when r is greater than g, in the constant barrage of spam emails, and every time a charismatic con man fleeces another mark. We humans are the weakest link in the supply chain. Markets buy and sell our mistakes more often than not.
And build it out
Technology freed humans from most manual labor. We no longer have to pull a plow, or spin wool. We built machines to do it for us. Every time we improve a process more resources are freed. We feed those resources back into improving the machines. Technological progress is a self-reinforcing cycle. As a result, we have more food, happiness, and peace than at any other point in human history.
Technology is beginning to free us from intellectual labor too. Research involves a few finger swipes instead of rooting through a card catalog. Accounting happens on indexable spreadsheets instead of paper ledgers. We spend so much mental time and attention on driving, and self-driving cars will free up so much of that for other pursuits. Those resources will be fed back into the system, and contribute to quality of life, research, art, philosophy, and other things of value to humans.
Only a small subset of us are qualified to build this technology. They do it by thinking slowly and carefully questioning the decisions they make at every part of each mental process. The vast majority of people will never learn this skill, but they'll benefit from it anyway. Our children will never have to learn to drive, never risk being hit by a drunk. We should be building a world where our descendants can live without worrying about making rent, retirement planning, or global economic crises.
Here's my proposal: humans should stop participating in markets. Not all at once, but over the course of my lifetime. Not all markets, but many of them. We're just not well-suited for it. We should go back to doing the things we were made for: eating, sleeping, living, and chasing happiness where we find it. It's time we truly accept that self-driving cars are better than human drivers, and follow it to its logical conclusion.
It won't be quick or easy, but it may be inevitable. If we can improve on human decision-making in an economic market even a little bit, then, in the long run, those algorithms will outperform all human actors in that market. Just like trucking companies will be forced to use self-driving cars to stay viable, all economic actors will have to use those algorithms to compete in the market. At that point, any human participation in that market will be a net economic negative.
Some markets are more susceptible to this than others, but over time, many markets will undergo this process. High-frequency trading is the start of this process, and it will only accelerate in the future. As we develop technology, this disintermediation of humans should be our explicit goal. As technologists, our endgame is to automate all work that can be automated. The systems we build should certainly optimize resource allocation, but it is more important that we free people from labor so they can pursue their own happiness.