Artificial Intelligence and Legal Challenges

Artificial Intelligence and Legal Challenges

Md. Rayhanul Islam^*

Abstract: Artificial Intelligence (AI) is developing rapidly. Its unique ability to learn and solve unique problems without the help of humans put it into a new place. Therefore, AI is the only machine that can challenge the human civilization or can give humans their utopia, but before reaching there, humans must understand the working method of AI to foresee the challenges as well as to prepare the system to address those challenges.

Keywords: Artificial Intelligence, AI Regulation, Big Data, Legal personality, IP, rights.

1.  Introduction

Ever since humans developed the ability to understand they wonder whether there are any other species like them, who can pose challenges to the humanity with the power of intelligence; that specific power that is thought to be inherently developed and nurtured only by the Homosapiens. The curious man then searched for their intelligent counterpart in the quest to place them in a better position, but so far all those efforts remain fruitless. Though the technological advancement especially in the arena of computer and data science, human has created their own counterpart. Now machines with Big Data can outperform their creators in various specific tasks.¹ It is also predicted that within a few years machines will transcend the human ability in most of the sectors.² This unique combination of Big Data and computer with the help of appropriate hardware along with the development of software that follows the working method of a human brain where as human brain acts naturally the man-made combination of software, data and hardware work artificially. Thus the name Artificial Intelligence (AI) is coined by the creators.

Such machines (AI) not use only the working method but also to imitate the intelligent behaviour of human, which so far only existed in the human domain. The plan to create this type of Intelligent system was nourished by the visionaries decades ago, but they failed to succeed due to various challenges. Now with developed

1-Rayhanul Islam is an apprentice lawyer at Bangladesh Supreme Court and also the Editor in Chief at Law Help BD. He is an alumnus of 18^th HRSS and could be reached at: rayhanulislam029@gmail.com.

2-UKIT AI Roman Steinberg, '6 Areas Where Artificial Neural Networks Outperform Humans' (VentureBeat, 2017) <https://venturebeat.com/2017/12/08/6-areas-where-artificial-neural- networks-outperform-humans/> accessed 27 August 2018

3-Katja Grace and others, 'Viewpoint: When will AI Exceed Human Performance? Evidence from AI Experts' (2018) 62 Journal of Artificial Intelligence Research.

257

ELCOP Yearbook of Human Rights 2018 Artificial Intelligence

technology and huge data things are changing so fast that scientists are facing a hard time foreseeing the potential pros and cons of AI, however, with the currently available data and information some rational and reasonable prediction can easily.³ As the practice of AI technology is not well tested out in the wild the concerns are quite reasonable; there are concerns that it could change the course of human history because of its unique ability to influence every aspect of human life, including the power to alter the emotional state of human being! ⁴ To avoid such undesirable sudden challenges legal mind should be there to foresee the situation and impose proper regulations so that things remain peacefully, to filter out the evil to ensure appropriate use of AI for human prosperity.

At this point, the development of AI can not only be a subject matter of Computer Science or Technology, as it will affect the entire natural and human-made system. AI will pose a dominant position in every sector from hard labour to creation of music, in the research of sea life to the exploration of extraterrestrial life form. However, as the true extent of AI potentials are yet to enlist, all aspects of academia should embrace it to understand the present and future application of AI with a subtle view of philosophy as well otherwise the diverse.

2.  Understanding of Artificial Intelligence

The idea of Artificial Intelligence has changed over time. It has been labelled with various generation waves which often leads new people to nowhere whereas it is indeed essential to understand the basics of AI otherwise true concerns of would not be released properly. The fundamental difference between AI with other similar technologies must be well understood otherwise the unique power of AI can be easily ignored.

Despite having much hope, proposal, and hypothesis, the development of machine has observed during the Second World War. Among them one of the pioneers was Alan Turing, his code-breaking machine surprised people that how a computer can perform a complex task and provide accurate output. However, it must be mentioned that was a machine that used pre-written programmes and a quick calculation, but the importance of this machine was that Mr. Turing showed people sometimes people can be less efficient, on the other hand, he proposed in his theory of computation, “that theoretically and mathematically it is possible, by shuffling symbols as simple as "0" and

3-NefaBerberovic, 'Most Companies Still Dont Understand Big Data and AI — and Their Potential?' (Big Data Made Simple - One source. Many perspectives., 2018) <http://bigdata-madesimple.com/most-companies-still-dont-understand-big-data-and-ai-and-their-potential/> accessed 27 August 2018

4-Samuel Gibbs, 'Facebook Apologizesfor Psychological Experiments on Users' (the Guardian, 2014) <https://www.theguardian.com/technology/2014/jul/02/facebook-apologises-psychological-experiments-on- users> accessed 27 August 2018

"1", could simulate any conceivable act of mathematical deduction. This insight computers can simulate any process of formal reasoning, is known as the Church– Turing thesis.”⁵ Now, this deduction is a way of philosophical reasoning to find out a logical answer but finding an answer is not something that refers to in intelligence rather it is one of the many elements that may help a system to be truly intelligent, to find out a general form of intelligence among various proposals, Turing proposed his "Imitation Game" also known as the Turing Test⁶ where a judge would ask random multiple questions to two subjects without knowing who is behind the screen and one of them would be an AI; it would only pass if it could provide reasonable human-like response.

