Artificial Intelligence - Challenges and Governance

Introduction

“AI is likely to be either the best or the worst thing that happened to humanity.”– Stephen Hawking

“Artificial Intelligence (AI), the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. (“artificial intelligence | Definition, Examples, Types, Applications, Companies, & Facts | Britannica,” n.d.). In the simplest terms, AI refers to systems or machines that mimic human intelligence to perform tasks and can iteratively improve themselves based on the information they collect (“What is Artificial Intelligence (AI)? | Oracle,” n.d.). AI helps a machine to develop generalized learning, reasoning, and problem-solving capabilities which makes machine Artificially Intelligent. AI enhances capability to adapt, reason, and provide solutions.

History

The history of AI started in late 19th and starting of 20th century with modern computers. Charles Babbage (Cambridge University Mathematician) and Augusta Ada Byron (Countess of Lovelace) invented first design in 1836 of a programmable machine. AI was early used in mid-1960’s to solve problems using mathematical formulas. AI was popularized at Dartmouth College in the United States in a conference that brought together researchers on a broad range of topics, from language simulation to machine learning in 1956. AI was started being used in business in 1980’s to develop software and hardware in computers and robots to build expert systems. Since 2001, researcher’s main challenge remains as large volume of data in databases.

Types

AI has two broad categories: Weak AI and Strong AI. Weak AI or Narrow AI focuses solely on one task which is machine trained and programmed for. For example, Alexa is trained to pick keywords to search and perform task. Strong AI is still a fiction robot that does not exist for now. These robots will be self-aware and eventually may even develop emotions.

?

?

Machine Learning

AI has many ideas, techniques, and viewpoints from different areas; however, it is considered to develop software and hardware to complete specific tasks. The difficulty in explaining this type of task by defining rules indicated that AI techniques needed the capability to extract patterns from data and to acquire their own knowledge is known as Machine Learning (“The strategic use of artificial intelligence in the digital era: Systematic literature review and future research directions | Elsevier Enhanced Reader,” n.d.). ?Machine learning is a technique to achieve AI and Deep Learning is in turn a subset of Machine Learning. (“Artificial Intelligence In 5 Minutes | What Is Artificial Intelligence? | AI Explained | Simplilearn - YouTube,” n.d.)

(“AI-vs-ML-vs-Deep-Learning.png (2048×861),” n.d.)??

Machine Learning is based on the inductive learning. With the help of observations, patterns are identified; and from patterns, hypothesis is derived which is used for predictions, and this hypothesis creates the theory. (Aggarwal, n.d.). There are several types of machine learning: supervised learning, unsupervised learning, and reinforcement learning.

Explainable Artificial Intelligence (XAI) is a set of processes and methods that allows human users to trust and comprehend the results and output created by machine learning algorithms. Explainable AI is used to describe an AI model’s expected impact, and potential biases. It helps characterize model fairness, accuracy, transparency, and outcomes in AI-powered decision making. Explainable AI is crucial for an organization in building trust and confidence while using AI models into production. AI explain-ability also helps an organization adopt a responsible approach to AI development (“Explainable AI | IBM,” n.d.).

1.?The strategic contexts for the use of AI technology in organisations including how the technology is used in digital transformation initiatives

In organizational contexts, AI can be considered as a technology that has been introduced as a means of emulating human performance with the potential to draw its own conclusions through learning, which can aid human cognition or even replace human in tasks that require cognition (Chakravorti et al., 2019). AI technology is enabling improvements in performance with respect to speed, scale, customization, flexibility, decision-making, and innovation.

AI is used in different ways by organisations as per their use. Some examples can be like AI is used in automation. With the help of AI volume and types of tasks to be performed can be increased. Software automates repetitive tasks done by humans. When machine learning is combined, bigger portions of jobs can be performed.

Another example of AI is self-driving cars. Vehicles use combination of image recognition, computer vision and deep learning to pilot the vehicle in a particular lane and avoiding unexpected obstructions. Self-driving cars are still in the innovation phase. Software is developed with the help of new learnings and database to overcome the obstructions and find better solutions. Tesla is a trending name and example of AI.

AI is used in the field of engineering to design and manufacture robots. Robots are programmed to complete tasks that are difficult for humans to perform or to perform consistently. Robots are used by automobile industry, NASA, and even in restaurants now to complete tasks with perfection.

