Manoranjan Shrivastawa, ITS officer-Department of Telecommunications, Ministry of Communications
In his famous book titled What Technology Wants, Kelvin Kelly responds to a question whether the self-evolution of technology will lead the society in desirable direction. His answer is Yes, as he believes that technology wants what society wants. It is based on the premise that the demand of a technology is function of the extent to which it increases the efficiency, opportunity, and freedom. While it is difficult to make sweeping generatizations, about the potential impacts and directions of technology, the experiences of history are rippled with conflicts between technological change and social progress. The disruptions caused by technological progress, in last century, economized production and brought fundamental changes in society, increasing wealth and prosperity and improving the quality of life for many. Yet, economic modernization was not without critiques. Many scholars of repute have questioned the models of development brought by globalization of technologies and production process in terms of increased inequality, rate of crime, and sustenance of poverty. The potential of technologies to bring destructions and create instability has also been realized by global leaders and experts. It made expedient for the world leaders to create multilateral conventions, rules, and byelaws to ensure that technologies can be used only for constructive and peaceful purpose. So, what should be clear that while technological progress brings opportunities for higher growth and improved standard of life making the process simpler, it also may contain seeds for destruction and instability making society vulnerable to its potential adverse impacts. The emergence of artificial intelligence (AI) should be seen in this context and effective intervention should be made to ensure that AI is used for the benefits of society and humanity.
The term artificial intelligence was coined by John McCarthy in 1956 as the topic of Dartmouth Conference (J McCarthy, ML Minsky, N Rochester, CE Shannon). The idea of programming, a machine to behave intelligently and make decisions based on context and reasoning was a plausible concept. However, it raised many issues related to the design and implementation of AI systems. Besides the capacity to program a machine and required processing capacity and memory; higher degree of freedom and flexibility for expression of creativity and controlled randomization of machine was required for machines to stimulate human brain. Designing an AI system also required extensive set of neural networks so that the machine can learn by examples and can do recursive self-improvisation.
AI can be defined as embodiment of intelligence in machines in order to enable them to take decisions and perform tasks on their own. They use models and algorithms to enable machines to take decisions using reasoning and experiences. Machine learning involves training those algorithms to adjust and improve. This training of algorithms requires huge amount of data to be analyzed and make suitable models for learning the patterns and developing intelligence. Thus, AI uses a combination of techniques, like big data analytics and machine learning to achieve the intelligence.
Major technological advances have taken place since 1990s in AI. Significant demonstrations in machine learning, intelligent tutoring, case-based reasoning, multi-agent planning, scheduling, uncertain reasoning, data mining, natural language understanding, and translation, vision, virtual reality, games, and other topics took place. In 1996, EQP theorem prover at Argonne National Lab proved the Robin Conjecture in Mathematics. The Deep Blue Chess Program defeated Gary Kasprow, a widely acclaimed chess player in May, 1997. NASA’s pathfinder mission made a successful landing and the first autonomous robotics system, Sojourner, was deployed on the surface of Mars. In 2000s the interactive robot pets became commercially available. The DARPA Grand challenge race in October 2005 was won by an autonomous driverless vehicle, Stanley.
The AI has potential to redefine humanity and civilization. The ability to take decisions based on context and infusing competency in machines to perform tasks in an autonomous manner reduces the gap between human being and a machine. AI is progressing rapidly. We have driverless cars on the roads. Robots have been made that act like human beings and also can learn from environment and respond to emotions by facial expressions. Recently, Saudi Arabia has even granted citizenship to a humanoid robot named Sophia. Facebook was made to abandon its AI experiment after two artificially intelligent programs appeared to chat with each-other in a language which only they could understand. These instances of using AI cause valid concerns of competent machines having goals misaligned with ours. The intelligent machines can work differently than expected and sometimes, bring a threat to human safety and security of systems. For example, if you ask an obedient intelligent car to reach to destination as quickly as possible it can wreak havoc on the road by causing accidents and may reach the highest possible speed, causing the passengers discomfort and jeopardizing safety. Similarly, an AI machine tasked with geo-engineering project might cause havoc in the entire ecosystem causing disruption in the natural balances of materials and affecting life and livelihood of people. The AI technology has reached a point where it is practically feasible to deploy autonomous weapons. It could have a potentially disastrous impact on international peace and stability and open the race among nations to acquire the autonomous weapons. The decision by autonomous weapons to intercept, target, and eliminate without meaningful human supervision is catastrophic and therefore, the tenets of international humanitarian laws governing warfare and other conduct among nations are at risk.
The applications of AI may vary across sectors and so does the nature and intensity of risks. For example, AI brings immense opportunities for investors while making financial decisions by giving better analysis of business and automating financial processes that brings efficiency and better customer experience. It could help in more efficient allocation of resources, lowering the operating costs, and improving risk management strategies. Nonetheless, it raises significant issues of monitoring as it may not always be possible to understand the communication mechanism used by machines. Defining responsibility in case of decisions causing losses becomes a major challenge. Terrorist organizations can also use AI for causing destruction of life and property without getting caught. Almost similar threats and vulnerabilities do emerge in using AI in other sectors too. Consider a driverless car causing an accident and leading to death. This can be an instance of a machine not behaving as per expectations or else, the machine can be programmed intentionally in such way that it kills people on road in an engineered accident. In such cases, the issue of determining culpability becomes challenging.
There have also emerged other legal and ethical issues related to privacy and security of data. They relate to the ability to access data being processed by AI tools. Data ownership rights, data privacy and cross-border flow of data are important issues to be evaluated in light of the new technologies, like AI, machine learning and big data analytics. Telecom Regulatory Authority of India (TRAI) has floated a consultation paper on privacy, security, and ownership of data in September 2017 and is in the process to finalize its policy recommendations. The issues raised herein needs to be analyzed in a broader framework of AI and its cross-sector implications. The infusion of ethical norms and value systems in machines in order to ensure the social acceptability of AI systems remains a major challenge. There is a possibility that AI may accentuate and reinforce the existing biases and prejudices in society. For example, the job selection algorithm may give preference to a particular caste or community and the college application screening algorithm may be used to filter a particular race or gender out. While, these possibilities are real and may accentuate structural inequality in society against the principle of social justice, the same AI technology also presents opportunities to reduce discrimination and inequality based on the programming choices.
The development of AI and its greater adoption provides immense opportunities to solve problems, bring economic growth, and add to wealth and prosperity. Its applications may range from improving transportation, creating smart machines, improving productivity, reducing poverty, and improving business decisions to deployment of autonomous weapons and creating artificial agents for defense of borders and managing disaster. While AI is still in a developmental phase and large-scale commercialization is yet to happen, it brings various legal and ethical challenges before national authorities. Greater adoption of AI technologies, its proliferation across multiple sectors, and uncertainties involved in a machine’s behavior shall make meaningful supervision of the AI industry expedient on national regulatory agencies. In order to ensure that AI technologies are used to accomplish only socially desired tasks and solve problems of humanity, regulations based on sound principles and robust enforcement mechanism shall be required.