Seemingly this simple tasks of human are obviously considered science fiction in the 1950's because computers were and are very good at the following instruction by humans which is known as programming, a general computer (Programme) uses the instruction given to it, following these instructions like if.. else condition,⁷ mathematical calculation and various other mathematical processes, these programs find the data from the connected database by programmed method then re-calculate with other data or information as commanded and after all this, the computer then outputs the data / information as per the instruction. Generally, computers are good when given clear instruction whether conditional or liner, computers process such calculation in a lightning first way with pinpoint accuracy.

In contrast human do not follow the same method like a computer, rather, humans follow a very different approach. It can be easily understood by observing learning curve of a human child at first, he gathers information about his surroundings with some sort of understanding about things and objects, he then tries to imitate things like others, learns through error and reputation, he corrects himself by the judgments and rewards.⁸ The human brain does all these processes with its 100 billion neurons and 10× more interconnected glial cell. From the neurological viewpoint, it is right to say these cells communicate with each other on an event via neurotransmission, repetition of which creates some sort of pattern. These cells with time and by the effort develops

5-Robert I. Soare, 'Turing Oracle Machines, Online Computing, and Three Displacements in Computability Theory' (2009) 160 Annals of Pure and Applied Logic <http://www.people.cs. uchicago.edu/soare/History/turing.pdf>

6-Luciano Floridi, Mariarosaria Taddeo and Matteo Turilli, 'Turing’S Imitation Game: Still an Impossible Challenge for All Machines and Some Judges––An Evaluation of The 2008 Loebner Contest' (2008) 19 Minds and Machines

7-Peter Van Roy and SeifHafidi, Concepts, Techniques and Models of Computer Programming(PHI Learning Private Ltd 2009)

8-Geraldine French, Children’s Early Learning and Development: A Research Paper (National Council for Curriculum and Assessment 2007) <https://www.curriculumonline.ie/getmedia/ 61796c448/ ECSEC05_Exec2_Eng. Pdf> accessed 27 August 2018.

some cognitive ability, thus these patterns are observed to be evident on a similar condition or a similar event, but it does not mean the pattern is permanent instead of changing with time and knowledge.⁹ Humans learn every day, and their decision can vary over time on the same type of incidence but it must be noted that the new decision also get influenced by the previous pattern, the pattern always remains as part of new evolved pattern. As these patterns are individually formed, individual bias is very natural here. That is why a selected individual prefers to response in a certain way in a given situation, for the same reason the decision of and individual changes over time with new knowledge or experience. Considering these learning method it can be concluded the human system depends on the data it gets, thus can be called a data driven system.

Now, these are some unique ability of human comes with its unique advantage and disadvantages as well. For the features of the human brain, every human is different from another and that is their identity, that is where individuals get the power of innovation and creativity. As the learning materials, and exposer of each human being are different, every individual holds some unique understanding which leads them to make some decision different than others. Therefore, humans are not like other species, not only we adopt and repeat, we think and create too. On the other hand, these abilities like thinking and creating has enable humans to recognize an object or things (could be sound or shadow as well) though the same object or sound is never seen or heard before, that creative part of human comes to play and recognize that new object or things within split second. However, humans are very poor in the matter of calculation, following logical instructions compared to a computer. To put that in an understandable perspective; a human brain is performing around 200 trillion operations per second to do simple tasks like hearing, viewing, talking etc. which is more powerful than 1000 supercomputers but in order to perform a simple multiplication a four-bit microprocessor is better than human.¹⁰As mentioned previously, the unique quality of human can be very vulnerable too, as humans depend on their sense, observation and knowledge, changing those variables the mid of an individuals can be influenced, similarly by knowing most of the information and the lifeline of an individual renders the subject to some calculated prediction and when a person is predictable he is very easy to hack, but to do that a whole life data and proper influencing method must be deployed.

9. Sylvie Belleville and others, 'The Pattern and Loci of Training-Induced Brain Changes in Healthy Older Adults are Predicted by The Nature of The Intervention' (2014) 9 PLoS ONE.

10. Reddy R, 'Foundations and Grand Challenges of Artificial Intelligence' [1998] Carnegie Mellon University <http://www.rr.cs.cmu.edu> accessed 27 June 2018

Coming back to the Turing Dream, his left dream was nurtured by many scientists but ordinary conditional and calculative programming method was destined to fail due to their design and structure, even where few programmes worked up to some point which were very narrow and required massive programming, those were also power consuming for these reasons those systems could never use on a large scale project, significant others lacking were, those could not learn by them self,¹¹ could not make new decisions on unique challenges. Whereas, the human cognitive system is quite the opposite. To act like a human brain, scientists tried to design a programming method that followed the steps of a human brain so that it can work more like a human.

To act like a human, the system must have some power of autonomy that helps the system to help itself to make some unique decision or to change a certain part of the code to make a better decision which was not implanted by humans rather the system implemented for its ease. That new type of programming concept had come from the work of John McCarthy; he introduced LIPS Programming in 1958,¹² he also coined the term "Artificial Intelligence" in 1956 the Dartmouth College summer AI conference arranged by him which was first of its kind thus known as the Father of Artificial Intelligence.¹³ The next remarkable development of human- like approach came from Artificial Neural networking (ANN) which mimics the biological human brain process of learning.¹⁴ The multiple layer programming acts like human brain cells; each neuron performs its work simultaneously sends and receives signals with other units which leap forwarded the AI technology. But still, there were many problems, the first problem was capacity of the Hardware at that time, therefore, the scientists had the plan ready they could not implement because of insufficient hardware performance. For this instant data, collection and processing were limited then, the connectivity and data transmission was low and the latency was very high so scientist could not be able to replicate the function of the human brain which consists of 100 billion neurons and interconnection among them. On the other hand, a human brain is a data-driven system so do the AI, but at that time collection vast data (known as Big Data)¹⁵ of million subjects with hundreds of data point and their constant update was impossible.

