Future of Artificial Intelligence: Safe AI Development

Abstract

From voice automated personal assistants to self-driving cars, there has been a tremendous progress in the field of Artificial Intelligence (AI). While the general notion of AI is depicted as robots behaving like human beings, AI can actually encompass anything from game playing agents to autonomous weapons. Though most of the research today is focused on solving one particular task like Image Processing, once combined, AI can overtake humans in almost all cognitive tasks. Progress in AI might be the biggest event in human history, however, some researchers are worried that it might also be the last. While digital life is augmenting human capacities, it is also turning manual effort in a huge number of tasks, redundant and unnecessary. AI isn’t just changing homes; it’s finding a way into their numerous industries and creating disruptions at the workplace. Thus the question lies in is this entirely beneficial? Many fear that the rise of AI will lead to machines and robots replacing human workers and view this progression in technology as threat rather than a tool to better ourselves. Thus it is extremely important to use technology for ethical purposes, keeping in mind we align our goals with those of development in AI.

Introduction - Evolution of AI

The concept of artificial intelligence (AI) is quite ancient. Intelligent robots appear in the fictions and mythologies of many ancient societies, including Greek, Arabic, Indian, Egyptian and Chinese. Today, the field of artificial intelligence is more diverse than ever earlier. Many industry leaders and frontline technology developers believe that we are on the threshold of discoveries that could change human society irreversibly, for better or for worse, the latter being more likely. Actually, AI is all around us in its primitive forms. It gets used when the aggregator cab apps suggest options to us for travel, gmail categorizes mails in inbox; more than 50,000 driverless cars of Tesla ply in the roads in USA, e-commerce softwares, electronic games, and so on. John McCarthy, who originally coined the term “artificial intelligence” in 1956, had said that “As soon as it works, no-one calls it AI anymore.”

To highlight this trend, today AI is regarded as a way of making a computer, a computer-controlled robot, or a software to think intelligently, in a manner in which intelligent humans think. It is accomplished by studying how human brain thinks, and how humans learn, decide, and work while trying to solve a problem. The outcome of such studies forms the basis for developing intelligent software and systems. To do this, the software program that runs the brain of the computer uses adaptive learning algorithms, in which the program uses knowledge of the outcome of an earlier action to modify the action for a more desirable outcome. Eventually, one day, artificial intelligence will enable a computer to think by itself by modifying its own program, without being programmed to do so.

Future of AI – Safe AI Development

Artificial Intelligence (AI) has the potential to improve productivity, efficiency and accuracy across an organization – but is this entirely beneficial? Many fear that the rise of AI will lead to machines and robots replacing human workers and view this progression in technology as threat rather than a tool to better ourselves. It is likely that artificial intelligence will soon replace jobs involving repetitive or basic problem-solving tasks, and even go beyond current human capability. AI systems will be making decisions instead of humans in industrial settings, customer service roles and within financial institutions. Automated decisioning will be responsible for tasks such as approving loans, deciding whether a customer should be taken on board, or, for identifying corruption and financial crime.

On the other hand, many businesses and individuals are optimistic that this AI-driven shift in the workplace will result in more jobs being created than lost. As we develop innovative technologies, AI will have a positive impact on our economy by creating jobs that require the skill set to implement new systems. 80% of respondents in the EY survey said it was the lack of these skills that was the biggest challenge when employing AI programs. Organisations will benefit from an increase in productivity as a result of greater automation, meaning more revenue will generated. This thus provides additional money to spend on supporting jobs in the services sector.

In the present phase of development, Artificial intelligence is referred to as narrow artificial intelligence, which is designed to perform a narrow task, e.g., only facial recognition or only internet searches or only driving a car. However, it is anticipated that in the future many researchers will try to create what is known as general artificial intelligence (AGI or strong AI). Narrow AI may learn to do better than humans its specifically designed task, like playing chess where the outcome of all possible allowed moves, no matter how unlikely, can be computed by the computer which the human brain can never do. However, it cannot do subjective common sense reasoning while taking decisions, such as choosing a safer path and adapting oneself for better balance while going down a hill. AGI on the other hand may be able to carry out all cognitive tasks in an integrated manner, and subject to development of suitable machines, outperform humans in all such tasks.

In the long term, the greater concern or hope, depending on the outcome, is that an AI system would become better than humans at all cognitive tasks, by potentially carrying out repetitive adaptive self-improvement, producing an intelligence level far exceeding the human intellect. Although, some experts feel that since designing a smarter AI system is itself a cognitive task, it can never exceed the smartness of the creator. However, this is a grey area where others feel that the software can be programmed to modify itself, and once such features are incorporated, it might do so in a totally unpredictable manner. Such evolutions of AI may turn out to be beneficial, and help us eradicate disease, hunger and poverty. In this way, the evolution of strong AI might be one of the biggest events in evolution of human beings. However, some other experts continue to express concern, that it might also be the last resulting in bondage of humans to machines, unless we develop safety mechanisms or means to align the goals of AI with those of human goals, before it becomes super-intelligent with its own goals which might not be desired.

Although there are counter arguments, to which most researchers agree, that a super-intelligent strong AI is unlikely to exhibit human emotions like love or hate, and hence there is no reason to fear that AI would ever become intentionally benevolent or malevolent. However, the fallacy in the counter argument is that even if it is designed to do beneficial tasks, it might self-develop into taking otherwise destructive steps in the long term, while trying to achieve the short term beneficial goal in the most efficient manner. Thus, what is of major concern is a half-baked partial knowledge based strong AI, which is more likely to be the intermediate stage product in the path of evolution of AI. This is almost inevitable, because even human knowledge is vastly limited when it comes to even mid-term outcome of actions such as in the field of environment, climate, sociology and human-animal interactions, and science and technology. Our limited knowledge of outcome resulted in long term hazards such as enhanced pollution and green house effects, weapons of mass destruction, nuclear accidents, unending human conflicts. Thus it might be an unavoidably incomplete and hence incompetent AI, rather than an evil AI, which might go out of control.

Bibliography:

1. Gods and Robots Myths, Machines, and Ancient Dreams of Technology - Adrienne Mayor, Princeton, 2018.
2. The Guardian view on the future of AI: great power, great irresponsibility, https://www.theguardian.com/commentisfree/2019/jan/01/the-guardian-view-on-the-future-of-ai-great-power-great-irresponsibility downloaded Feb 27, 2019.
3. The Future of Artificial Intelligence In The Workplace, Vishal Marria, Cybersecurity, Jan 11, 2019, downloaded on Feb 27, 2019 from https://www.forbes.com/sites/vishalmarria/2019/01/11/the-future-of-artificial-intelligence-in-the-workplace/#208afa4f73d4
4. Benefits & Risks of Artificial Intelligence, Max Tegmark, Future Life Institute, downloaded from https://futureoflife.org/background/benefits-risks-of-artificial-intelligence/?cn-reloaded=1
5. What is AI? Ten things you need to know about the future of artificial intelligence, Duncan Geere & Vaughn Highfield, Aug 2018, downloaded on 27 Feb 2019 from https://www.alphr.com/artificial-intelligence/1002792/what-is-ai-ten-things-you-need-to-know-about-the-future-of