An Economy Simulated by Artificial Intelligence

 

Salesforce engineers have developed an artificial intelligence (AI) model to try to distinguish a tax policy that would offer a fairer world,reports the site SiecleDigital.fr. The system they created is capable of identifying the consequences of different tax models in a simulated economy.

Currently, governments around the world collect taxpayers' money based on their income and then use it to finance public projects or redistribute it in social assistance. Explaining in this way, the model seems ideal, but for a few years now, we have noticed that inequalities are constantly increasing.

Balance is obviously very difficult to find. Many researchers have worked on this issue without finding better solutions. Faced with this impossibility, why not use the machine?

Salesforce engineers believe that AI would create a fairer model. So they developed AI Economist. This model identifies optimal tax policies through a simulated economy.

An economy simulated by AI

The model used by Salesforce interacts workers with a two-dimensional world. These workers are responsible for collecting wood and stone and then exchanging these resources with other workers. As in the real world, they all have different skill levels. The most qualified use the accumulated resources to buy wood and stone to build their homes.

At the end of each year, workers are taxed at a rate designed by a manager created by the AI. The objective is to allow workers to increase their productivity and income. This same simulation is repeated millions of times in order to find the most efficient model.

Scientists believe the model will ultimately determine the best solution because both workers and the manager are managed by AI. This allows the two to adapt over the simulations.

Thus, some workers reduce their productivity to fall into a lower tax bracket. We can also see concessions from decision-makers and workers, which makes it possible to create a particularly realistic simulation.

At the moment, AI fails to capture all of the complexities of the real world and human behavior, but it is a matter of time, according to the researchers.

An idea difficult to accept

However, leaving a decision of this political importance to a machine is not easy for society to accept. 'It would be completely innovative to make tax policy less political and more data-driven,' said Alex Trott, a member of the Salesforce research team, who understands that novelty can be scary.

However, initial results point to a model that would be 16% more correct than those currently used, in terms of maximizing productivity and equal income.

Cite this page