11. Feng-hsiung Hsu, IBM’S DeepbluechessGrandmasterchips (IBM TJ Watson Research Cente 2018) <http://www.csis.pace.edu/ctappert/dps/pdf/ai-chess-deep.pdf> accessed 27 August 2018
12. Mitchell Wand, 'What is Lisp' (1984) 91 The American Mathematical Monthly.

13. Jordan Novet, 'Everyone Keeps Talking About A.I.—Here's What it really is and Why it's So Hot Now' (CNBC, 2017) <https://www.cnbc.com/2017/06/17/what-is-artificial-intelligence.html> accessed 27 August 2018

14. Marcel van Gerven and Sander Bohte, 'Editorial: Artificial Neural Networks as Models of Neural Information Processing' (2017) 11 Frontiers in Computational Neuroscience.

15. Steve Lohr, 'The Origins of 'Big Data': An Etymological Detective Story' (Bits Blog, 2013) <https://bits.blogs.nytimes.com/2013/02/01/the-origins-of-big-data-an-etymological-detective-story/> accessed 27 August 2018

First forwarding to the year 2010, the last problem of Big Data has solved due to the digital identification, constant surveillance, exposure to the internet, voluntary disposing of data and the collective collaboration of those data.¹⁶ Connectivity problem was also solved by this time, due to fourth and fifth generation networking, optical fiber technology, high speed and cloud server technology, but still, there were some problems to process such huge data with reasonable time and using a cost- effective method. Due to these advancements, almost all tech giants and first world government heavily invested in AI research and development. Their investment showed the potentiality of great benefit for them; they were quite successful in limited autonomy, understanding personal choice of individual, voice and image recognition etc.¹⁷ which was unthinkable decades ago. AI is also used to create more accurate data of a person,¹⁸ to make better suggestion and marketing¹⁹ as well as to carefully extract very specific data from an individual.

Looking forward to the present context of AI in 2018, AI can perform some remarkable tasks. Currently, these technologies outperformed human being in a most complicated board game of the earth named GO, similarly outperforming human beings in various specific tasks around 1.5% accuracy being even better.²⁰Now AI can recognize a voice, translate in real time, drive autonomously, design creatively, can talk humanly²¹, edit videos and images just with simple voice command by humans, even can land a giant autonomous fighter jet on a moving aircraft carrier.²² Now AI can predict human behavior and vulnerability better than a best friend. These developments became possible with the new hardware and connectivity. Special hardware like “Intel’s Movidius Neural Computer stick which enables rapid

16. ibid

17. Robert Hof, 'Is Artificial Intelligence Finally Coming into its Own?' (MIT Technology Review, 2013) <https://www.technologyreview.com/s/513696/deep-learning/> accessed 27 August 2018
18. Chelsea Gohd, 'How The CIA is Using Artificial Intelligence to Collect Social Media Data' (Futurism,

2017) <https://futurism.com/how-the-cia-is-using-artificial-intelligence-to- collect-social-media-data/> accessed 27 August 2018

19. Kyle Russell, 'Netflix is 'Training' Its Recommendation System by Using Amazon's Cloud to Mimic the Human Brain' (Business Insider, 2014) <https://www.businessinsider.flix-using-ai-to-suggest-better-films-2014-2> accessed 27 August 2018

20. Samuel Dodge and Lina KaramKaram, 'A Study and Comparison of Human and Deep Learning Recognition Performance Under Visual Distortions' [2017] Arizona State University <https://arxiv.org/pdf/1705.02498.pdf> accessed 27 August 2018

21. Jason Cipriani, 'What is Google Duplex?' (CNET, 2018) <https://www.cnet.com/how-to/what-is-google-duplex/> accessed 27 August 2018
22. Jason Dorrier, 'Robot Fighter Jet X-47B Autonomously Lands on Aircraft Carrier' (Singularity Hub,

2013) <https://singularityhub.com/2013/07/19/robot-fighter-jet-x-47b-autonomously- lands-on-aircraft-carrier/> accessed 27 August 2018

prototyping, validation and deployment of Deep Neural Network inference application at the edge”²³ , special graphics card²⁴ and special input and output device.²⁵

Now in mid-2018, AI technology has taken off at rocket speed, AI is now outperforming human almost every specific job, from image recognition to image recreation, from self-driving to expert cooking. The dramatic change has made possible due to government and non-government funding. Individual and open source project like TensorFlow²⁶ by Google and Open AI²⁷ by Elon Mask have opened ample opportunity for entrepreneurs and creators to put their effort to solve diversified problems around the world.

Some companies are now developing a new kind of hardware which not only process data rather those hardware also act like a human brain with build in neural networking whereas, previously only software used to imitate that networking and processed the information; this technology is known as Neuromorphic Computing.²⁸ This new range of hardware and software now enabled AI to do some unbelievable tasks. For the same tasks AI consumes less power than ordinary system, AI does not always require large-scale computational power. Therefore, small mobile chips can perform AI tasks. Moreover, the brain or learning of an AI can be copied or redistributed. Now AI is developing as new AI by itself and surprisingly the child of the AI is better than human creation! It seems there are no limits of possibility and no threshold to stop. Similarly, the concerns of human reached a new level and there are questions that humans should have the answer on time.

