The Benefits and Risks of Artificial Intelligence
The world is hurtling toward a dramatic change, and I’m not so sure it is ready for it. Let me elaborate:
In the eighteenth and nineteenth centuries, the first industrial revolution completely upended the way we made and sold things. For thousands of years, goods were handcrafted by lifelong artisans, which kept the rate of consumption low and promoted a slower, mostly-local economy. Around the middle of the eighteenth century, however, advances in steam power and machine tooling made it possible to rapidly produce new goods, which increased the rate of consumption and the scope of business influence, and it revolutionized the world economy. These changes improved the average quality of life for much of the western world. However, it was also incredibly disruptive, and the transition period was rife with poor labor conditions, lost employment and displaced workers, income inequality, and political upheaval. It took decades for all the details to smooth, and for the modern world to emerge.
In much the same way that steam-powered machinery changed the way we lived and worked, artificially intelligent machines are going to have a profound impact on the way we live and work. In fact, it may prove far more disruptive (and in certain ways destructive) than the industrial revolution. It is very important, for the sustainability of society and our culture, that we collectively consider the benefits and risks of these systems; and adjust our expectations and preparedness, and regulatory frameworks, as necessary. And when the situation calls for it, we must consciously opt to keep humans in the loop (even if not for a technical lacking).
Let’s first consider the impact of the narrow form of AI, otherwise known as “weak” AI, because that is what you will encounter today and in the near future. These narrow systems are inarguably non-sentient (this isn’t a new form of digital life), but are supremely excellent at very specific but complex tasks. For example, autonomous driving is a type of narrow AI. The autonomous vehicle will not talk to you about the weather, or pick a favorite song to play, or ask you how your day is going, but it will be exceptionally good at monitoring and understanding road and environmental conditions, and will make appropriate real-time decisions about how the car should react. Lola uses narrow AI by creating systems that will learn all about you so that it can make perfect flight and hotel recommendations.
Narrow AI will remain the dominant force of intelligent machinery for the next few decades, with encroachment on driving (both recreational and professional), the service industry, rote work involving data entry and discovery, financial transactions, and a host of other tasks. In many ways, these intelligent machines will provide us with huge benefits, but there are dangers.
The most prominent risk from narrow AI, which has been discussed at length (but perhaps not with the requisite level of urgency) is job loss. Unlike the industrial revolution, which did in fact displace a great deal of labor but did not eliminate most low-skill work, narrow AI could potentially wreak havoc with the low-skill labor market by destroying entire job categories without an obvious path for worker transition. The trucking industry alone employs 3.5 million people in the United States (over 1% of the entire population). Another 3.67 million people are employed in the foodservice industry, and many millions more work in retail. With just a handful of automation advances, if pushed, these three industries could be nearly eliminated over the next few decades, dumping over 10 million unemployed on the open market.
Additionally, automation will likely cause a continued (and accelerated) rise in the rate of income inequality. The companies that are responsible for this automation technology will reap enormous financial reward, and many will likely become effective monopolies.
But we should also consider the potential benefits of AI. For example, autonomous cars and automated services will dramatically reduce the number of accidents and incidental costs (nearly 33,000 people die every year from fatal auto accidents, and 150 people die per day from industrial accidents; automation could save most of those people). AI will also increase production capacity, and will be able to make many goods cheaper and more accessible. And expanding the scope of autonomous machinery will also significantly expand the demand for high-skilled technicians that must service and maintain these systems, increasing wages and quality of life for technical professionals. While the transition from a low-skill labor force to a high-skill labor force is not an easy hurdle, targeted training programs can help ease the shift. And expansions (or new creations) within the social safety net can protect people throughout the transition, including controversial ideas like Basic Income.
And, of course, a good implementation of many narrow AI systems will still keep humans in the loop. Most job categories are not yet able to be automated end-to-end, but will instead have the scope of responsibility decreased by automating parts of the job. To date, most companies have unfortunately approached this reduction in scope by dramatically downsizing the labor force, but there are alternative approaches; for example, we could instead decrease the length of the workday to fit the demands of the position. We should demand that automation work to improve our lives rather than simply to replace them.
Beyond Weak AI
Beyond the narrow application, however, it is worth remembering that there does loom a potential disruption without any useful modern comparison: general AI. That represents a risk whose impact is far more difficult to predict. General AI might not even happen; or it could be a few decades away, or centuries away. But there are large academic and professional interests working tirelessly on systems that might one day be capable of producing a true artificial consciousness. If this were accomplished, it would be a chaotic new paradigm that would rival anything imaginable The world would have birthed a new, ceaseless, universal labor force that is capable of fulfilling every skill demand that humans currently occupy. It would shake the very foundations of capitalism, and cause humanity to confront an entirely new form of society. And, of course, we would have created a new form of life, along with all of the demands associated with that (rights, laws, social norms).
Whatever the future holds, there is no question that the very broad applications of machine learning and artificial intelligence hold the potential to create the same (or greater) benefit and terrible disruption from the great industrial revolutions of the past. Society will need to be watchful, and prepare for creative solutions to whatever problems might arise from this dynamic change to how our economy and society functions.