The Economics of Artificial Intelligence
“Artificial Intelligence is a technology like any other, it will create new jobs to replace the ones it makes obsolete”. This view is based on a category mistake about the nature of Artificial Intelligence. Artificial Intelligence is a substitute for the labourer, not merely for certain of their routines. This presents profound challenges to our society and economy in the future which may be addressed through developing general knowledge.
Here’s something governments need to be thinking about yesterday: Artificial Intelligence technology. The challenges this technology pose for our future are profound. And if, as is incredibly likely, the government doesn’t start thinking about this, then you, reader, need to for yourself.
In a recent paper, I developed a model of Artificial Intelligence technology which I then released into a model of the economy to see what would happen. In three particulars, the effect of modern Artificial Intelligence technology was profound: labour employment, production technology and income distribution. There is no future for routine labour, our production technologies will begin to resemble those of a technological utopia and our distribution of income will in all likelihood become more unequal and not reflective of returns to labour.
There is a solution to address the inability to obtain remuneration through routine labour, which is the development of general knowledge. I’ve written elsewhere about how general knowledge supports economic development by facilitating the ability to develop specialised knowledge more quickly and in more “directions” of how and why to engage in economic behaviours. It transpires also that it facilitates creativity and promotes the ability to exercise judgement.
What is modern Artificial Intelligence technology
Artificial Intelligence technology is defined by its replication of human psychology in the machinations of a machine. One of its greatest developers, Herbert Simon, was also a founder of cognitive psychology, using computers to replicate and simulate the workings of the human mind, thereby creating an “artificial intelligence”. The advent of machine learning means that not only can we write into a computer a replication of the algorithms (behavioural rules) which exist in the human mind; we can write into a computer a process whereby they may be updated.
So a computer can replicate the structure, function and evolution of the human psychology. All that is required is to connect those computations to a physical mechanism and we can also replicate the behaviour that psychology supports. The psychologies and behaviours it can replicate are bounded only by our ability to uncover the routines of thought underlying human behaviour.
It is therefore a category mistake to say of Artificial Intelligence that it is but an extension of human capabilities as other technologies in the past have been. It is not merely a compliment for human labour by automating certain of its routines. It is a substitute for routine human labour as a whole.
There are only two particulars in which Artificial Intelligence cannot replicate human behaviour, and be a substitute for human labour as a whole. It can be argued (and Godel’s theorem can be understood to in fact prove) that the ability to be genuinely creative — realise connections outside the delimitations of process and procedure — is reserved for the consciousness. Until we resolve the problem of asexual reproduction of a conscious entity, we will find reserved for humanity those tasks which require creativity.
Creativity is the ex nehilo creation of connexions (the root of all art and science), but also the creating of connections between statements and their “truth-values”, the assent to a proposition by the consciousness which is the exercising of judgement. For these non-routine tasks and for these alone is Artificial Intelligence technology not substitutable in principle.
Establishing that Artificial Intelligence technology is a substitute for all routine human behaviour is all we need for placing it within a model of the economy as a complex evolving network of individuals acting on the basis of their psychology and social position to assess its economic impacts.
The economic consequences of Artificial Intelligence technology
The most profound economic impact of Artificial Intelligence technology is the direct result of its being a substitute for all routine labour. It is relatively straightforward to demonstrate that in theory Artificial Intelligence technology will replace all routine human behaviour with certainty.
If it is a substitute for routine human labour Artificial Intelligence technology can “do” the same “things” as routine human labour and both cheaper and more effectively. The only labour reserved in principle for humanity is that which is non-routine, that which requires creativity and the exercise of judgement: the artists, the professional creative, the professional decision-makers and strategists.
Moreover, being not merely the replication of a human being, but one where the executing of algorithms is bounded only by the friction afforded electrical currents within the machine itself, the only real constraint (outside of time to build) on production technology is the availability of base resources. The labour exerted in ordering an Artificial Intelligence to produce 1000 widgets is really equivalent to that exerted in ordering it to produce 1 widget. So we are really moving toward a technological utopia, the one which both Keynes and Marx predicted.
The effect of this on the distribution of income would be profound. The “scaling” ability of production technology would amplify the “winner-take-all” dynamic introduced into the distribution of income by modern information communication technology. Income would not reflect labour exerted at all, but rather who has the rights to the income accrued by Artificial Intelligence. Not only can we expect the distribution of income to become more unequal, we can expect it to not reflect returns to labour exerted.
How do we maintain capability to lead a Good Life?
Amartya Sen has written convincingly of the imperative of the liberal state to ensure, within reason, the liberty of individuals to live the life they choose. The greatest impediment to this in the foreseeable future is the inability to obtain remuneration through routine labour caused by Artificial Intelligence. Redistribution has a part to play in addressing this problem to be sure, but it addresses the symptom, not the cause of the issues raised by Artificial Intelligence technology.
The only non-luddite manner in which to rectify the issues raised by Artificial Intelligence technology is to ensure individuals have the ability to engage in non-routine labour.
That ability can only exist if individuals have the ability to be creative, and to exercise judgement: to make connexions in their minds. But creating connections isn’t really the problem, and we really can’t do much to encourage a process which is actually by definition free of the delimitations of process. The problem is increasing the likelihood individuals will assent to the new connexions they create — incorporate them into their minds where they can influence their thoughts and actions.
I’ve written elsewhere about the “Made to Stick” theory (in the context of political violence, rhetoric, financial markets and economic development). Knowledge grows, new connexions are incorporated more readily, when the knowledge being built upon is of a more general nature. General knowledge (contra specialised knowledge) forms a core on which new knowledge may build in a consonant (rather than dissonant) fashion, requiring fewer connexions to be incorporated to “complete” any new idea in the mind because a number are already present.
The having of general knowledge increases the likelihood any new connexion will be incorporated, and this increases the ability of the individual to be creative and to exercise judgement. That, in turn, increases their ability to engage in non-routine labour and obtain remuneration which is inoculated against Artificial Intelligence technology.
If governments won’t do it, do it yourself
Martin Ford has written of a terrifying vision of a jobless future brought about by Artificial Intelligence technology. Modern economic theory explains exactly why this vision has not only credibility, but plausibility. But economic theory also suggests a manner in which one may inoculate oneself and one’s children against the economic impacts of the technology: developing general knowledge improves the ability to be creative and to exercise judgement, to engage in the non-routine labour which is, in theory, the sole preserve of humanity until we solve the problem of asexual reproduction of consciousness.
Governments are by definition the agents of the dominant entities in society. They do not, reader, work for you. They can’t, otherwise they would lose office. So do not count on them to inoculate you or your children against the effects of Artificial Intelligence. Still more so when that would require radical reform of the education system to a) radically overhaul curricula and b) vastly improve the quality of teachers by radically improving pay and training.
In this instance however reader you are lucky. This weekend, pick up a new book, go to a museum or a concert, and begin the process. As I’ve said elsewhere: learn a little of everything. Learn science and mathematics and technology. But also learn art theory and history, learn music, learn literature and classics, learn history, learn linguistics, learn religion, learn philosophy. By simply reading, or going to a museum or concert you can inoculate you and your children against the economic impacts of Artificial Intelligence technology.