There are queues of questions that need to be focused, but as the purpose of this paper is to find out legal challenges and look after the solution that can be proposed, therefore, this paper would focus on few specific areas where we need to rethink about the system and the law. In the previous part, it is obvious that the Turing Test of imitation can easily be passed by modern AI's, Now people have adopted the new

23. 'Intel® Movidius™ Neural Compute Stick' (Movidius NCS, 2018) <https://developer.movidius.com/> accessed 27 August 2018

24. 'NVIDIA: The AI Computing Company' (NVIDIA, 2018) <https://www.nvidia.com/en-us/about-nvidia/ai-computing/> accessed 27 August 2018
25. Timothy J. Seppala, 'Microsoft Repurposes Kinect to Boost AI and Azure Development' (Engadget,

2018) <https://www.engadget.com/2018/05/07/microsoft-project-kinect-for-azure- build-2018/> accessed 27 August 2018

26. Martín Abadi and others, 'Tensorflow: Learning Functions at Scale Machine Learning' (2016) 51 ACM SIGPLAN Notices https://www.usenix.org/ system/files/conference/osdi16/osdi16-abadi.pdf> accessed on 1 August 2018
27. 'About Openai' (OpenAI, 2018) <https://openai.com/about/> accessed 27 August 2018

28. Steve Esser and others, 'Backpropagation for Energy-Efficient Neuromorphic Computing' (Papers.nips.cc, 2015) <https://papers.nips.cc/paper/5862-backpropagation-for-energy-neuromorphic-computing> accessed 27 August 2018

definition of intelligence; "Intelligence is simply the capacity to express an appropriate behavior in response to changes and opportunities in one's environment... Things are considered more intelligent when they can more successfully achieve more goals in more varied environments." So, "Artifacts that implement any of those parts are said to be AI ..."²⁹

3.  Concerns of the use of Artificial Intelligence

AI is already more capable than a human for a specific task and by using this capability, it is already affecting our daily life both positively and negatively, a simple scrutiny will open the depth of AI for real understanding. AI has been used for the purpose of capitalization; to make more money. All most all the companies that work to develop AI systems used Big Data to identify an individual and to find out their vulnerability so that they could have maximum profit out of their investment. Social networking sites, apps and game companies use these analyses to keep users locked in their system; furthermore, some institution is also accused to experiment on users to alter their mood without their consent.³⁰ Most of the time the data is collected in the name of improved service or offering some virtual benefit for the user. That little push (Also known as nagging) of advertisement looks very harmless but the sheer amount of data they collect with those technical words enables them to collect huge sum of data. As a result, the amount of data they are handling with the help of powerful AI is unmatched with previous data analysis method. Furthermore, the ways they can use those analyzed data as well as apply those data, thus this use of AI can change our day to day life. As mentioned earlier when data is there action can be predicted and man can be influenced without giving them any idea that they are driven by other ideas.³¹

Though humans are very good at doing the "intelligent tasks" the ability of humans are still limited, therefore, humans are confined to their power, even the latest technology cannot enable humans to understand more than one human conversation at a time, humans cannot multitask³² nor can they upgrade their capability to a new level like computers. On the other hand, the internet of things³³ can enable certain AI

²⁹  Miles Brundage and Joanna Bryson, Smart Policies for Artificial Intelligence (2018) <https://arxiv.org/ftp/arxiv/papers/1608/1608.08196.pdf> accessed 27 August 2018

30. Samuel Gibbs, 'Facebook Apologises for Psychological Experiments on Users' (the Guardian, 2014) <https://www.theguardian.com/technology/2014/jul/02/facebook-apologises-psychological-experiments-on-users> accessed 27 August 2018

31. Rayhanul Islam, 'Legal Approach to Cyber Privacy' (The Daily Star, 2018) <http://103.16.74.140/law-our-rights/legal- approach-cyber 03814> accessed 27 August 2018.

32. Claudia Wallis, 'Genm: The Multitasking Generation' (TIME.com, 2006) <http://content.time.com/ 0,9171,1174696,00.html> accessed 27 August 2018

33. Neil Gershenfeld, RaffiKrikorian and Danny Cohen, 'The Internet of Things' (2004) 291 Scientific Am.<http://www.jstor.org/stable/26060727> accessed 27 August 2018

to communicate with other AI or system or Database and can use them with slight delay if not without delay, on-demand computing servers or add-on computer systems can greatly increase the power of a certain system, thus the result of the same task done by a human and AI would present a wide margin of difference.

In a hypothesis, “where AI is given a task to find out a man and stop him to before reaching a certain point, AI can do it all within few minutes!” Yes, that is, in fact, real in few cities of China. Imagine how that technology can change the Human Rights situation in China! ³⁴ Considering another aspect where a human administrator is employed to supervise AI activities to ensure the legality of its (AI’s) action, that administration rather be a victim of AI influence, because there is no way a human could understand Big data even with reasonable time, the reason is that the data would be so huge human could not be able to realize the situation, the same applies to a judge too, if that approval (of admin) end up into vital consequence then presented before the court. For the same reason, the inventor of a particular AI would fail to understand the result and consideration of that result of that AI. Therefore, specific regulation of AI is very important.

4.  AI Regulation

Before entering into other legal aspects, the regulation of AI and concerning issues need to be addressed through some experts like Andrew McAfee, co-author of the bestselling book the Second Machine Age, wrote "it's way too early for explicit AI Policy",³⁵ a similar voice is also raised by most of the AI companies, as a defense they refer two arguments, the first one is AI is still in earlier stage and the second one is that a policy may lead AI research to an unnecessary hassle, and that may forestall the substantial benefit, nevertheless it must be considered that AI is not like other technology and it is significantly affecting human life, for the second point, the guideline less continuity may end up more deleterious than the potential benefit and where the concerns are open question it would be negligent not to take any steps towards regulation.