AI is very helpful in healthcare. AI is helping to create better machines to diagnose humans with the help of machine learning. One of the best-known healthcare technologies is IBM Watson. It understands natural language and can respond to questions asked of it. The system mines patient data and other available data sources to form a hypothesis, which it then presents with a confidence scoring schema (“What is Artificial Intelligence (AI)? - AI Definition and How it Works,” n.d.). Machine learning is helping to develop software to scan and diagnose x-ray chart with a normal phone camera.

AI is currently used in education, business, finance, banking, transportation, law, manufacturing and in many other ways. AI and machine learning are widely used to develop software and hardware as per the tasks to be performed. AI is also playing an important role in cyber security to counter attacks and even predict them before.

Elon Musk & his team is developing a chip that could fit inside human brain. With the help of it, humans may control and make machines work as per their will by just thinking about it. How exciting it is, you can make your coffee, control electrical systems, and even plan your day by just thinking about it.

?

2.?The different challenges associated with the use of AI technology

Development and innovation in AI technology is a continuous process. This aims to solve challenges being faced after implementation. However, still some challenges are associated which are:

1.?????AI technologies need a human support as expert to problem domain for hypothesis creation and selecting features to solve. But humans may be unwilling to insert information that can lead to their job elimination. The cognitive AI restrict human beings to understand and explain its behaviour in certain cases. This makes decision making a major challenge. However, this needs to be understood that AI-human equation is important to align business needs and digital strategies to implement applications successfully.

2.?????Involvement of employees and customer are important to develop AI. But many changes in workforce and fear of job elimination can lead to have low confidence on AI’s recommendations. AI creates a competitive advantage for improving customer’s experience and applications are designed as per digital strategies. Companies should create digital strategies which involve their workforce engagement.??

3.?????AI is widely used in automation. As a part of digital business strategy, business should involve rules to harness automation which can lead have advantages. To avoid errors, automation should be used in creating competitive advantage but also under human supervision.?????

4.?????New products and services are designed with the help of AI. But still, AI cannot completely comply with human behaviour, emotions, and cognitive action taking capabilities currently. AI can be a curse if not controlled correctly.

5.?????AI era may have been started but it is still in progress. A large amount of data is being generated every second. This is a big challenge for AI. This is an obstacle in building generalized learning. The “black box” complexity of deep learning techniques creates the challenge of “explain-ability,” or showing which factors led to a decision or prediction, and how.

6.?????Countries have different rules and regulations which makes AI a restricted technology. Also, since AI is expensive technology, not every country can afford and benefit from this technology. A big question is, world is ready to change, but is every country ready to adopt and survive change?

7.?????Not every job can benefit from AI. Some tasks require human expertise to solve issues and take quick decision in case of any emergency. AI can only perform dedicated task as it is narrow. For example, working in offshore oil rigs and coal mines cannot be performed by AI.

8.?????AI performs as per the database value. Sometimes. AI may perform in a biased manner as per the data. For example, a company was using AI as screening process to hire candidates and process was rejecting candidates based on their gender which was not assigned. This error occurred due to machine learning and deep learning’s algorithm bias.

9.?????AI also has a challenge of securing the data. Valuable information is used which if miss-used can lead to huge loss to organisation and even individuals.

10.?With major companies such as Google, Facebook, and Apple facing charges regarding unethical use of user data generated, various countries such as India are using stringent IT rules to restrict the flow. Thus, these companies are facing the problem of using local data for developing applications which would result in bias. The data is a very important aspect of AI, and labelled data is used to train machines to learn and make predictions. Some companies are trying to innovate new methodologies that are focused on creating AI models and can give accurate results despite the scarcity of data. With biased information, the entire system could become flawed (“Top 7 Challenges in Artificial Intelligence in 2021 | upGrad blog,” n.d.).

Along with the above challenges, AIRS has identified risks of AI as per below figure:

(“AIRS-AI-Risk-Categories_Figure-1_Dec2020.png (782×549),” n.d.)

Artificial Intelligence/Machine Learning Risk & Security Working Group (AIRS) is an informal group of practitioners and academics from various backgrounds, including technology risk, model risk management, information security, legal, architects, privacy, and others working for academic institutions, technology organizations, and financial organisations. The AIRS working group is based in New York and was initiated in early 2019. It has grown to nearly 40 members from dozens of institutions and continues to grow (“Artificial Intelligence Risk & Governance - Artificial Intelligence for Business,” n.d.).

?

