What Should We Reasonably Expect From Artificial Intelligence?

41 Pages Posted: 25 Jul 2022

Date Written: July 14, 2022


Artificial Intelligence (or just AI) is one of the most pervasive and cutting-edge technologies of our time. It is already present in a variety of sectors, such as agriculture, industry, commerce, education, professional services, smart cities, cyber defense, and so forth. However, there seems to be a misalignment between what AI systems can currently deliver to mankind and the results some people expect from it. This original misalignment leads to two unwanted outcomes. Firstly, some people expect results from AI that it – at least in its current stage of development – simply cannot deliver. Secondly, people are dissatisfied with what AI is already capable of providing, even though such provisions may be enough in many contexts.

In this article, the author points out that this original misalignment stems from the false premise that AI systems should always provide high accuracy rates, many times higher than human standards, no matter the context. By unfolding different market applications included in the general term “AI”, the author demonstrates that it is not a single, monolithic concept. On the contrary, AI embraces a wide variety of sector-specific applications, each of them with different purposes, inherent risks, and desired accuracy. The article then demonstrates that each purpose should target a different accuracy and transparency rate, on a case-by-case basis. Depending on the context, AI systems are more than welcome, even if they eventually provide accuracy or transparency rates lower than human standard. Consequently, the author advocates for a realignment of expectations fine tunned to 1) connecting the debate to what AI is both in the lab and in the field ; 2) better understanding its potential as well as its accepted levels of accuracy and transparency in each context, considering different purposes and inherent risks of the activity to be automated; and 3) providing more solid guidance to regulators, developers, and customers.

Keywords: Artificial Intelligence, Accuracy, Transparency, Regulation, Innovation

Suggested Citation

Parentoni, Leonardo, What Should We Reasonably Expect From Artificial Intelligence? (July 14, 2022). Available at SSRN: https://ssrn.com/abstract=4162871 or http://dx.doi.org/10.2139/ssrn.4162871

Leonardo Parentoni (Contact Author)

UFMG Law School ( email )

Av. João Pinheiro, 100
Belo Horizonte, Minas Gerais 30130-180

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics