When AI takes your job...

(Source: here) 

Argentina's President Javier Milei opined in the Financial Times that Artificial Intelligence (AI) must be permitted to develop "without premature regulation". Many will disagree, because they fear AI-enhanced machines will eventually destroy jobs with severe consequences for both employment and inequality.

It is not unusual for new technologies to attract criticism, which then usually fades away as fresh areas of comparative advantage emerge and create opportunities and jobs elsewhere in the economy. Indeed, this type of economic renewal explains why Western economies have been able to sustain high or full employment through countless cycles of technological innovation from steam and electricity via modern transportation to computers and the internet.

Having said that, new labour-saving technologies do create serious short-term disruptions in labour markets and, in the case of AI, such concerns may be particularly pertinent, since some prominent proponents of AI claim that AI may even replace all labour.

To see how this might happen, imagine a simplified economy with just one worker, one machine, and one technology, AI. Since AI is embedded within the machine, it follows that advances in AI technology make the machine more capable over time. Eventually, the machine becomes so ‘productive’ that it is preferred to labour, rendering the worker redundant as far as production is concerned.

This scenario is precisely what many AI critics have in mind - a scenario in which machines do all work, while the vast majority of erstwhile working people are thrown into unemployment. Mass unemployment brings extreme inequality in its wake as a few accumulate great wealth from owning AI, while the masses have no income at all.

Personally, I do not think the nightmare scenario of AI critics is likely to materialise, although we may well move a great in that direction before things change. For one, the nightmare scenario does not make much economic sense. If the entire population - barring a handful of billionaires - has no income, then who is going buy the products made by the machines? Without anyone to buy goods, prices collapse and so do the returns to machines and AI. Nor do I see the nightmare scenario as politically plausible, because voters - the vast majority of whom stand to lose their jobs to AI - will surely vote against policies that guarantee their own impoverishment.

Still, so far there have been no serious attempts on the part of Western governments to prevent the nightmare scenario from materialising. If anything, the trend is heading in the opposite direction. Successive United States administrations have shied away from applying readily available anti-trust legislation to the big AI monopolists for fear of upsetting the stock market, while the Trump Administration actively de-regulates the business environment and grants AI executives political influence at the highest level of government.

My own view is that AI will require early and intelligent regulation to guaranteed that the benefits of the technology accrues to all of humanity. The argument for regulation rests on the view that AI may indeed be different from other types of innovation. Specifically, in previous cycles of innovation workers were able to maintain complementarity with successive vintages of technology by accumulating human capital, but is not the case if AI does indeed render workers entirely obsolete.

Should this begin to happen, a number of possible policy options spring to mind. At one extreme, governments could ban AI technology altogether. However, history has not been kind to Luddites. Bans on technology are often motivated by emotion rather than rationality and tend to be myopic rather than far-sighted. They are usually futile too. A far better way to guarantee economic and political sustainability in a world, where work is carried out by AI is to ensure that workers receive an income stream from AI even if their work is no longer required.

One way to deliver income to obsolete workers is to tax quasi-monopolistic AI companies and distribute the proceeds somewhat akin to unemployment benefit. Unfortunately, this approach suffers from several short-comings. For one, CEOs of large tech companies are effective lobbyists, who may well scupper efforts to tax their businesses.

Another problem is that even if taxation is possible there is no guarantee that income will be fairly distributed, because most governments favour some groups over others. For example, anti-immigrant political parties might withhold income from residents with immigrant backgrounds, while Left-wing governments might over-allocate income to poorer segments of the population at the expence of wealthier groups.

In addition, there is the obvious problem that levelling recurrent heavy taxes on AI companies year after year could disincentivise them from investing, thereby killing the goose that lays the golden egg.

The better way forward is to democratise AI by means of a one-off redistribution of AI shares to the entire population along with a credible commitment not to repeat such a redistribution. Once the AI shares are distributed, then everyone receives a share of the income stream generated by AI-augmented machines forever.

To illustrate how this might work, let us briefly return to the simple stylised model from earlier in which we now make one important change, namely that the machine is owned by the worker instead of the monopolist. Nothing else changes. The AI-machine still does all the work and the worker is still redundant, but the income stream from production now ends up in the pocket of the shareholder-worker. The income in turn gives the worker the purchasing power to buy the product made by the machine. The big inequality problem from earlier is eliminated and unemployment is no longer associated with stigma or poverty, since workers have both income and free time to engage in leisure on a full time basis. Philip Larkin’s toad no longer squats on humanity.

Needless to say, redistribution of AI shares will be challenging, at least as challenging as any other major wealth redistribution in history. The silver lining is that the logistics of redistribution of AI shares should be fairly straight-forward. I propose the establishment of a system of income accounts along the lines of pension funds. Each resident in the country is assigned an account at birth or at the point of acquiring residency. Listed shares in AI companies are then distributed equally to the accounts and dividends provide a living wage indefinitely.

Since no one works and everyone receives a steady income for life from AI shares, there is no longer need for conventional pension funds to guarantee income after retirement. Assets from existing pension plans, such as bonds, currencies, commodities, hedge funds, and other investments can therefore be transferred to the new income accounts to provide additional income and diversification to mitigate risk.

Income account holders should be allowed to have a meaningful say in how their income and investments are managed. For example, income account holders should be free to borrow for, say, education (for the sake of acquiring knowledge rather than increasing productivity) and then repay the debt over the rest of their lives.

Democratisation of finance along these lines outlined above would have two important positive effects. First, it would improve the population's knowledge of finance significantly and since current income (as opposed to distant pension income) is more directly affected by government policies more widespread ownership of financial assets should pressure politicians to run the economy prudently. Second, broader access to financial assets should help to reduce poverty, especially in poorer countries. Access to financial instruments is still mainly the domain of the rich and the inability of the poor to access to financial services remains one of most important factors keeping the poor poor.

Conclusion 

It is already evident that AI will replace many jobs in Western economies, starting with jobs in white-collar sectors and then eventually in blue collar sectors as AI augments the performance of robots. It is therefore not pie-in-the-sky to suppose that workers could one day be rendered obsolete by AI as far as production is concerned. When this happens, the only way that guarantee economic and political sustainability is to socialise the ownership of AI company shares so that everyone continues to receive an income. Since redistribution of AI shares will prove extremely challenging, it makes sense to begin a debate about this tricky subject right away.

The End