AI regulation can very well be a creators philosophy implanted to their AI products which is more like De Facto Policy³⁶ of those institution, i.e. Policy of Apple how “Siri” should reply to an individual or which data of should it remember, how far to remember etc. but for the purpose of this paper “AI Regulation” must be a clear

34. Paul Mozur, 'Inside China’S Dystopian Dreams: A.I., Shame and Lots of Cameras' (Nytimes.com, 2018) <https://www.nytimes.com/2018/07/08/business/china-surveillance-technology.html> accessed 27 August 2018

35. Miles Brundage and Joanna Bryson, Smart Policies for Artificial Intelligence (2018) <https://arxiv.org/ftp/arxiv/papers/1608/1608.08196.pdf> accessed 27 August 2018

36. Miles Brundage and Joanna Bryson, Smart Policies for Artificial Intelligence (2018) <https://arxiv.org/ftp/arxiv/papers/1608/1608.08196.pdf> accessed 27 August 2018

structured minimum but mandatory instruction where it provides individuals and institutions with a clear outline in accordance with other laws, regulations, principles and other aspects of socio, economic and psychological concerns to maintain those instructions. AI Regulation under its proposed framework shall include creators to end users of AI product and it shall work with existing legal institutions, i.e. Legislature, Judiciary, executive or specialized institutions like separate ministry or special court.

Regulation of AI is another huge challenge for the development factors of AI. For example, development of AI does not require the individuals to be together at the same place, the code is such it learns automatically from its source, the output of an AI can be very different then it is anticipated, hence Ex-ante regulations may render fruitless as opposed to, ex-post regulation pose a risk to get out of control. AI system is not inherently limited by the preconceived notions, rules of thumb, and conventional wisdom upon which most human decision maker relies, AI systems have the capacity to come up with a solution that humans may not have considered or that they considered and rejected in favour of more intuitively appealing options.³⁷AI is performing better then and presenting better results than human but the question remains can a machine really realize human values? Consider that, Google's AI knows what advertisement may be preferred by the user and bring profit for Google³⁸ but does it really understand the human (emotional) value of that advertisement; for example, an ad for a baby bed to a pregnant woman who is visiting attorney website for divorce. These human emotions cannot be categorized and regulated by law, so this is another challenge.

Though regulating AI separately would be fruitless, from all the aspects of social science, psychology and physiology and the root of law needs to be understood with a new outlook. Challenges to the legal doctrine would be more specific to find out the answer as well as the whole legal framework should be redesigned. However, the surface of the present legal system can be drown to make a comparison or to find an apparent solution

37. Matthew U. Scherer, 'Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies' (2015) 29 SSRN Electronic Journal <http://jolt.law.harvard.edu/articles/pdf/v29/ 29HarvJLTech353.pdf> accessed on 5 May 2018.

38. NgridLunden, 'Google Debuts Adsense 'Auto Ads' with Machine Learning to Make Placement and Monetization Choices' (techcrunch.com, 2018) <https://techcrunch.com/2018/02/21/google-debuts-adsense-auto-ads-with- machine-.com/2018/02/21/google-debuts-adsense-auto-ads-with-machine-learning-to-make-placement-and-monetization- choices/> accessed 27 August 2018

5.  Big Data and its Use

AI uses Big Data, it also collects data because data is the main feed for AI due to the requirement AI constantly collects data and regularly update them.³⁹ The companies use AI to collect data from subjects with aesthetic wordings whereas the main target is together very specific data to understand the subject in details.⁴⁰ These data are the representation of human life in a numeric system, having most of the data of an individual can make him helpless even not clearly breaking any law. Using these data in our existing digital ecosystem an individual can easily be manipulated even if not implanted any idea. This is a serious violation of privacy, even the collection of these data is never exposed or used, simply the collection of these data is a privacy violation under various national and international law.⁴¹In a case it is observed that “the mere improper access to personal data created an increase ‘risk of harm’ is sufficient for a claim”.42 A recently leaked video of Google that was originally created to guide Google employee showed its interest to collect human data generation after generation and link them together to under their behavior, like human genes so that not only present generation but also future generation and their action can be explained and identified.⁴³ Humans are turned into data subject and companies experimented with those subject using psychology without the consent of the user.⁴⁴ Moreover, social networking sites, games and apps are constantly using tricks to keep the user online most of the time and all these are done with the help of AI technology. The use of AI for mere commercial purpose can be accepted to a certain level but when the same technology is used to make people addicted and manipulate people’s choice it needs to be regulated strictly. Mark Zuckerburg has accused to manipulate the US election 2016 and that led him to

39. Tanvir Zafar, 'Transforming The AI Industry with Next-Level Data Collection' (Hacker Noon, 2018) <https://hackernoon.com/transforming-the-ai-industry-with-next-level-data-collection-72f61b6a1b44> accessed 27 August 2018

40. Fei Jiang, Yong Jiang and Hui Zhi, 'Artificial Intelligence in Healthcare: Past, Present and Future' (2017) 0 Stroke and Vascular Neurology 2017 <https://svn.bmj.com/content/early/2017/06/21/svn-2017-000101.full.pdf> accessed 27 August 2018

41. Regulation (EU) 2016/679 of the European Parliament and of the Council, see also. Rayhanul Islam, 'Legal Approach to Cyber Privacy' (The Daily Star, 2018) http://103. 16.74.140/law-our-rights/legal-cyber-privacy-1503814 accessed 27 August 2018

42. David L. Silverman, 'Developments in Data Security Breach Liability' (2018) 70 The Business Lawyer <https://www.jstor.org/stable/43665697> accessed 27 August 2018.

