In car factories, robots assemble cars. In logistical centers and warehouses, robotic machines sort merchandise, bar code it and stack it. At home, we ask a small box to play a song or list nearby restaurants serving gyro. On the phone, we exchange information and manage transactions by talking to a bot powered by AI. The growth of AI applications in a short period of time is already impressive and will become even more pervasive and dominant in the work place in the near and distant future. Advances in AI will have far reaching implications for jobs and incomes, employment and the nature of work. Parents with children in elementary school should already be discussing jobs with their children so they don’t end up with dead-end career choices.
To understand the implications of AI on jobs, we can think of it as an example of outsourcing. The first significant outsourcing of work was to machines and took place after the first industrial revolution. The second outsourcing of work was to computers. The third outsourcing of work was to send it abroad, mostly to countries with low labor cost or unique knowledge expertise in some field. AI is the fourth phase of outsourcing of work, with the difference this outsourcing can go far beyond what is possible with machines and computers. Beyond doing work that handles hardware and software, AI machines have the potential to learn on the job and execute tasks out of reach to humans.
Even without going into the most far-reaching capabilities of AI, we can ask what jobs are more likely to survive AI and what the consequences might be for human employment. Based on present estimations, jobs with a good chance to survive are those that require:
- Social intelligence and people skills
- Creativity and coming up with ad hoc solutions
- Working in unpredictable environments
As advances in AI threaten more and more jobs, the question is how AI and human work will co-exit in the future. There is considerable disagreement about the impact of AI on employment. One possible and very dire scenario envisages AI machines gradually displacing humans in a growing range of jobs. The result will be massive unemployment, first in lower-skill jobs and then progressively in higher-skill occupations. We can call this the substitution scenario. Under this scenario, AI will destroy human jobs faster than it will create new ones. Critics of this scenario point to the historical experience from previous industrial advances. More jobs were created than destroyed. However, we cannot ignore the fact that old factory jobs that paid very decent wages have already been lost to the machines. AI is much more potent than previous technological revolutions and as such it precludes reliable projections from past experience. The alternative scenario is one of a symbiosis between AI and human work. AI machines work alongside human work and help to expand economic output (GDP) beyond the level possible only with human employment. We can call this the complementarity scenario.
Both scenarios face a couple of crucial constraints. The first constraint is related to the purchasing power and size of the economy. For example, the substitution scenario must explain how an economy will retain its purchasing power to absorb the aggregate output if machines keep displacing more and more human workers, thus destroying incomes. The complementarity scenario must explain how far an economy can grow in size to accommodate both machines and a fully-employed human labor force. To the extent economies are constrained in size, cost efficiencies will likely favor the substitution over the complementarity model.
The second constraint relates to availability of a work force possessing skills for a knowledge-driven economy. Expansion of the usage of machines requires increasing numbers of workers capable of handling robots. Speaking of the US only, there is currently a significant deficit in this type of workers. The present state of education in the US, especially in analytical and technology-related fields, can stymie the inroads of AI.
AI has the potential to lead to unusual consolidation of incomes and wealth among a limited number of firms that master superior efficiencies in the use of AI technologies and among fewer workers who master high-level skills. The result will be lower or no incomes for broad segments of working-class people. How should then societies respond to this problem? One proposal is the adoption of UBI (Uniform Basic Income) offered to all and high enough to cover basic needs. Another proposal is for governments to enhance the purchasing power of workers through subsidies that lower or eliminate the cost of education, health care, raising children and caring for family members, to mention some. A third, innovative proposal is the Social Investment Stipend (SIS).* SIS will be paid to those who dedicate their work (work released thanks to AI) to advance socially beneficial goods, like care of others, education and community service. (I would also add advancement and appreciation of the arts.) SIS is supposed to be a better solution because it avoids the idleness that goes with UBI. Work plays an important role in giving people a sense of contributing to society and, hence, a sense of dignity. Therefore, SIS is a way to reward and incentivize individual engagement in society.
Obviously, the above and other solutions, require decisions that will reflect political choices. Ignoring the downside risks of AI can spur popular resistance to technological advancement. Ignoring the right of all people to benefit from the immense benefits AI can bring can very well create a backlash similar to the one we see now regarding international trade. Sharing common windfall gains is not new to market economies. Alaska and Norway use oil revenues to fund income-boosting and pension programs. A rational approach dictates that societies weigh the benefits and costs of alternatives and adopt arrangements that maximize the social good.
* SIS is advocated by Kai-Fu Lee, Chairman and CEO of Sinovation Ventures and former president of Google China; his essay appeared in the WSJ, Sept. 25, 2018.
Besides the book Life 3.0 by Max Tegmark (mentioned in the previous post), other books on AI and jobs are The Future of Work by Darrell M. West and Human + Machine by Paul Daugherty and H. James Wilson.
Let's Reason 