Movies have often filmed doomsday scenarios where Artificial Intelligence (AI) driven robotics impose a threat upon humanity. A classical example is the movie Terminator 2: Judgement Day, where Skynet attempts to rule Earth using AI driven killer robots. These depictions have led to debates where optimists and pessimists argue whether embracing AI is beneficial for society. Pessimists would argue that a brief software malfunctioning may lead to the robot willing to decimate the owner. However, perhaps it is too simplistic to blame human-build AI robotics for their malfunctions.
A critical element to consider is the risk of individuals exploiting AI technology for harmful needs. Several global leaders in AI such as Google recognize the dire need for preventive measures and therefore have established countermeasures such as Adversarial Logit Pairing (ALP) and CleverHans. Nevertheless, the fear remains that individuals with harmful intentions possess the capabilities to assert disaster. To minimize the risk, a sophisticated correction mechanism must be imposed to eradicate harm. Several institutions have attempted to establish a responsible framework for the creation and usage of AI. For example, the Institute for Ethical AI and Machine Learning emphasizes the importance of conscious human augmentation (assess the impact of one’s input upon the output created by AI).
Besides the possible explicit harmful intentions, harm caused by AI can also arise due to human error embedded in the coding. As expressed by software developers, the likelihood of AI malfunctions rises as the software grows in complexity, implying a near impossible likelihood of perfect functioning software. An example is the test deployment of Tesla’s autonomous vehicle which resulted in a collision. The AI was unable to distinguish an object from the environment and this ability is dependent whether the developers have provided sufficient instructions to the system. Unfortunately, the collision is the physical evidence of embedded human error.
To conclude, Hollywood’s depiction of AI driven robot killers appears misled. Rather, the possible “threat” imposed by AI is likely sourced from humans with harmful intentions or by coincidental human error. Nevertheless, it remains important for organizations willing to merge society with AI to prevent human misconduct and identify current/future human errors inherited by AI.
Sources:
https://ai.google/static/documents/responsible-development-of-ai.pdf
https://ethical.institute/principles.html
https://humanerrorsolutions.com/problems-and-solutions-ai-and-human-error/
https://www.information-age.com/artificial-intelligence-set-fix-human-error-123466675/
https://thenextweb.com/neural/2020/09/18/a-beginners-guide-to-the-ai-apocalypse-killer-robots/
