Author(s),: SupriyaGhosh Artificial Intelligence AI ethics : The framework for building Ethical AI Artificial Intelligence is considered a revolution by many people in the technology world. This brings up its own challenges and problems. This leads to the question: Will AI be our driving force in the future? Oder Will AI be able to fight for our safety? Oder will AI manage human healthcare? Oder will AI take over agriculture? There are many other questions. These questions can be a constant source of confusion and encourage people to think beyond the practical capabilities and ethical implications of AI systems. This is the main challenge: to create a system that serves humanity’s common good and has ethical considerations. It is important to first understand what you want your system to achieve, and then design it to do so. Three major areas that are of greatest concern in the advancement of AI include: 1. Privacy and surveillance, 2. Bias and Discrimination The Ethical Avatar addresses all of these issues by providing correct judgment and accountability. What exactly is ethical AI? Ethical AI is the adoption of AI that is transparent, accountable, and responsible. This means AI should be consistent with all laws, regulations and norms. An ethical AI must be capable of protecting against bias, discrimination and incorrect judgement, as well as respecting privacy at all scales. Source — https://revistaidees.cat/en/thinking-about-ethics-in-the-ethics-of-ai/ Let me depict a few Use cases in AI which will make you understand the underlying serious ethical concerns. Healthcare Healthcare: Ethical AI Examples. Although healthcare is a field where AI was first adopted, it has many problems such as issues with patient engagement, care delivery and privacy. If not rectified, bias in the society can be reflected in historic health data. AI systems may make biased decisions if this isn’t corrected. Who gets healthcare management services on priority based on the racial composition of the study population and the gender split? This is discriminatory and a source of malpractice. What ethical AI considerations can be used to address the Healthcare problem? FDA has added regulatory requirements to decrease bias. It is strict on firms’ obligation to report regularly on real-world performance and monitor them. Companies are constantly being directed to ensure that the decisions they make regarding customers, partners, data team composition, and data collection contributes towards minimising bias. Google Health is one example of a company that is working to detect breast cancer before it occurs. It is not only religiously improving its performance, as well as reducing costs and validating algorithm performance in various clinical settings, but also investing large amounts of money in ensuring the algorithm is fair across racial backgrounds. An example of ethical AI: Autonomous Vehicle. Imagine an autonomous vehicle driving at maximum speed toward an elderly person in the -80’s. The child is 10. One can save one’s life by allowing the algorithm to deviate a bit. The car’s algorithm will make the final decision, and not the driver. Which would you choose, an old man or a child? This is a moral dilemma that requires Ethical AI. There’s no one right way to solve it. What ethical AI can do to solve this Autonomous Vehicle problem? The algorithm should be highly efficient to take a controlled decision to both avoid hitting the streetwalker/pedestrian as well as cut the risk for the people inside the car. Therefore, control and liability are essential. Ethical AI businesses have many roles for existing bodies such as the National Highway Transportation Safety Association. This organization oversees vehicles safety. Ethical AI example: Court of Law AI is used in the judicial system to assess cases and administer justice more efficiently than a judge. Because machines are faster than human judges and have large data storage capacity, they can better evaluate and weight relevant factors. It is important to ask whether the decisions are free from bias or subjectivity. These ethical issues are not unique: Transparency of AI algorithms and tools. 2. AI does not work in isolation. Decisions based upon human-fed data can be biased, inaccurate, or discriminatory. What ethical AI can solve this Court of Law problem? To make legislators act, there must be an “emotion of urgency”. The Algorithmic Accountability Act was introduced by U.S. legislators. If fully enforced, it can resolve known issues such as algorithmic bias and privacy and security concerns. Although ethical AI plays a significant role, context is also important at every level including the tactical. Ethical AI is focused on reliability and safety in industrial manufacturing applications, while fairness and in-discrimination are priorities in public and consumer services. Source — NEW YORK (NY) NOVEMBER View of a poster on ethical AI at the 2018. New York Times Dealbook, November 1st in New York City). Photo by Michael Cohen/Getty images for The New York Times. Another concern with current artificial intelligence technology is the possibility of adversarial instances. Aversarial instances manipulate AI systems’ behavior by small, invisible changes to input data. Because AI algorithms operate in a way that is fundamentally different to the human brain, this happens. These adversarial examples may occur by chance or can be intentionally injected in to harmful adversarial attacks on critical AI systems that can pose a threat to safety. For ethical, trustworthy and reliable artificial intelligence to be developed, the European Commission has set out several requirements. The European Commission recommends that AI systems be capable of switching from machine learning to rules-based systems, or asking for human intervention. The Human-in the-loop system is required for human intervention. To understand Human-in the-loop systems in detail, you can refer to my previous write-up. Incorporating Human-in the-Loop (HITL), in machine learning applications is not an option. The end-user must also know the reliability and confidence levels of the AI systems they use. Guidelines recommend that AI developers ensure their data sets include all information to prevent biases against specific groups. This is a crucial step to take now in order to build ethical AI. The consequences down the line can prove far more devastating than most people realise. It is important to continue working for ethical AI. Source — https://devopedia.org/ethical-ai Final Thoughts The adoption of ethical AI principles is essential for the healthy development of all AI-driven technologies and self-regulation by the industry will be much more effective than any other effort. This AI-based decision must be explained and continually monitored. All AI systems are powered by data. Therefore, it is crucial that the use and collection of customer data be closely monitored, particularly in commercial settings. These types of safeguards are likely to become more common as more people and businesses realize the value of ethical AI. Thank you for reading! Follow me on LinkedIn and Medium: Supriya Ghosh. Twitter: @isupriyaghosh. Why was Ethical Artificial Intelligence Avatar created?
