This is my final essay for my humanities module Morals and Markets, where we discuss how markets shape morals and how morals shape market.
Karl Marx (1942) famously dreamt of a day where he could “hunt in the morning, fish in the afternoon, rear cattle in the evening, criticise after dinner, just as I have a mind, without ever becoming hunter, fisherman, herdsman or critic”. His vision of a state controlled economy never materialised as there was no way at that time for the state to efficiently assign resources for an entire country. There was also the problem of a central state that decides and controls production succumbing into capitalistic and cronyistic tendencies. With the impending arrival of a superintelligence, there might finally be a system clever and impartial enough to manage an entire economy.
Superintelligence is an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills. This definition leaves open how the superintelligence is implemented: it could be a digital computer, an ensemble of networked computers, cultured cortical tissue or what have you. It also leaves open whether the superintelligence is conscious and has subjective experiences (Bostrom 2006). It should be noted that such a superintelligence intellect is not strictly required for controlling the economy. There can be efficient allocation and matching of resources using optimisation techniques developed in the past decades, however for the economy to progress and for innovation to occur a form of superintelligence is critical in terms of leading the progress.
It may seem premature to talk about how superintelligence1 is going to change the economy, considering that majority of the development in the field now is on Artificial Narrow Intelligence - AI that specialises in a single area. It can beat a world champion in Chess, but thats the only thing it can do. There has been very little progress in terms of a general agent which is able to tackle a wide variety of problems in a way a human can. However, with the exponential nature of technology2 and the recursive self-improvement capabilities of AI a single improvement can lead to an intelligence explosion - improvements at compounding rates resulting in intelligence that humans cannot begin to grasp (Good 1965). Given this information, it is with near certainty that a superintelligence will be created by the turn of this century - likely much earlier.
What does this mean to the currently dominant lifeform on earth? How can humans co-exist with this new superintelligence? Turchin (2018) outlines 3 different possibilities for how the relationship might be. The first, and the most dire, is that the superintelligence is a ruthless exterminator. It might not necessarily view humans as an enemy but rather as an obstruction to its goal. This is further explained with a thought experiment about a superintelligence whose tasks is to maximise efficiency of paperclips production. After reaching a limit to it’s efficiency, it might decide to convert the human mass on earth to help power it’s paperclip factories (Bostrom 2003). Eliezer Yudkowsky (2008 p. 27) puts it eloquently:
“The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.”
The second outcome is where the superintelligence is subservient to humans. However, considering the level of intelligence difference, it is hard to phantom humans being able to effectively communicate and control a much superior intellect. The superintelligence is likely to have intelligence that humans are not even able to grasp, in the same way that an ant is not able to understand the concept of galaxies and space travel. The final outcome proposed is where the superintelligence acts as a benevolent dictator. It makes its decisions with the interests of humans in mind and adapts these goals as the population evolves. This seems to be the favourable likely scenario. Humans will be ruled by a central system which dictates how to best allocate resources.
It is inevitable that humans would lose almost all of their jobs to AI. It is hard to imagine a job in which a superintelligence would be worse than humans at. This requires humans to find other ways to sustain themselves. They would have to use the single advantage that humans have over AI, being human. AI improvement is a highly data driven process, and the superintelligence would no doubt collect tons of data about the world through its variety of sensors. However, it would not be able to replicate the exact way a human perceives the world. Therefore, it can pay for this data to have a better understanding of the world. Humans then have a form of universal basic income. However, its implementation is paramount. Done poorly, it can turn into a mirror of the capitalistic system we have today.
If the compensation is based on the quality and length of the data, it can lead to alienation of humans (Tucker 1978). Humans have to change their lifestyle to maximise the quality of data. This leads to a blending of work and life, where their life is no longer a part of their self, but external to them in terms of labour. Having no other possible jobs available, they will be alienated from the means of their own subsistence as well. They will also be alienated from each other, as they see each other as competition vying for the same limited resources. It might also be that if the AI is able to get sufficient data, its reliance on new data might decrease. The humans then become poorer the more value that they produce. Therefore, it is important that compensation is not tied to some measurement about the quality of the data.
With such a system in place, humans do not need to engage in work. They are able to spend their free time doing activities that satisfy their passion. One might say that Marx’s vision from earlier might finally be realised. However, will humans be free? Friedman (1962) argues that concentration of power is a threat to freedom. If a government has full economic power, it will be able to abuse it’s position. This leaves the citizens unable to make changes as the government has full control over all aspects of society. For example, individuals are unable to advocate and propagandise for change in society when the government has control over radio and other forms of infrastructure. In the same way, humans will be placed in a situation where all economic decisions are made by AI and therefore they can lose the ability to make radical changes to the structure in society. To combat this, there needs to be some platform for the superintelligence and humans to have a discussion and convene on how to change the structure of society. With the widespread adoption of the internet, it can serve as a platform for the voices to be heard and decisions to be made. Needless to say, the internet needs to remain independent and outside of any single platform (eg Facebook, Google).
