PROS & CONS

WHAT ARE SOME OF THE BENEFITS AND DRAWBACKS OF AI?

Many companies often only think of the benefits of AI, but "60 percent of those companies fear liability issues and 63 percent say they lack the skills to harness AI's potential", according a study by IBM's Institute for Business Value (Rossi, 2018). Does the good outweigh the bad? Some of the pros and cons of AI are listed below.

PROS OF AI

  • Allows for an enhanced system for collecting, processing, and categorizing information (Voinea, Vica, Mihailov, & Savulescu, 2020).
  • Virtual assistants that use AI assist users in device and internet navigation (Voinea, Vica, Mihailov, & Savulescu, 2020).
    • Open-sourced code
      • Allows anyone to fix and improve the application (Voinea, Vica, Mihailov, & Savulescu, 2020).
      • Autonomy for user to know how the software works and enhance it to meet their needs (Voinea, Vica, Mihailov, & Savulescu, 2020).
      • Creates an environment for users to share knowledge with others and make additions to the code to support their own needs (Voinea, Vica, Mihailov, & Savulescu, 2020).
  • Provides user independence and autonomy (Voinea, Vica, Mihailov, & Savulescu, 2020).
  • Cyber security
    • Recognition of system weakness missed by humans with the ability to fix before an attack happens (Taddeo, 2019).
    • Forces attackers to work harder and smarter to attack the system (Taddeo, 2019).
    • Easier to collect data from opposing system and use it later for an attack (Sharikov, 2018).
    • Use of swarms, or new botnets, to infiltrate opposing system with little intervention (Sharikov, 2018).
  • Software that can test and fix itself faster and better than humans can (Taddeo, 2019).
  • Frees up people from mundane tasks (Taddeo, 2019).
  • Quick response time (Taddeo, 2019).
  • Machine Learning
    • Algorithm learns over time and adjusts the rules of the system to become streamlined and more efficient (Bohyun, 2019).
  • Help to lessen or completely get ride of an pressing issue (Floridi, Cowls, King, & Taddeo, 2020).

CONS OF AI

  • Lack of transparency in some AI algorithms make it difficult to determine the accuracy, determine the danger, and be impartial in social applications (Osoba & Welser IV, 2017).
  • Cyber security
    • Difficulty in seeing who is responsible for cyber attacks (Demchak, 2019).
    • Attacks can be used multiple times (Demchak, 2019).
    • No knowledge that an attack has happened (Demchak, 2019).
    • Recognition of system weakness missed by humans and then attack system (Taddeo, 2019).
    • Machine learning and deep learning allows for quicker and more forceful attacks (Taddeo, 2019).
    • Only leaning on AI testing could lead to human cyber security experts becoming less skilled and negligent (Taddeo, 2019).
  • Every user interaction is collected and tracked (Taddeo, 2019).
  • Treacherous when integrated with "cloud computing, big data and Internet of Things (IoT)" (Sharikov, 2018).
    • Coupled with big data, cracks in opposing systems can be found and automatic determination of the best time for attack is known (Sharikov, 2018).
  • Machine Learning
    • Lacks sensory aspect, human reasoning and logic, and interaction (Williams, 2019).
    • May learn incorrectly and as a result deliver unacceptable or inappropriate response to user (Osoba & Welser IV, 2017).
  • Removes the ability to choose the persona that we want to represent ourselves. We are only known by our accumulation of data within the system (Williams, 2019).
  • Lacks consent or the ability for the user to decline help or access (Floridi, Cowls, King, & Taddeo, 2020).
  • Data may be predisposed to a certain area or group (Floridi, Cowls, King, & Taddeo, 2020).
  • May have an understanding different than intended (Floridi, Cowls, King, & Taddeo, 2020).
  • Cannot define all interpretations and perceptions (Floridi, Cowls, King, & Taddeo, 2020).
  • Restriction of learning passed from one function to a similar function (De Spiegeleire, Maas, & Sweijs, 2017).
  • Not as good at language translation as humans (De Spiegeleire, Maas, & Sweijs, 2017).
  • Data can be manipulated (Floridi, Cowls, King, & Taddeo, 2020).