Pursuing Low Error Rates with Technologies

 Pursuing Low Error Rates with Technologies
Date
Inefficient Machine Learning
Machine learning (ML) is a fairly new concept among the emerging technologies of the 21st century, which denotes algorithms and statistical models that allow computer systems to perform some tasks and make inferences from large datasets (Mahesh, 2020). With the huge amount of data collected by the various computer systems present today, the need to have machines capable of making inferences from the data and solve complex problems becomes vital. Currently, there are multiple areas where machine learning algorithms have been put into practice: analyzing medical scans, game playing, language translators, and most recently, image recognition systems that have become crucial in enhancing security and authentication (IEEE Spectrum, 2021). Efficiency and the low error rate have been the ultimate objective of most ML algorithms with regards to image recognition, bearing in mind the need for precision with security systems where image recognition has been greatly put in use. As much as there is a myriad of benefits with the use of ML algorithms in image recognition, there are also problems associated with too much dependence on ML. One of the greatest concerns with pursuing low error rates with ML in image recognition takes an ethical dimension. Pursuing lower rates in image recognition has unwelcome environmental impact, which is unethical from both the deontological and utilitarian points of view.
The most commonly employed ethical theories in the contemporary era, the deontological theory and utilitarianism, are the most applicable ones in the issue of pursuing low error rates in image recognition with ML. Deontological ethics, otherwise referred to as Kantianism, is based on the argument that certain acts are to be pursued as an observance of moral obligation or acting based on the duty that one is ought to attend to (Forcehimes, 2018). Utilitarian ethics, on the other hand, is based on the notion that the morality of a person is dependent on the ability to do the most good (Forcehimes, 2018). Utilitarianism is a perfect example of consequentialism because the morality of an action is determined based on its consequence. Approaching the issue of low error rates in image recognition from both deontological and utilitarian perspectives paints a picture of total disregard to morality with the application of machine learning.
From a deontological perspective, ethics demand that humans have a responsibility towards the environment. Those who subscribe to this school of thought may argue that anything that destroys that which has an intrinsic value, such as the natural environment, is morally wrong (Dursun & Mankolli, 2021). There is tangible evidence that pursuing low error rates in image recognition systems to as low as five percent would require an enormous amount of computing resources and energy that consequently result in a large carbon footprint that can only be equated to that produced by large cities (IEEE Spectrum, 2021). There is a corresponding relationship between the amounts of carbon dioxide (CO2) produced and the computing power, with higher computing power resulting in more CO2 released to the atmosphere. In the case of efficient...