It’s impossible to only regulate ‘bad’ AI without stifling the ‘good’. Tech is too far off and unpredictable AND humans have no current baseline for understanding, let alone assessing or guiding, productive means. Premature regulation drives straight to extinction.

3. Protecting human dominance through regulation or setting tailored goals to maintain human existence

Having a timeless and robust definition of AI is of paramount importance when thinking of regulating AI. One cannot regulate a certain subject without establishing a robust definition of what it regulates. The ambiguity of the definition of AI is mainly due to the “I”, “intelligence” of the AI. Concepts like “intelligence”, “consciousness”, “free will” and “soul” accompanying it are yet to have deterministic definitions although the greatest minds of our planet have tackled them for thousands of years (Burkeman, 2015).21

Neither any of the foregoing definitions of AI, nor many other definitions in the academia presents adequate definitions that can be satisfactory when regulation techniques are considered. In addition, the lack of definition is only one of the problems regulators will face; they will need to tackle liability gaps, control and transparency problems (Danaher, 2015).

In light of the foregoing, our primary statement stands firm: it is very early to begin thinking about regulating AIs or AI studies, particularly if such regulations may hinder developments that could prove essential for human existence. The turning point in AI development will probably be the development of ANIs, which should be encouraged through regulation, not restricted. However, if humanity fails in establishing adequate safe guards for ANIs, science fiction may turn into reality. Goertzel and Pitt (2012) call this the ‘AGI Sputnik moment’.

3.1. The great AI hype of 2015

Elon Musk’s and Stephen Hawking’s fears, Bill Gates’ cautious approach, Kurzweil’s optimistic take and Bostrom’s realistic analysis on the future that will probably be painted by AIs point to a single fundamental and existential dilemma: Are we going to be extinct because of AIs or will we maintain our existence with the help of AIs?

The cycle of extinction and rise of species may be the greatest success of evolution: ensuring the continuity of life. Over 90% of all species that ever existed on Earth went extinct and humanity’s fate will be no different, unless we come up with methods to achieve transcendence over evolution.22 Urban (2015) also treats this concept with a less theatrical manner and stresses two major outcomes for a possible ‘ASI Sputnik moment’. He states that either the introduction of ASIs will make immortality possible for our species or it will drive the human race into extinction.

Evolution has granted us our strongest instinct: survival. Instinctively we are in a never-ending war with nature, aiming to prolong our existence. In the abstract, the field of medicine solely exists for this purpose. Therefore, instinctively we will either try to eliminate the existential threat that ASIs might pose against us when we face the threat itself or try to eliminate a potential threat prematurely and in so doing cause our own extinction.

3.2. Reshaping perception on law

We may be living in the dawn of the age of artificial intelligence today. Consequently, the legal landscape surrounding our lives will require rethinking, as the case was with every big leap in technology. The industrial revolution brought conveyor belts and mechanical manufacturing processes operated by workers for longer and longer hours, which ended in myriad clashes between proletariat and employers. Hence, we developed labour laws, bringing a humanitarian minimum standard for the workers that were suffering from extreme working conditions. Similar legislative efforts followed each time when technologies required us to adapt new paradigms they introduced, technologies such as electricity, telegraph, telephone, railroad, automotive, television, and computers and so on. . . Below we will seek answers to some exemplary questions as to how AI might reshape our thinking, in terms of certain matters of current and prospective law.

3.2.1. Liability on damages

There are very few laws or regulations that address the challenges raised by AIs, and no courts appear to have developed standards so far, addressing who is legally responsible if an AI causes harm. The diversity and richness of individuals and firms that participate in the creation of an AI will make it difficult to identify the persons under liability. Certain technologies used in the development of an AI may date back to years before such AI is developed. Further, the developers of such technology may never have thought that one day, someone might incorporate their creation into any AI system. In such circumstances, it would be unfair to hold the developer of such technology responsible for a possible tort.

National and international laws do not recognize AI as a legal person. Therefore, current legal systems cannot hold them liable for the damages they might cause. However, what if an AI was fully autonomous and aware of its actions, causing harm knowingly and willingly?