3.??What the governance of AI technology entails and the different disciplinary perspectives on the governance requirements

AI technology is under the governance of General Data Protection Regulation (GDPR). This regulation lays down rules related to the protection of persons regarding the processing of personal data and rules related to the free movement of personal data. It also protects fundamental rights and freedoms of persons and in particular the right to the protection of personal data. The free movement of personal data within the Union shall be neither restricted nor prohibited for reasons connected with the protection of natural persons regarding the processing of personal data (“General Data Protection Regulation (GDPR) – Official Legal Text,” n.d.). The final interpretation of the GDPR is exclusively within the jurisdiction of the European Court of Justice. However, the opinions of the supervisory authorities are of considerable practical relevance due to their supervision through their investigative and corrective powers.

The highlighted key issues by GDPR are following:

Consent must be specific, given, informed and unambiguous. It must be given on voluntary basis. The basic requirements for the effectiveness of a valid legal consent are defined in Article 7.

Data Protection Officer is a concept established by GDPR in Europe. Decisive for the legal obligation to appoint a DPO is not the size of the company but the core processing activities which are essential to achieve the company’s goals. If these core activities consist of processing sensitive personal data on a large scale which is particularly far reaching for the rights of the data subjects, the company must have a DPO.

Email Marketing and newsletter mailing processing is only allowed by GDPR if either the data subject has consented, or there is another legal basis. Recital 47 of GDPR states that the law also applies to the processing of personal data for direct marketing in a legitimate interest of the controller. According to Art. 21(2), (3) GDPR the data subject always has the right to object the processing of personal data for direct marketing purposes.

Encryption of personal data can reduce the probability of the data breach. Risk management is playing an important role in cyber security. The Regulation also recognizes these risks when processing personal data and places the responsibility on the controller and the processor. In Art. 32(1) of GDPR implementation of appropriate technical and organisational measures to secure personal data are stated.

Fines/Penalties must be effective, proportionate, and dissuasive for each individual case. For especially severe violations as per listed in Art. 83(5) of GDPR, the fine framework can be up to 20 million euros, or in the case of an undertaking, up to 4 % of their total global turnover of the preceding fiscal year, whichever is higher. But even the catalogue of less severe violations in Art. 83(4) of GDPR sets fines of up to 10 million euros or in the case of an undertaking, up to 2% of its entire global turnover of the preceding fiscal year, whichever is higher.

Personal data are any information which are related to an identified or identifiable person. The data subjects are identifiable if they can be directly or indirectly identified, especially by reference to an identifier such as a name, an identification number, location data, an online identifier or one of several special characteristics, which expresses the physical, physiological, genetic, mental, commercial, cultural, or social identity of these natural persons. The same also applies to IP addresses.

Privacy by Design and Privacy by Default have been frequently discussed topics related to data protection. The first thoughts were expressed in the 1970’s and were incorporated in the 1990’s into the data protection directive.

Privacy Impact Assessment is the instrument introduced in Art. 35 of the GDPR. This refers to the obligation of the controller to conduct an impact assessment and to document it before starting the intended data processing. One can bundle the assessment for several processing procedures.

Processing regulations is connected to activities of an establishment within the EU. One must differentiate between processing and joint control as per Art. 26 of the GDPR, where both parties jointly define the purposes and means for the data processing and are thus also jointly responsible for these.

Records of processing activities must include significant information about data processing, including data categories, the group of data subjects, the purpose of the processing and the data recipients. This must be completely made available to authorities when requested. As per Art. 30 of the GDPR, written documentation and overview of procedures by which personal data are processed should be available.

The right of access plays a central role in the GDPR. On the one hand, because only the right of access allows the data subject to exercise further rights. On the other hand, because an omitted or incomplete disclosure is subject to fines.

The right to be forgotten derived from the case Google Spain SL, Google Inc v Agencia Espa?ola de Protección de Datos, Mario Costeja González (2014). For the first time, the right to be forgotten is codified and to be found in the GDPR in addition to the right to erasure.

The right to be informed is about the collection and use of the personal data, which leads to a variety of information obligations by the controller.

Third countries are in the view of the international trade and cooperation. It is essential these days to be able to also transmit data to third countries.

4. Concrete examples of governance models, frameworks and solutions that have been applied to address the governance of AI technology

Many organisations hold both, public and private information about individuals. This has led to the data protection laws. In the EU, this is governed by GDPR. Specific laws also deal with such matters as criminal investigations. There are other additional laws in each EU member state. In Ireland, the Data Protection Acts and other regulations include these laws.

