AI's autonomy

In a hyperconnected, fully digitalized and fast-paced world, automation is an indispensable resource. Techniques and parameters give shape to electronic agents that interact with human beings - they are most commonly known as Artificial Intelligence. These autonomous are fundamentally composed of information, further translated into coded instructions for the machine to manage tasks on its own. Taking this in consideration, it is necessary to reevaluate the concept of autonomy - after all these entities should have an equal share of both power and responsibility. 

Nowadays these artificial agents are part of our daily activities: they can be an app with traffic information, they emulate the human voice in bank calls, they can help doctors to save lives (1) and also drive cars causing fatal accidents (2). In Groundwork of The Metaphysics of Morals (3), Kant speaks about autonomy in a way that doesn’t take into account actions perpetrated by nonhuman agents with the capability to  self-legislate. The autonomy of intelligent machines has been a topic in the European Parliament recently: they recommended the creation of legal devices specific for robots; the parliament points out that more sophisticated autonomous machines should be seen as “electronic humans”, therefore, they should be convicted for any damage or harm they may cause. They further suggest that, in a case where robots interact with humans in an independent, autonomous way, an “electronic personality” is to be attributed to the machine (4). It is natural to imagine that the creator of an autonomous machine is fully responsible for its behavior and actions, but isn’t that the same as grounding parents for how their children misbehave? After all, we generally conceive that children’s behavior and actions are shaped by the “input” (or education) they got from their parents. 

I end this essay with a demand for further and deeper insights on the topic of AI’s autonomy and the ways we should analyze the machine’s responsibility. An agent, whether human or AI, can act independently and therefore should be deemed responsible for its own actions. These AI devices’ personalities are no longer shaped only by codes conceived by their programmers, but also by acquired learning. It is necessary that, just like humans, machines become responsible for what they do autonomously.  

—

References

1. R7: Inteligência artificial ajuda médicos a salvar vidas em hospitais. Available on: <https://noticias.r7.com/saude/inteligencia-artificial-ajuda-medicos-a-salvar-vidas-em-hospitais-03122019>.   

2. Tecmundo: Carro autônomo mata pedestre: quem é o culpado?. Available on: <https://www.tecmundo.com.br/mobilidade-urbana-smart-cities/204718-carro-autonomo-mata-pedestre-culpado.htm>.  

3. KANT, Immanuel. Fundamentação da metafísica dos costumes. In: Crítica da razão pura e outros textos filosóficos. Col. Os pensadores. São Paulo: Abril, 1974.

4. Report with recommendations to the Commission on Civil Law Rules on Robotics (2015/2103 (INL)). Available on: <http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML+REPORT+A8–2017–0005+0+DOC+PDF+V0//EN>.