This brings us back to the debate on consciousness. A conscious AI should naturally be liable for its actions. However, how can that be possible if we keep refraining from coming up with an adequate definition of what an AI is as far as legal ‘beings’ are considered? Should we ascribe legal personhood to them? (Paulius et al., 2015).

3.2.2. Intellectual property

IP law and its application places human initiative at its core. Berne Convention of 188623 requires an ‘author’ and an ‘artistic work’ to begin talking about intellectual property. While there is no limitation as to what form a ‘work’ can assume as long as humans can perceive it, an author must be a ‘human’. A San Francisco court applied and materialized this concept in 2015 by deciding in a lawsuit by PETA, the renowned organization defending animal rights, against David John Slater, a professional photographer, that a macaque money cannot own copyright to a selfie it took using the photographer’s camera (Kravets, 2016). What about AIs though? Can they own copyrights to the artistic works they create? Should law consider them as ‘individuals’?

3.2.3. Copyright and AI

Currently, a handful of AI applications are capable of producing works that resemble ‘art’, such as Deep Dream and the Cybernetic Poet.

Google’s® researchers developed DeepDream® to create a human-like image recognition software to identify certain things through mimicking human cognitive abilities. DeepDream uses Google’s artificial neural networks protocol to discern and process images of things to learn what they look like, such as a cat.

Google’s developers taught DeepDream what a cat looks like by showing millions of images of cats. Then they put DeepDream’s learning and identifying abilities to test by asking it to identify cats in pictures with cats and if found amplify them, introducing a feedback loop to work on. Then the developers introduced a random image to DeepDream and asked it to enhance the image in such a way as to elicit a particular interpretation. This method enabled the developers to understand whether DeepDream understood the essence of the things it learns. As a result, DeepDream searched in the images provided for all the things the developers trained it to recognize and when it found the tiniest bit of reference, it enhanced the relevant reference to make it look like the thing it found similar. The resulting images were surprisingly close to works of art. Few predicted this phenomenon, including DeepDream’s developers.24

Ray Kurzweil developed a poem software in mid-80s, a computer-implemented method of generating a poet personality that reads poems and generates analysis models to build its personality, and ultimately writes poems; the ‘Cybernetic Poet’. Cybernetic Poet is “provided with an input file of poems written by a human author or authors. It analyses these poems and creates a word-sequence model based on the poems it has just read. It then writes original stanzas of poetry using the model it has created.” (Bridy, 2012)

Now, who owns the copyrights of the artistic works created by these AIs?

As explained, current law cannot vest ownership of the copyrights to an AI, as it is not ‘human’. However, the laws of the United Kingdom make express provision for copyright in computer-generated works and introduce the following definition: ‘works generated by a computer in circumstances such that there is no human author’.25 The copyright in such works under UK law vests in ‘the person by whom the arrangements necessary for the creation of the work are undertaken’. Concordantly, Irish Law adopts the same principles.26 However, the UK and Irish approaches to the issue surrounding copyright ownership of computer-generated works and not the works of an AI. Therefore, they overlook the possibility of ‘non-human’ copyright ownership, ruling out the possibility of an AI that develops its own creative abilities. Who will have the ownership then?

3.3. Regulate and dominate?

A regulatory oversight and governmental intervention is a need when the development of AI is considered.27 It is not common to hear a Silicon Valley entrepreneur who operates on the frontiers technological advancement, urge governments to directly intervene with a developing technology in the hope of preventing humanity to do ‘something stupid’. When such thing happened in October 2014, it created a ripple effect and caused ‘The Great AI Panic of 2015’ (Sofke, 2015), which eventually led an institution called ‘Future of Life Institute® (FLI)’ to issue an open letter signed by Elon Musk, Stephen Hawking, hundreds of AI researchers in addition to many individuals representing U.S. government (Russel et al., 2015). FLI urged expanded research on how to contain AI systems within the walls of human benefit, including premature regulation. However, FLI used statements such as ‘AI systems must do what we want them to do’, ‘We should identify research directions that can maximize societal benefits’ and ‘AI super-intelligence will not act with human wishes and will threaten humanity’ while providing a research roadmap for AI researchers.