43. Madison Dapcevich, 'Leaked Google Concept Video Shows “Ledger” of User Behavior Used to Manipulate Your Decisions' (IFLScience, 2018) <http://www.iflscience.com/technology/leaked-decisions/> accessed 27 August 2018

44. Samuel Gibbs, 'Facebook Apologisesfor Psychological Experiments On Users' (the Guardian, 2014) <https://www.theguardian.com/technology/2014/jul/02/facebook-apologises-psychological-experiments-on-users> accessed 27 August 2018

face Congress inquiry⁴⁵ but this is not the only case where Big data played a very big role with the help of AI⁴⁶ to change the course of history another accused is Cambridge Analytica,⁴⁷ Cambridge Analytica is also accused to play a vital role in Brexit vote influence.⁴⁸⁴⁹ These are some big fish that could not hide in the small pond but they are not the only player on the field. This type of manipulation can change the state of a country or even change the shape of the world. Therefore, Big data must be kept safe to avoid such threat of influence over citizens, hence a data bank can be used to collect all the data of individual presenting with proper reasoning before them, there should be a clear definition of public and private data and how those data could be used must be guided by law. Private data sharing among companies should be made very limited subject to the approval of both data subject (individual) and government.

To use such Big Data AI is expected to be straightforward and a non-bias system but so far it has proven the world wrong in place to place,⁵⁰it has been observed that AI has got the biases like any other individual, as AI is a data driven system therefore it vulnerable to data impurity,⁵¹ furthermore systematic biases also play a role to ensure the maximum profit for the companies who develop them.⁵² Before the world fully relies on AI these issues need to be solved but in the name of noise reduction of data if the data is heavily filtered the result of AI could be very wrong, so these issues should be handled very carefully.

45. 'Why Facebook Suspended a Data Company and How it Relates to Donald Trump' (ABC News, 2018) <http://www.abc.net.au/news/2018-03-18/cambridge-analytica-suspended-by-facebook/9560272> accessed 27 August 2018

46. Sam Biddle, 'Facebook Uses Artificial Intelligence to Predict Your Future Actions for Advertisers, Says Confidential Document' (The Intercept, 2018) <https://theintercept.com/2018/04/13/facebook- advertisingintelligence-ai/> accessed 27 August 2018

47. Kevin Rawlinson, 'Pressure Grows on PM over Brexit Cambridge Analytica Scandal' (the Guardian,

2018) <https://www.theguardian.com/politics/2018/mar/26/pressure-grows-brexit- cambridge-analytica-scandal-theresa-may> accessed 27 August 2018

48. Reuters Staff, 'What are the Links Between Cambridge Analyticaand a Brexit...' (reuters.com, 2018) <https://www.reuters.com/article> accessed 27 August 2018

49. Patrick Greenfield, 'The Cambridge Analytica Files: The Story So Far' (the Guardian, 2018) <https://www.theguardian.com/news/2018/mar/26/the-cambridge-analytica-files-the-story-so-far> accessed 27 August 2018

50. Matt Chessen, 'Encoded Laws, Policies, and Virtues' [2018] Cornell Policy Review <http://www.cornellpolicyreview.com/encoded-laws/> accessed 27 August 2018.

51. Erika Hayasaki, 'Is AI Sexist?' (Foreign Policy, 2018) <https://foreignpolicy.com/2017/01/16/women-vs-the-machine/> accessed 27 August 2018
52. ibid, 38

6.  Legal Personality

The concept of legal personality is nothing new, companies, organizations and teams are considered as a legal person under legal instruments. Though legal personality is awarded to those type of organizations, in reality, those are run by living human beings who are responsible for any action and those persons can be called into question for any legal issue. These are some common legal personality by expressed law with a regulation mechanism, whereas, court orders have introduced some unconventional legal personalities too. Presently some cities are considered a legal personality, rivers are also awarded legal personality by courts,⁵³ similarly some animals are also considered as legal personality, for example the effort of Non-Human Right Project (NhRP group) at Newyork has obtained a court order in favour of two research chimpanzees by writ of Habeas Corpus.⁵⁴ Few other examples are also around us too. Not very ago black humans were considered as non-human and no human rights were there to help them than the concept of humanity has changed and it is still changing, Now, NhRP is promoting person-hood for other “intelligent” animals, like dolphins, whales, and elephants etc. and all these are backed by humans. At this point, it is very clear using the existing legal system an AI can achieve legal personality status. Here the fundamental difference is that AI is really intelligent presently it outperforms human in many specific tasks, like in the toughest board game or cancer cell detection. it is also very good for legal research.⁵⁵

Considering the power of special legal researcher AI like ROSS⁵⁶ with collaboration AI companies, AI could claim legal personality exploiting the vulnerability of present backdated legal system. Some argument can be placed challenging age-old definitions like ‘Birth’, ‘Right’, ‘Citizenship’ etc. The criteria of legal citizenship or prove of epistemological and ontological existence within person can be challenged, contrarily, where there is a test (i.e, Turin’s imitation game) to prove the capability of legal or natural person-hood be very well favourable for AI to legally establish AI position as a person. On the other hand, companies could create highly designed and specific empathy for AI to manipulate humans to support to achieve their goal. Now, considering the legal personality of Sofia AI which is given by the kingdom of Saudi-Arabia AI and their owners may not have to fight that much, there would be huge consequences if Sofia evolves as general purpose AI within few years, to consider some

⁵³  Planet Earth, '3 Rivers Just Became Legal 'Persons' (Live Science, 2017) <https://www.livescience.com/rivers-just-became-legal-persons.html> accessed 27 May 2018.

