FIN 639 Problems with Current Applications of AI PDF

Preview

Summary

In this course, you will review 2 to 3 subject per class since it is 3 hours long, they will help you for the midterm, this is the number 23...

Description

Problems with Current Applications of AI AI used in driverless vehicles, speech and facial recognition, language translation, lip-reading, combatting spam and online payment fraud, detecting cancer, law enforcement, logistics planning, and language translation. Much of this AI is what can be described as narrow AI, that is, AI designed to solve a specific problem or familiar task, such as to play chess. 1. Bias - The coalescing of AI and big data opens significant possibilities for the synthesis and analysis of that data, but it also stands to compound problems that presently exist in that process. These include unintended racism, sexism and discrimination in the outcomes of data analysis. 2. Safety - AI is being touted as a solution to a number of social problems. However the safety risks present in autonomous vehicles include the risk of accidents that may not otherwise have occurred; accidents created by even minor software or hardware errors, flawed or deficient programming of software, or unethical decision-making in the face of a high-risk, multi-risk scenario. These risks are most acute with personal care robots. Trust and confidence in AI assisted robots may be hard-won in personal care situations given that they have traditionally involved humanto-human interaction. 3. Legal Decision-Making - AI has been applied in highly specific legal tasks such as sentencing and judicial interpretation in an effort to improve transparency and consistency in judicial decisions. However, these systems have been criticized as lacking capacity to exercise discretion and make situational value judgments. Concerns have been raised about mechanistic reliance upon these applications of AI and their capacity to influence and shape the behavior of people involved in the decision-making process. 4. Privacy - The leaps in advancement that are the promise of AI will sometimes turn on the quality and quantity of information available to it to inform AI learning. Public regulators will need to regulate to protect the privacy of individuals if large data sets are disclosed to tech companies with AI capabilities. 5. Unemployment - The socio-economic and socio-political impact of AI is a serious risk for public regulators. The deployment of AI in workplaces via algorithms, robotics or automation targeting increased speed, efficiency, or safety is expected to radically change the workforce. These concerns speak to a fundamental issue beyond the economics of increased productivity. The sheer scale of the disruptive impact on wages and employment is unlikely to be matched by increased productivity and may instead ‘exacerbate inequality rather than promote greater opportunity and shared prosperity...