I am an optimist about the role of technology, automation, and AI in contributing to the productivity levels and economic output of our species. Yet, from a metaphysical perspective, I am particularly wary of what can happen when AI is able to gain a level a general intelligence at par with our species.
If we were to plot a graph of global human economic output vs time, from the beginning of our time on earth till today, we would see three sharp exponential phases of increasing economic output: the agricultural revolution, the industrial revolution, and the modern computing and artificial intelligence revolution. There is little doubt that the artificial intelligence revolution has started and will continue to enhance human productivity at a scale that rivals previous two. Artificial intelligence refers to the ability of computing systems to carry out tasks that previously only humans could perform. Today AI can perform highly specific tasks like image recognition, electronic trading, remote sensing, medical diagnosis, etc. with accuracy and speed greater than humans. The silicon chips on which AI are programmed perform at lightning speeds that are much faster than firing of the synapses and neurons in our brains. Furthermore, these silicon chips can store far more data than our brains can. This enables our economy to exponentially increase output while reducing labor input.
I believe that purely economic arguments against AI and automation should not be entertained, especially arguments that reject AI and automation because they displace working class jobs. These arguments are anti-development in nature and resist the pioneering spirit of innovation and adventure that this country was founded upon. In fact, technophobic ideas against automation and AI the resemble the Luddite movements of the industrial revolution. There will always be a segment of the workforce that gets left behind when there are capital and technological gains in the economy. Simultaneously, it is true that the trajectory of this particular revolution will likely lead to millions of citizens could losing their jobs. The sheer number and proportion of individuals that will lose their jobs is staggeringly high, much higher than it was in the industrial revolution (McClure, 153).
While it is unfortunate that so many working class citizens will be replaced by autonomous machines, I believe that they must adapt and build new skills in order to participate in the economy. Charles Darwin, once wrote that ‘it is not strongest of the species, nor the most intelligent that have the greatest probability of survival; It is the ones that are most adaptable to change’. Morally speaking, it is up to the rest of society to help this segment of the labor force make this transition. We are already seeing vast amounts of venture capital investment influx into online platforms that allow individuals to learn new skills from the ground up such that they can contribute to the economy. And this influx is only going to increase in the future. Historically, automation has led to a decrease in certain jobs but a net increase in overall jobs. This will likely continue. Thus, I am not at all pessimistic about AI and automation when it comes to their impact on the labor market. Our planet has a cornucopia of untapped resources and AI will force us to find and utilize them.
The worrisome aspect of AI is in regard to its existential and metaphysical nature. Today AI is at the stage where it can carry out highly specific tasks with high accuracy and speed. This type of AI is known as weak AI; I am not fearful of weak AI as it is merely another capital tool that enables us to be far more efficient than we have ever been. Strong AI, on the other hand, refers to the type of AI that is no longer specific to the task at hand. I.e. The AI becomes capable of successfully performing a variety of non-specific tasks with greater accuracy and speed than humans. There are various proposed manifestations of Strong AI. Some of which include human brain interfaces (1), human brain emulation (2), general AI (3) and more. Either of these versions of Strong AI pose a direct and terrifying threat to the very existence of the human race. Nick Bostrom, in his book Superintelligence: Paths, Dangers, Strategies, explains various possible scenarios in which the default outcome of strong AI is doom. The most convincing and easy to understand of these scenarios are Infrastructure profusion and Perverse instantiation.
Perverse instantiation refers to the scenario in which a superintelligence discovers some way of satisfying the criteria of its final goal even though it violates the intention of the programmers that actually defined the goal. Let’s say we were to program an AI to make us happy. What ends up happening via perverse instantiation is that the AI begins to parallelize human facial musculatures to constant beaming smiles by manipulating facial nerves. It figured out a way to realize its final goal to a greater degree than the methods we would normally use. For obvious reasons, this is highly concerning.
Infrastructure profusion, on the other hand, refers to the notion that the only thing that the AI cares about is its reward signal; it is programmed to do whatever it takes such that it completes its tasks and receives its rewards for doing so. Thus, it will devote all the possible resources it has access to in order to carry out its task. For example, say the goal of the AI is to make ten paper clips. One would think that the AI would stop once the ten paperclips are made. But, if not programmed correctly, the AI would keep carrying out tasks in order the ensure that the astronomically low probability that it is unsuccessful has been minimized. It will recheck the paper clips again and again ceaselessly. In doing so it could possibly use up an unimaginable amount of resources. While this example is a microcosm, its implications are clear; If we give the strong AI some highly complex task, it will not know when to stop. It could potentially consume all of the resources and energy available in the entire universe in order to achieve its end goal.
I understand that these are just two theories representing one highly abstract point of view. Furthermore, many unprovable arguments against perverse instantiation and infrastructure profusion exist. However, the nature of these threats is so grave that even the tiniest probability of their occurrences deem that we must not ignore them. This is not to say that we must stop striving to make progress in the field of AI, but that we need to pay copious attention to the developments that are taking place. I commend Elon Musk and Sam Altman’s OpenAI project (4) and support Stephen hawking’s view that some kind of global regulating body must be set up so that we are in preemptive mode and not reactive mode, given that reactive mode will be too late (Caughill, 3).
In conclusion, I am optimistic about technology, automation, and AI when it comes to the global economic output, but the existential threat cannot be ignored.
Footnotes:
References:
Franssen, Maarten, et al. “Philosophy of Technology.” Stanford Encyclopedia of Philosophy, Stanford University, 6 Sept. 2018, plato.stanford.edu/entries/technology/.
Patrick Caughill, Artificial Intelligence Is Our Future. But Will It Save Or Destroy Humanity.
Nick Bostrom, Superintelligence: Paths, Dangers, Strategies