Salesforce has proposed using artificial intelligence (AI) to develop a tax program.
American researchers proposed using artificial intelligence with reinforcement learning to form taxation algorithms that would maximally take into account the current economic situation in a particular country and could respond flexibly to changes in the level of income and productivity of all participants in the system.
The project, called AI Economist, does not use the previous experience of mankind in its work and focuses only on the current situation. In an experiment that was conducted in a playful way in a virtual world, the AI developed by Salesforce allowed to create a 16% more efficient model of progressive taxation than that created by economist Dr. system. During self-study, AI Economist came to the conclusion that the optimal taxation system would be a higher rate for the rich and poor and a lower tax for the middle class.
The system proposed by AI Economist tries to equalize the income level of all inhabitants of the virtual world, taking into account the degree of productivity of a particular “person”, which is formed taking into account, among other things, the desire to work at the tax rate established by the “state”, as well as the skills of a particular character. During the simulation, the “world” created by Salesforce was able to “live” for several million years, during which the tax rate was continuously changing, taking into account the situation in the simulated reality.
Salesforce stressed that creating a gaming world with virtual residents and training artificial intelligence based on this simulation has confirmed the potential for using AI to adjust the tax system. At the same time, AI is not perfect enough for the solutions it proposed to be applied in reality right now - for this it is necessary to significantly complicate the "game" on the basis of which the system will learn. But, as the researchers note, some solutions formulated by artificial intelligence may be introduced into the real taxation system in the future.