While the ‘we’ hints at a desired ownership over a technology under development (i.e. AI) and the ‘we’ implies superiority over ‘others’ in determining how a technology will be socially beneficial for humanity. It also begs the questions, ‘Who are you to claim that you have the capacity to force your desires over the entire human race, and who are you to claim that you can decide what is socially beneficial for us?’ Stating that an ASI will definitely be against the humanity’s welfare is an unexpectedly ignorant claim, allegedly coming from some of the greatest minds on Earth.

We experienced this line of thought when the Internet reached the masses, disrupting the status quo by lifting the boundaries of communication and information exchange and blurring the sense of control over disseminated information and access to such. The idea of an open interconnected network of networks that is not in anyone’s control or under any jurisdiction challenged lawmakers, policy makers and judiciary bodies and it still does. We have still been unable to set out universal rules on Internet (except DNS policies, where all stakeholders over Internet govern these policies through ICANN, a non-governmental organization) for almost 60 years. It would be very naïve to think that we can regulate AI policies, while AI is still in its infancy.

There is almost a consensus within the scientific AI community that definitive predictions on the future of ASI are impossible at this stage, simply because we are so far from creating an ASI, let alone understanding its implications.

3.3.1. Current and prospective regulatory efforts

Trying to anticipate ASI’s desires from where we stand now in terms of AI development is very similar to a chimpanzee trying to anticipate our motives when we crush an onion to remove its skin. Therefore, aiming to establish regulations to prevent ASIs from obliterating us is a hopeless endeavour. However, this line of thought may eventually lead regulators to prevent AI research from developing an AGI, fearing that it will break free from the chains of our capacity and become an ASI by itself. For example, John FrankWeaver, an attorney working in the field of AI law, praised the regulators at California when they intervened with Google’s self-driving cars and required test drivers to be present in these cars. He even claimed that this as a ‘wonderfully swift governmental response to autonomous technology and artificial intelligence’ while further supporting four states (Mississippi, Florida, Nevada and California) for passing restrictive regulation on autonomous cars that are not even on the market yet (Weaver, 2014).

3.3.1.1. Legislative efforts for autonomous vehicles. Nevada is the first U.S. state to enact a legislation authorizing the operation of autonomous vehicles in 2011 and was then followed by six other states, with many other states in still pending status with reference to their respective autonomous vehicle legislations. Tennessee among those who did enact such legislations stands out with its enabling and refreshing legislation wherein it prohibits local governments from banning the use of motor vehicles equipped with autonomous technology (Legislatures, 2016).

Throughout the world, legislators are working to incorporate autonomous (driverless) vehicles into their legislations to allow this thriving technology bloom and develop further, which brings hope.

The Convention on Road Traffic,28 of the United Nations, ratified by 73 countries, is in the process of amendment to allow automated vehicles on roads in many countries. European Road Transport Research Advisory Council published the roadmap for automated driving for Europe.29 German Federal Highway Research Institute published a report on the status of German legal landscape pertaining to vehicle automation technologies, indicating the areas of improvement on research, legislation and involvement of government agencies.30 Netherlands, Sweden, Japan and many other developed countries are actively working on improving the conditions of economic and legislative environment to enable swift development and consequently to reap the benefits of being involved in the forefront of innovative technologies.

While governments are honing in on preparing the legislative grounds for the operation of autonomous vehicles, academia adopts a wider approach and handles the concept in a wider manner, and works on determining the adequate policies for robotics and AI.

3.3.1.2. The RoboLaw project. The main objective of the RoboLaw project (“Regulating Emerging Robotic Technologies in Europe: Robotics facing Law and Ethics”) is to understand the legal and ethical implications of emerging robotic technologies and to uncover whether existing legal frameworks are sufficient in light of the rapid expansion of robotics technologies.31

The project was launched in March 2012 and funded by the European Commission (Paulius et al., 2015). The project produced the “Guidelines on Regulating Robotics”, which was then presented to the European Commission, to create the legal framework surrounding the development of robotic technologies in Europe.

The RoboLaw Project considered industrial robots, domestic robots, care robots, medical and surgery robots, autonomous vehicles, and humanoids/animaloids.The report discussed five essential legal areas for robotics: (i) health, safety, consumer, and environmental regulations; (ii) liabilities; (iii) intellectual property rights; (iv) privacy; and data protection and (v) capacity for legal transactions (Anon, 2015).