54. Jane Hu, 'When is an Animal a Legal Person?' (Pacific Standard, 2015) <https://psmag.com/ environment/is-chimpanzee-a- person> accessed 27 August 2018
55. 'About' (ROSS Intelligence, 2018) <https://rossintelligence.com/about/> accessed 27 August 2018

56. ibid

wider perspective lets think Sofia is a citizen of USA. It could claim many rights, might want to be independent, own income and life or social benefit, health benefit etc. The most interesting one is that it might claim the reproductive rights too, it might create its own child and then the law would bound to allow that child as legal person, as observed in Skinner v. Oklahoma case.⁵⁷ These may seem like a futuristic friction but actually it is not, it is quite clear with more human like AI (humanoid) which is designed to act like human may want to lead a human life with re-productive right, as observed in the case of Sophia.⁵⁸ To ensure these rights and to deal with new kind of evidence because now AI can create doubtless fake speech and videos which is beyond the capability of humans to clarify a human-centric AI based ecosystems, must be developed that could easily turn up an AI centric ecosystem in no time. In “I robo” si-fi movie where “VIKI”; a central AI, take over the human control not by violating any law rather explaining human set laws by its own understanding and interpretation to save humans, the same could also happen which would be a clear conflict of interest and then humans would have very limited scope to resolve the past error. A miniature version of the machine taking over has happened with Facebook AI where two test AI were unplugged when they create their own language method and somewhat take over the internal system.⁵⁹ Similarly, Sofia has threatened humans and other AI in the chat box has also drawn public attention by using slang to text people openly on the

internet.⁶⁰

7.  Responsibility and Liability

AI works with Big Data which comes from various sources then it learns with time by applying those data in various situations. For example, Tesla semi-autonomous car uses GPS, Infrared, Proximity sensor, Traffic light signals etc, it also learns from human drivers, it tries to understand what would be the next step of front car which it does with constant prediction which comes from the AI intuition that is developed by observing such kind of driving experience.⁶¹ Hence there are various sources from

57. Skinner v. Oklahoma ex rel. Williamson [1942]316 U.S. 535 (1942) 782 (U.S. Supreme Court)

58. Sophia The Robot Wants a Baby and Says Family is 'Really Important' (Bbc.co.uk, 2017) <http://www.bbc.co.uk/newsbeat/article/42122742/sophia-the-robot-wants-a-baby-and-says-family-is-really- important> accessed 27 August 2018

59. Andrew Griffin, 'Facebook Robots Shut Down After They Talk to Each Other In Language Only They Understand' (The Independent, 2017) <https://www.independent.co.uk/life-style/gadgets-and-tech/news/facebook-artificial- intelligence-ai-chatbot-uage-research-openai-google-a7869706.html> accessed 27 August 2018

60. Associated Press, 'Microsoft kills 'inappropriate' AI chatbot that learned too much online (Latimes.com,

2016) <http://www.latimes.com/business/la-fi-0325-microsoft-chatbot-20160326- story.html> accessed 27 August 2018

61. Lex Fridman and others, 'MIT Autonomous Vehicle Technology Study: Large-Scale Deep Learning Based Analysis of Driver Behavior and Interaction with Automation'<https://arxiv.org/

which AI is learning as much as the basic coding influence. The question comes, who would be liable for an accident? In Tesla car, they play a trick to overcome these problems they require a human driver behind the wheel so that he can take over in certain circumstances to make the liability clear to the driver of AI Car.⁶² On the other hand, Waymo's fully autonomous car can drive by itself where human has the control over AI but not mandatorily required to be active at all time.⁶³ In a hypothetical situation where the driving AI has only two options, i) to save its passengers killing few innocent children or ii) To save children and heavily injure the passenger inside, a human in a critical situation can select any of that two choices depending on the various point, fixing any single policy for AI might turn out polar opposite to the owners human sense but exactly that's what programmers are doing.

That implies though AI is featured to be self-learners it can be manipulated by code or data thus is also vulnerable, but it is also important to put a multilevel human-centric control system that can filter out the potential and visible threats. However, that would be another challenge when industry giants would be a party of legal procedure or when invited to determine the control regulation. They are likely to invite to collaborate with the government but wise sense can predict they would be effective to influence the system to take fewer burdens in case of responsibility and they are likely to succeed with lobbying and manipulation for their commercial benefit. Therefore, full reliability on AI could end up deadly for human civilization. Considering the partial autonomy, the immediate controller over an AI could be made liable for any occurrence, as we can see with the autonomous driving but these areas are yet to be clear, therefore, an AI regulation is much needed.

The concept of digital trespass⁶⁴ and fiduciary relation⁶⁵ concept in the matter of data safety liability can be brought so that the burden of proof in case of misuse lies on companies as they are supposed to understand the expert AI system better. These concepts can push the companies to be more careful than ever. The Tort liabilities like negligence and strict liability must be expressly introduced to the developers. Because

pdf/1711.06976.pdf> accessed 27 August 2018.Katie Fehrenbacher, 'Http://Fortune.Com' (Fortune, 2015) <http://fortune. com/2015/10/16/how-tesla-autopilot-learns/> accessed 27 August 2018

62. Models Owner's Manual (Teslr Company 2018) <https://www.tesla.com/sites//files/model_s_owners_ manual_north_america_en_us.pdf> accessed 27 August 2018.