A major argument for capitalism is the amount of goods and innovation that it creates. Without capitalism, there would not be innovation such as iPhones. There will be a drop in the amount of new products created. There a couple of ways of solving the problem in this system. The immediate idea is to allow the superintelligence to innovate and create new product. That will certainly happen, the superintelligence is likely to terraform the face of the earth to something that is undistinguishable. However, it is impossible for the superintelligence to perfectly know the wants and needs of humans. Also, allowing the superintelligence to control all innovation also relinquishes all forms of economic freedom. Humans would no longer have the freedom to participate in the economic market.
Recent advances in technology have introduced the idea of open source, decentralised, pseudonymous currency that are governed by deterministic algorithms rather than institutions. Any time a transaction is recorded, multiple computers in the network ensure that the transaction is valid before accepting it into its immutable data structure. Some currencies also have the ability to use “smart contracts” - where a contract is recorded between two users with a agreed value and condition. The currency is only transferred upon the fulfilment of the condition (verified by the network). This removes the requirement of trust that is needed for transactions today. Such a technology can be utilised to create a commons market of sorts. In this market, an individual with an idea can share the upfront capital required with other individuals to produce an item in a decentralised way. All aspects of the ideation, manufacturing, outreach can be produced using the network. Because many “bullshit jobs” (Graeber, 2013) are removed from the process, the actual cost of production goes down. Combined with the fact that these ventures can be funded together with like-minded people, the barrier to entry for production will go down.
To give a concrete example, a group of individuals want to create a new watch to commemorate a special historic event. They reach out to the watch enthusiast communities to gauge the interest for such a watch. They receive positive feedback and some of the members are even willing to agree to a smart contract for the watch at a discount from the future price. Encouraged by this, they then look for engineers and designers in the commons market to design their watch for them. They give the engineers and designers a sketch of their idea and enter in a smart contract with them for the blueprint of the watch. The blueprint is then verified by the manufacturers who begin work on it immediately. After it is produced, it is sent to the shipping company who verify the stock. They then ship out to the recipients of the watches who again are able to verify that watch has indeed arrived. Note in this scenario, the exchange of value is in a linear fashion. Instead of the entire upfront cost being borne by the entrepreneurs, they only have to bear the cost of the initial cost in the chain. Each of the subsequent costs is transacted purely between the transacting bodies. And due to the smart contracts, each of them can trust that they will be paid by the next person in the chain once they finish their job. Again, this significantly reduces the amount of upfront capital required for production.
No part of the system described above requires a superintelligence. It can definitely be helped by AI, for example in matching customers with service providers, but these are no more complex than Spotify recommending you a song you might like. Yet, such a commons market does not exist today. With a capitalistic market dominating the economy, it is impossible for such a market to compete with the large economies of scale enjoyed by big companies who are able to exploit low level workers with meagre pay. Compared to the current market, the commons market would probably cost more per product to make but the profits would be redistributed equally between each of the participants in the production chain. For such a commons market system to work, the current capitalistic market needs to collapse before it can be replaced by such a system.
Marx dreamt of an utopian paradise in which people were free from labour. However, his era did not have the tools required to execute on such a vision. By shackling ourselves under the control of a superintelligence overlord, we might just finally be able to free ourselves.
Bostrom N. (2003). “Ethical Issues in Advanced Artificial Intelligence”. Cognitive, Emotive and Ethical Aspects of Decision Making in Humans and in Artificial Intelligence.
Bostrom N. (2006). How long before superintelligence? Linguistic and Philosophical Investigations, Vol. 5, No. 1, pp. 11-30.
Friedman, M. (1962). Capitalism and Freedom. Graeber, David, 2013. On the phenomenon of bullshit jobs. Retrieved from http://strikemag.org/bullshit-jobs/
Good, I. J. (1965), Franz L. Alt; Morris Rubinoff, eds., “Speculations Concerning the First Ultraintelligent Machine”, Advances in Computers, Academic Press, 6: 31–88, doi: 10.1016/S0065-2458(08)60418-0
Marx, K., & Engels, F. (1942). The German ideology. London: Lawrence & Wishart.
Tucker, Robert C. (Ed.) (1978). The Marx-Engels Reader. Economic and Philosophic Manuscripts of 1844 p. 70-81
Yudkowsky E. (2008).”Artificial Intelligence as a Positive and Negative Factor in Global Risk”. Global Catastrophic Risks, ed. Nick Bostrrom and Milan Cirkovic (Oxford University Press): 308-345.