In the scientific article of Kossuth radio held on the First of August 2023, Szabados Levente, senior advisor and co-founder of Neuron Solutions answered arisen questions about the potential application of artificial intelligence in economic tasks and defining economic policies.
Nine robots attended the briefing at the International Telecommunication Union event of the UN, held in the beginning of July in Geneva. Amongst these robots, one could have found Sophia, first robot innovation ambassador of the UN Development Programme (also known as UNDP). Is it possible that robots are capable of making better economic decisions and economic policies than humans?
According to Levente, the analysis of the definitions of both “better” and “decision” is required separately in this case.
In 2020, during an experiment conducted by Harvard University, a machine model was given the task of establishing a tax policy by adjusting the values of different taxation types. This policy had to be as fair as possible and had to maintain a motivational factor as well as to strive for a uniform distribution. The result of this experiment was an optimal tax policy. Thus, it was proven that a machine is able to make better decisions than the current ones.
However, how is it defined what is “better” from the point of view of the artificial intelligence?
The researchers from Harvard relied on their own knowledge of the system and their own concept of fairness when they trained the model. The maximization of tax income was a secondary objective yet being an economically significant one. Since machine learning models do not have an intuitive understanding of fairness or the difference that makes one state of the system being better than another, these concepts must be declared by the designer of the given model. This leads to the artificial intelligence reflecting the views of the designing person or group. Defining what is right, what is the goal and how to achieve it is a deeply human task.
Will these models ever be applied in intellectual work?
Levente told us that there are already numerous organizations that apply such artificial intelligences.
Well defined intellectual work that could be automated is basically intellectual manual labor, like writing marketing scripts based on a template or answering routine questions at customer service or handling assignments in a background system that directly depend on the request. Jobs like these are perfectly fit for automation by artificial intelligences. Nonetheless, designing good templates, reviewing generated texts and answers and booking these are responsibilities reserved only for humans.
In conclusion, artificial intelligence can find optimal solutions to economic assignments and optimize economic policies, but defining the details and context of these assignments remains to be a human task.
The full dialog is available by this link.
Follow us, for we will have more themes to explore soon!