These data protection laws mean that your personal data should generally only be stored where there is a lawful basis, such as your consent or a legal obligation. There are several rights under data protection laws, like the right to access the personal data held and the right to have it corrected or erased.

Article 5 of GDPR sets out key principles which lie at the heart of the general data protection regime. The following is a brief overview of the Principles of Data Protection found in article 5 GDPR:

Lawfulness, fairness, and transparency: Any processing of personal data should be lawful and fair. It should be transparent to individuals that personal data concerning them are collected, used, consulted, or otherwise processed and to what extent the personal data will be processed. The principle of transparency requires that any information and communication related to the processing of personal data should be easily accessible and easy to understand. It should be clear and plain language to be used.

Purpose Limitation: Personal data should only be collected for specified, explicit, and legitimate purposes and not further processed in a manner that is incompatible with those purposes. However, further processing for archiving purposes in the public interest, scientific, or historical research purposes or statistical purposes, in accordance with Article 89(1) GDPR, is not considered to be incompatible with the initial purposes.

Data Minimisation: Processing of personal data must be adequate, relevant, and limited to what is necessary in relation to the purposes for which they are processed. Personal data should be processed only if the purpose of the processing could not reasonably be fulfilled by other means. This requires ensuring that the period for which the personal data are stored is limited to a strict minimum.

Accuracy: Controllers must ensure that personal data are accurate and kept up to date where necessary; taking every reasonable step to ensure that personal data that are inaccurate, having regard to the purposes for which they are processed, are erased, or rectified without delay. Controllers should accurately record information they collect or receive along with the source of that information.

Storage Limitation: Personal data should only be kept in a form which permits identification of data subjects for as long as is necessary for the purposes for which the personal data are processed. To ensure that the personal data are not kept longer than necessary, time limits should be established by the controller for erasure or for a periodic review.

Integrity and Confidentiality: Personal data should be processed in a manner that ensures appropriate security and confidentiality of the personal data, including protection against unauthorised or unlawful access to or use of personal data and the equipment used for the processing and against accidental loss, destruction, or damage, using appropriate technical or organisational measures.

Accountability: Finally, the controller is responsible for, and should be able to demonstrate, the compliance with all the above-named Principles of Data Protection. Controllers must take responsibility of their processing of personal data and how they comply with the GDPR and be able to demonstrate their compliance, through appropriate records and measures, to the DPC. (“Principles of Data Protection | Data Protection Commissioner,” n.d.)

?

5. Challenges in adopting the governance solutions

AI is developing, making it important to have flexible rules and regulations which can be changed or updated after the implementation observations. With respect to EU, below are current challenges:

-?????????The convergence between data-driven AI and the global data-processing infrastructure raises risks for citizens as the power of AI can be harnessed for surveillance and manipulation. The combined powers of AI and big data can restrict user’s options or influence their opinions and manipulate them into making choices that do not serve their best interests.

-?????????Both legal regulation and social empowerment are needed to ensure that AI is developed and deployed in ways that preserve and enhance individual interests and the social good.

-?????????Legal regulation must focus on principles including individual rights and social goals, as well as on existing regulatory frameworks.

-?????????Multiple legally relevant interests are affected by AI, such as data protection, fair algorithmic treatment, transparency, explicability, and protection from undue influence.

-?????????The extent to which algorithmic price discrimination is acceptable in online markets should be clarified.

-?????????Inacceptable practices in targeted advertising and nudging directed to consumers should be defined and addressed.

-?????????Discrimination in ads delivery should be countered.

-?????????Citizens and consumers should be provided with effective ways to turn off personalisation.

-?????????The development and deployment of AI tools that empower citizens and consumers and civil society organisations should be incentivised.

(“IPOL | Policy Department for Economic, Scientific and Quality of Life Policies,” n.d.)

6. Your critique of the extant literature on the governance of the selected digital technology and suggested areas for future studies

Artificial Technology is developing drastically. Some of my critique with respect to AI are as per below:

-?????????In times of rapid change of technology in society, the regulatory balancing is difficult. So, the laws should be extremely dynamic and adaptive.

-?????????Ethical guidelines may be formalized, focusing on the importance of trust.

-?????????Transparency, traceability, and human oversight should be clearly understood under current legislation.

-??????????

Suggested areas for application of AI future studies:

-?????????AI is being used to improve healthcare industries. To detect fatal diseases like cancer, AI is being developed. Fitness band is an example which collects information of body activities. In future, AI may be able to prescribe medication as per the disease.