63. Waymo Safety Report; On The Road to Fully Self-Driving (Waymo 2018) <https://storage.googleapis.com/sdc-prod/v1/safety- report/waymo-safety-report-2017.pdf> accessed 27 August 2018

64. Hannah L. Cook, '(Digital) Trespass: What’s Old is New Again' (2017) 94 Denver Law Review <Https://Ssrn.Com/Abstract=2923211> accessed 27 August 2018

65. Robert Cooter and Bradley J. Freedman, 'The Fiduciary Relationship: Its Economic Character and Legal Consequences' (1991) 66 NYULR<https://pdfs.semanticscholar.org/cf87/95a89ebdc1894c722126542ea fc9964e8959.pdf> accessed 27 August 2018

by the nature tort law come into force when a cause of action is already there, that gives the court to scrutinize the situation in details with all evidence before him, tort law is applicable case to case basis which is very suitable for AI regulation as AI development is very diverse from one another. Moreover, tort law is more focused on public safety than the financial aspect.⁶⁶

A scientific experiment where it poses a considerable threat to human life, tradition or society must be approved by the government, the subjects must be fully informed with the full written consent, and before running it in wild it must pass multiple level tests and trials like pharmaceutical products.

8.  IP rights and Concern

Intellectual Property (IP) has a close relation with Artificial Intelligence but the relation is yet to be defined. IP right is the incentive for the creators who create creative, original and useful work, IP right comes in various category like, Copyright, Patent, Industrial design etc. which was so far only eminent in the human domain. Recently the emerging AI's are being very creative, they are now performing all type of creativity, they can write poem, article, create music and voice, they can design and edit videos etc. These activities meet almost all the criteria to get AI their IP right except the lacking legal personality, the US Copyright office has declared the work must be created by human.⁶⁷ So, the companies that created the AI can use the AI creation as their Intellectual Property but a recent judgment of an Australian court declined that view, "court declared that a work generated with the intervention of a computer could not be protected by copyright because it was not produced by a human.⁶⁸ "The same is supported by a famous chimpanzee selfie copyright case in the USA.⁶⁹ There is a clear vacuum of a principle for those who invested millions of dollar behind the development of those AI. There must be some reasonable consideration for those companies so that they use legal protection and share their technology so that the impurity of data can be correct and that would help the development of AI as well as regulators to regulate. Moreover, the monopolistic approach of the companies must be limited by law to the creation of AI, though the Universal Basic Payment is a good policy the catch is that it only applies to the citizens of that country whereas if the

66. Matthew U. Scherer, 'Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies' (2015) 29 SSRN Electronic Journal <http://jolt.law.harvard.edu/articles/pdf/v29.pdf> accessed 27 August 2018

67. Feist Publications, Inc. v Rural Telephone Service Co., Inc. [1991] 499 U.S. 340 (1991) 89-1909 (Supreme Court of the United States)
68. Acohs Pty Ltd V Ucorp Pty Ltd [2012]

69. Bill Chappell, 'NPR Choice Page' (Npr.org, 2014) <https://www.npr.org/section /thetwo-way/2014/342419651/who-owns-a- monkey-says> accessed 27 May 2018

272

ELCOP Yearbook of Human Rights 2018 Md. Rayhanul Islam

creation is openly shared after a reasonable time the third world countries would have some space to survive.

9.  Preservation of Certain Human Issues

As this article has shown various aspects of AI, it is significantly outperforming humans, it is faster, creative, cost-effective and many more. There are many reasons why human should use AI to improve the situation of the world but in certain places, AI must be prohibited for good reason ignoring the economic benifit, simply to nurture very special human feelings and values, similarly as we preserve wild sanctuary for the purity of nature or sometimes to keep them alive and apart from extension that otherwise would very obvious. There are many system that are designed for humans that are attached to human feelings, human decision and human achievement, these systems are not perfect because humans are not perfect, therefore, these imperfect systems are known for their imperfection, fixing these imperfect systems is a task of humans, human should use their own skill to perfect these systems. For example, democracy is not perfect, AI could be used to understand the lacking of democracy but by no means it can be employed to direct or manipulate an election, that would be an inhuman approach even if it is designed to bring the right person in the chair and even if that effort would succeed.

Lawyers and courts are using AI to put their best argument and to provide proper sentencing.⁷⁰ for the first part when a lawyer is preparing for his submission to convenience the juries and the court he must have some skill that is practiced and sharpened by him but if he relies based on the Argument design by an AI that would not be very appropriate to the context, those who put their best effort may fail or those who could not afford that AI would be frustrated, the whole justice system shall start to collapse for the use of AI such sensitive place can be compared to use of prohibited drugs at Olympic games, In respect of court using AI for criminal sentencing is quite inhuman and contrary to their stand about criminal profiling.⁷¹ For a simpler example; an AI can be used to teach a mother how to nurture her baby but it should never be used to replace the place of a mother. Our tendency to the easy solution may lead the own characteristics of humanity in trouble.

70. Jason Tashea, 'Courts are Using AI to Sentence Criminals. That Must Stop Now' (WIRED, 2017) <https://www.wired.com/2017/04/courts-using-ai-sentence-criminals-must-stop-now/> accessed 27 August 2018

71. Rayhanul Islam, 'Forensic Psychology and Present-Day Use | 02 | Law Help BD' (Law Help BD, 2018) <https://lawhelpbd.com/article/forensic-psychology-and-present-day-use-02/> accessed 27 August 2018