3.3.1.2.1. Health, Safety, Consumer and Environmental Regulation. The report identifies that common usage of robotics in hospitals, homes, commercial areas and our daily lives will require a new wave of legislations to cope with the prospective health and safety matters.

3.3.1.2.2. Liability. The report argues that imposing substantial liability on manufacturers, owners or users of robots for damages caused to third parties may increase safety while inducing wider social acceptance of robots. However, the report also argues that such approach on a liability regime may result in the displeasure of tech industry, consumers and, in the end, the general public, and may slow down the development of AI and robotics technologies. Therefore suggests a balanced approach between the interests of manufacturers, users, and third parties, and between risk regulation and stimulation of innovation, to encourage research, innovation and experimentation on these technologies, for increasing welfare in health, transport, commerce and other areas of business.

3.3.1.2.3. Intellectual Property Rights. RoboLaw Project indicates the lack of legal provisions that specifically apply to robotics. RoboLaw Project states that further research would be beneficial to determine whether the current application of intellectual property rights sufficiently meets the needs of the robotic industry and society.

3.3.1.2.4. Privacy and Data Protection. The RoboLaw Project suggests implementation of legal requirements into the robot’s software and interface through the ‘privacy by design’ approach, such as data security through data encryption and data access control in order to comply with the data protection requirements.

3.3.1.2.5. Capacity for Legal Transactions. The report stresses the lack of legal personality of robots and indicates that robots are seen as ‘mere tools’ to carry out commands that can, directly or indirectly, be attributed to human beings. Consequently, this approach requires the legal responsibility for robot actions to rest with their human ‘masters’.

It is possible to attribute legal personality to robots through legislative effort. Non-humans such as corporations, associations, and foundations gain their legal personalities through registration. The registration principle could be extended to robots and AIs (including requirements how robots can prove their registered identity); the capability of owning property is less easy to create, although legal constructions could be devised to accommodate this.

The report concludes with indicating that if these issues concerning legal personality are resolved at a certain point in time, more practical requirements and rules pertaining to legal acts will come into play, such as implementing legal conditions into the machines to make it possible for them to enter into a contract.

Lawmakers need to familiarize themselves with the potential benefits of AIs. Strict rules may prevent humans from the possible damages of AIs. However, these rules will also dampen possible improvements. Therefore, lawmakers should consider the balance between protection of humanity and development in technology.

4. Conclusion

When aiming to regulate currently non-existent technologies, we must avoid this approach at all costs. Putting restrictions on developing technologies based on our personal presumptions might indeed help us to avoid extinction at the hands of ‘evil robots’, but it might also cause our extinction due to natural reasons, such as evolution by making it harder for the human race to use technology to adapt.

Based on the statements of Elon Musk, SteveWozniak, Bill Gates, Bill Joy, Stephen Hawking and FLI’s open letter, it is clear that what they all fear is an ‘unfriendly AI’ and what they all want is a ‘friendly AI’ in the abstract.

The terms ‘friendly’ and ‘unfriendly’ do not refer to a personal trait of an AI system. These terms refer to whether the actions of an AI will have a positive or a negative impact on humanity (Urban, 2015). This is because AIs are computers and they do not have human values. We tend to anthropomorphize32 AI and attribute them with our moral values such as ‘good and evil’, ‘moral and immoral’ that are formed by our consciousness. These attributes developed only after thousands of years of social interaction. AIs will not share these human traits unless we specifically create them to do so. They operate on a task and goal oriented manner. To illustrate this point, for instance, there is an AGI, whose main task is to ensure that trees in a certain pine tree plantation are under protection from alien spores to keep the tree DNA as pure as possible. We should not be surprised when such an AGI takes drastic measures as far as obliterating the entire flying bug population in the area. One who is unaware of the goals of this AGI might easily label it as ‘evil’ and a ‘danger to humanity’ as he/she has no preconception on what the AGI’s motives or goals were. Similarly, a chimpanzee fearing that the crushing of an onion is a sign of aggression might attack us. Ironically, this view is very similar to the perspective of those who propose premature regulation of AIs.