-?????????Robots are being used for education purpose.

-?????????AI algorithm can be used to manage investments funds.

-?????????AI is being used to make weapons which can be operated without any human intervention for the safety of nation. Robot military may be the future.

-?????????Cybersecurity is being strengthened with the help of AI. Cyber attacks can be intercepted.

-?????????Self-driving cars will be more advanced in safety and security to avoid obstacles.

It is difficult for an organisation to understand the AI decision making algorithms. Explainable AI help humans to understand this machine learning algorithms, deep learning, and neural networks. Explainable AI is one of key requirements for implementing AI responsibly. It can troubleshoot and improve business performance. (“Explainable AI | IBM,” n.d.)?

References

Aggarwal, C.C., n.d. Artificial Intelligence A Textbook (accessed 12.20.21)

AIRS-AI-Risk-Categories_Figure-1_Dec2020.png (782×549) [WWW Document], n.d. URL https://ai.wharton.upenn.edu/wp-content/uploads/2020/12/AIRS-AI-Risk-Categories_Figure-1_Dec2020.png (accessed 12.21.21).

AI-vs-ML-vs-Deep-Learning.png (2048×861) [WWW Document], n.d. URL https://www.edureka.co/blog/wp-content/uploads/2018/03/AI-vs-ML-vs-Deep-Learning.png (accessed 12.24.21).

artificial intelligence | Definition, Examples, Types, Applications, Companies, & Facts | Britannica [WWW Document], n.d. URL https://www.britannica.com/technology/artificial-intelligence (accessed 12.24.21).

Artificial Intelligence In 5 Minutes | What Is Artificial Intelligence? | AI Explained | Simplilearn - YouTube [WWW Document], n.d. URL https://www.youtube.com/watch?v=ad79nYk2keg (accessed 12.24.21).

Artificial Intelligence Risk & Governance - Artificial Intelligence for Business [WWW Document], n.d. URL https://ai.wharton.upenn.edu/artificial-intelligence-risk-governance/ (accessed 12.22.21).

Chakravorti et al., 2019 |Which countries ate leading the data economy?, B. Chakravorti, A. Bhalla and R.S.Chaturvedi, Harvard Business Review (2019) (accessed 12.26.21).

Explainable AI | IBM [WWW Document], n.d. URL https://www.ibm.com/watson/explainable-ai (accessed 12.24.21).

Fujita, H., Selamat, A., Chun, J., Lin, W., Ali, M., n.d. Advances and Trends in Artificial Intelligence Artificial Intelligence Practices. Proceedings. (accessed 12.29.21)

General Data Protection Regulation (GDPR) – Official Legal Text [WWW Document], n.d. URL https://gdpr-info.eu/ (accessed 12.31.21).

IPOL | Policy Department for Economic, Scientific and Quality of Life Policies, n.d. (accessed 12.29.21)

Larsson, S., 2022. On the Governance of Artificial Intelligence through Ethics Guidelines ?(accessed 12.30.21).

Principles of Data Protection | Data Protection Commissioner [WWW Document], n.d. URL https://www.dataprotection.ie/en/organisations/data-protection-basics/principles-data-protection (accessed 12.31.21).

ProQuest Ebook Central - Reader [WWW Document], n.d. URL https://ebookcentral.proquest.com/lib/nuim/reader.action?docID=5877761 (accessed 12.31.21).

The strategic use of artificial intelligence in the digital era: Systematic literature review and future research directions | Elsevier Enhanced Reader [WWW Document], n.d. URL https://reader.elsevier.com/reader/sd/pii/S0268401219317906?token=F74AD8CE6513DF625F3430285D8B2806B3582228C0CB27527E732335B920FB66FE7DB6A192A7D5E17FF79AC253BB2073&originRegion=eu-west-1&originCreation=20211229132759 (accessed 12.29.21).

Top 7 Challenges in Artificial Intelligence in 2021 | upGrad blog [WWW Document], n.d. URL https://www.upgrad.com/blog/top-challenges-in-artificial-intelligence/ (accessed 12.31.21).

What is Artificial Intelligence (AI)? | Oracle [WWW Document], n.d. URL https://www.oracle.com/artificial-intelligence/what-is-ai/ (accessed 12.24.21).

What is Artificial Intelligence (AI)? - AI Definition and How it Works [WWW Document], n.d. URL https://www.techtarget.com/searchenterpriseai/definition/AI-Artificial-Intelligence (accessed 12.31.21).