1. The Premise.

Artificial intelligence (AI) is becoming a general purpose replacement for all forms of human productivity. It is widely understood that every month, AI gains the ability to complete more tasks at a rate cheaper than human labor, while simultaneously maintaining or increasing quality.

1. Most economies distribute income based on wages, which is in turn based on individual economic output and productivity.
2. Most public budgets are funded primarily by taxing labor and consumption.
3. Most households hold little (relatively) free capital and depend on a paycheck.

The current arrangement requires human labor to remain a scarce resource. If this breaks, then every point of the chain also breaks.

1. Economies distribute income based on wages: give the wages to artificial intelligence, and the work gets completed for less capital, at higher quality, and at a faster speed.
2. Most public budgets are funded by taxes on labor and consumption: As households receive less wages as AI begins to take make labor an abundant resource, both wages and consumption will drop. Directly, public budgets will crash.
3. Most households hold little free capital: nobody is prepared effectively for the reckoning to come.

2. Economic Safety.

We use economic safety to mean three conditions that should hold during all stages to highly automated production.

1. Income security: Every household must retain a reliable share of what the economy produces, whether or not it can sell labor.
2. Broadly held ownership: The returns to the machine doing the work must be decentralized so that economic and political power do not become the same.
3. Resilient public finance: Governments must be able to fund existing services even if payroll and income taxes shrink as a share of the overall economy.

This is not an argument against automation. The reasonable individual can come to the conclusion that individual surplus in resources is generally a good thing. Abundance created by machines is the best plausible outcome, if gains are widely shared. We aim to address the gap between that outcome and the current trajectory of the established institutions.

3. Lag.

Existing economic safety nets rely on the assumption that employment is the normal state of being for any adult (or, at the very least, for a household) and that wages are the main thing to tax. Unemployment insurance pays out when you lose a job, not when the entire job category disappears. Pensions and retirement accounts compound and are funded by a working life. Healthcare, in many countries, is directly tied and often paid for by an employer.

Each of these institutions needs to be, and can be, reformed. Institutional reform, however, is measured in decades, and gains in AI capabilities are measured in months and weeks. The mismatch between speed is the key risk. We understand automation will arrive, but we must increase the speed of reform such that the institutions are prepared and that the transfer to an automated society is safe.

4. Directionality.

Truss PAES is not meant to promote or endorse a single program. Its purpose is to develop and pressure-test a portfolio of mechanisms that could, in combination and in theory, deliver economic safety.

We aim to be structured as a working group, combining economists, engineers, policymakers, and practitioners. We aim to publish short working papers, like this one, and maintain open questions centered around the future.

In the near term, we aim to conduct more research into the tax and benefit systems exposed to labor displacement, quantify fiscal exposure, and identify which mechanisms are deployable at which scales.