Exploring Potential Flaws and Dangers Involving Machine Learning Technology

Missouri S&T’s Peer to Peer Volume 1 | Issue 2 Article 4 May 2017 Exploring Potential Flaws and Dangers Involving Machine Learning Technology David Nicholas Skoff Follow this and additional works at: https://scholarsmine.mst.edu/peer2peer Part of the Computer Sciences Commons Recommended Citation Skoff, David N.. 2017. "Exploring Potential Flaws and Dangers Involving Machine Learning Technology." Missouri S&T’s Peer to Peer 1, (2). https://scholarsmine.mst.edu/peer2peer/vol1/iss2/4 This Article - Journal is brought to you for free and open access by Scholars' Mine. It has been accepted for inclusion in Missouri S&T’s Peer to Peer by an authorized administrator of Scholars' Mine. This work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the permission of the copyright holder. For more information, please contact . Skoff: Dangers of Machine Learning Technology DANGERS OF MACHINE LEARNING TECHNOLOGY Skoff 1 David Nicholas Skoff Computer Science at Missouri University of Science and Technology EXPLORING POTENTIAL FLAWS AND DANGERS INVOLVING MACHINE LEARNING TECHNOLOGY Published by Scholars' Mine, 2017 1 Missouri S&T’s Peer to Peer, Vol. 1, Iss. 2 [2017], Art. 4 DANGERS OF MACHINE LEARNING TECHNOLOGY Skoff 2 Abstract This paper seeks to explore the ways in which machine learning and AI may influence the world in the future and the potential for the technology to be misused or exploited. In 1959 Arthur Samuel defined machine learning as “the field of study that gives computers the ability to learn without being explicitly programmed” (Munoz). This paper will also seek to find out if there is merit to the current worry that robots will take over some jobs based in cognitive abilities. In the past, a human was required to perform these jobs, but with the rise of more complex automation a person may not be necessary. Many of the sources cited throughout this paper focus on the innovation of machine learning and AI and how dangerous the over automation of the world could be. Machine learning and the resulting AI’s have their place in the world and more than likely they will do nothing but push the world towards a more fruitful future. Looking at potential risks of letting lines of code make important decisions is crucial given the consequences that negligence can have. There is a need to explore these topics because losing the human element in decision making can have some big implications if the AI is not programmed correctly. Machine learning has one of the greatest opportunities to impact the world. The need for caution however cannot be understated because of the potential dangers it may pose to jobs, security, and the overall stability of an ever changing world. https://scholarsmine.mst.edu/peer2peer/vol1/iss2/4 2 Skoff: Dangers of Machine Learning Technology DANGERS OF MACHINE LEARNING TECHNOLOGY Skoff 3 Exploring Potential Flaws and Dangers Involving Machine Learning Technology Humans are always looking to evolve and automate tasks. Programming has come a long way since the early programming languages of FORTRAN and the like. Programming is now a complex task which creates complex solutions to problems plaguing all aspects of humanity. One of the complex solutions is artificial intelligence or AI. Machine learning and AI have created the potential for complete automation at home and in the workplace. There are of course problems with removing a human element from complex tasks. The potential effect on the workplace cannot be understated. Complete automation may even lead to more pressing issues. While the possibility of rogue AI seems straight from a science fiction film, the dangers of full automation are extensive. This danger could come from someone intentionally creating malicious AI or from a simple and innocent error in algorithm construction. In the future, there may need to be certain restrictions and sanctions targeting algorithms that could be used to create powerful AI’s that could impact more than just the workplace. As the world nears complete automation in some sectors, security becomes paramount in ensuring safe execution of tasks. Machine learning can be a great tool for shaping the future, but its potential perils cannot be understated. Workplace Impact AI taking over the workplace removes the human element from decision making and introduces the potential for malicious attacks upon critical systems. Carl Frey and Michael Osborne explored the fact that jobs that usually require high cognitive ability are being replaced by an automated solution. They say, “Text and data mining has improved the quality of legal research as constant access to market information has improved the efficiency of managerial decisionmaking” (Frey & Osborne, 2017). This means that in the near future, tasks believed to require a human may become automated. Frey and Osborne specifically mentions such tasks as legal Published by Scholars' Mine, 2017 3 Missouri S&T’s Peer to Peer, Vol. 1, Iss. 2 [2017], Art. 4 DANGERS OF MACHINE LEARNING TECHNOLOGY Skoff 4 writing and truck driving may be taken over by computerization (Frey & Osborne 2017). Darrell West from The Center of Technology Innovation at Brookings says, “Telemarketers, title examiners, hand sewers, mathematical technicians, insurance underwriters, watch repairers, cargo agents, tax preparers, photographic process workers, new accounts clerks, library technicians, and data-entry specialists have a 99 percent chance of having their jobs computerized” (West, 2015). This does not necessarily mean that more complicated jobs such as those in the medical and legal fields can be computerized. In fact, West says that these jobs have a less than one percent chance of being replaced (West, 2015). If phased out by robots then the workforce potentially gains efficiency and accuracy but loses the human element. Another concerning factor is the potential breach of algorithms that dictate AI for critical systems. In these situations, a real person would be unaffected by such malicious attacks on critical systems. These types of attacks may become more probable as time goes on. Researchers from Stanford and Georgetown dissected the fact that making viruses has never been easier. They state, “To complicate matters, writing malicious programs has become easier: There are virus kits freely available on the Internet. Individuals who write viruses have become more sophisticated, often using mechanisms to change or obfuscate their code to produce so-called polymorphic viruses” (Kolter & Maloof, 2006). Surely, the security on critical systems which house essential AI would be strong. This however, has never stopped determined hackers from trying to crack through every firewall and security protocol. The computerization of certain jobs is coming and being prepared for such a future would be beneficial for the whole world. Security Concerns The potential dangers an (...truncated)