The advent of computers propelled society and spurred into action a lot of businesses and industries that would otherwise have been impossible. Computers and computer technology changed the way humans worked and functioned. The very first computer was a behemoth. Today, everything that you need is available on the computer through the internet and even the most complex computations can be carried out by these modern wonders. But, what does the future have in store for these computers? Are there any more frontiers that these brilliant systems need to face and overcome, and as a result, become even more powerful computationally? Let’s find out.
Computers are already ubiquitous, but, they possess the ability to enhance their presence even more. Computers are on desks and countertops, in bags and pockets, but, soon, they might be a part of everything imaginable. Yes, everything conceivable, even the tiniest of devices and commodities, might soon have computers embedded within them. There is an increasing demand for products that can carry out the computations that they need on their own and this has led to the need for devices that have computers embedded in them.
Some people perceive the final goal of computers to be that of ensuring that computers are inextricable from society and human life. Computers would be so involved in every process that the very thought of its non-existence would make life impossible! The aim is to ensure that computers are entangled with human life in such a way that computers would be indistinguishable from human life.
The notion going around that the future of computing is limited to Artificial Intelligence is incorrect. Artificial Intelligence is still burgeoning and is expected to be an essential part of day-to-day life, but, the future of computing is not limited to just Artificial Intelligence, there is much more to computing.
The first challenge involved with Artificial Intelligence is the definition of intelligence. Intelligence has long been considered to be subjective and each person’s interpretation and definition of intelligence is generally different. Is a machine intelligent if it is able to communicate? Is it intelligent if it is able to learn? Or, is it intelligent if it can merely solve problems? One way to measure intelligence is through processing power. Machines can potentially gain computing powers that are greater than the processing power of all human beings put together. This situation is termed as the Singularity. The term was coined by John von Neumann and refers to the point when human affairs would cease, and computers would completely take over.
The low-end repetitive tasks have already been replaced by computing systems that automate it. The need for automation has made it imperative that we develop the computing potentials of computers. But, this growth has already caused a lot of concerns. There are several concerns over whether machines would someday surpass human intelligence. Currently, human intelligence, based on processing power, is considered to be 100,000 times more than that of the most powerful processing computer. But, it is estimated that computers can reach the processing power of humans within the next decade, at the rate of growth they are showing. By 2050, it is estimated that they would have processing power that is more than the sum total of all human brains combined. That is a scary thought indeed.
So, computers can take away our jobs and surpass our intelligence, what does that suggest? Is the world doomed? Not quite. Although one cannot quite rule out world domination by super intelligent computers and artificial intelligence, it is unlikely to happen due to the complexity brake. Scientists argue that computers and systems cannot grow to gain such super intelligence because as the processing power grows, the complexity increases and the complexity would reach a level that would restrain the growth of intelligence. Even if computers somehow broke free from the shackles of the complexity break, it is hypothesized that there would be enough ‘obedient’ intelligent systems that would keep the ‘misbehaving’ ones in check.
Let’s now get back to defining artificial intelligence. Since intelligence cannot be quantifiable, a generalist definition of artificial intelligence would be “machines and computers that carry out computations and tasks that would otherwise be carried out by humans”. This definition clearly does not capture the whole picture, but, is a reasonable attempt at defining A.I.
Evolution of Artificial Intelligence
The evolution of artificial intelligence can be viewed with regards to the forms of problems that it intended on solving. The early works in artificial intelligence focused on formal tasks, such as game playing and theorem proving. Checkers playing programs and chess playing algorithms were the earliest works in the field. Theorem provers such as the ‘The Logic Theorist’ aimed at solving theorems.
A.I. then began to be implemented in solving common sense reasoning problems. A.I. was used to simulate the human ability to make presumptions about everyday activities. Reasoning about physical objects and their relationships, and about actions and their consequences were the problems that came under this field.
When A.I. began to grow, it started being built for perception problems. These problems were especially difficult because they involve analog data such as speech and vision.
A.I. is now being used for natural language understanding and natural language processing. Humans possess the ability to communicate through expressions and languages. Perceiving the languages and various constructs of the languages and being able to process them, are all problems that A.I. was created to solve. These fields are still not fully developed, but, systems that carry out natural language processing with a high degree of accuracy are already in place.
A.I. is now being used in tasks where human experts are needed. These tasks, such as engineering design, scientific discovery, medical diagnosis and financial planning are being carried out using artificial intelligence with a great degree of accuracy and precision. Expert systems are computer systems that use inferences and knowledge bases to mimic the decision making ability of human experts.
Domains of Artificial Intelligence
Artificial intelligence consists of countless domains. It would be highly improbable to list all of them. Here are some of the interesting domains of Artificial Intelligence:
Heuristic techniques involve finding approximate solutions to problems instead of accurate ones. Heuristic techniques are generally used when the time for computing the solutions is restricted or when it is computationally too expensive to look for the global optimal solution.
This form of reasoning involves using probabilities and probabilistic theorems to determine most likely occurrences. Logical rules based on probabilities are created to determine what event is more likely to occur, given certain conditions.
Natural language understanding and processing
Natural language understanding involves the interpretation of natural language spoken by humans. Natural language processing, on the other hand, involves both interpretation and generation of natural language spoken by humans.
Expert systems are involved in carrying out tasks that would otherwise require human experts. These systems use inferences, which, are generally if-then rules, and the knowledge of previous problems and inferences derived from these previous problems, to provide solutions to the problem at hand.
Fuzzy logic systems
Fuzzy logic systems remove the notion of crisp boundaries and use fuzzy sets instead of crisp sets. In fuzzy logic, everything is viewed in terms of degree of membership. The hotness of a day is generally not quantifiable, but, fuzzy logic aims to quantify such occurrences using the concept of membership degrees. A temperature of 45°C, which, most people would perceive to be hot, could, therefore, belong to a set called ‘Cold’ to a certain degree (a small membership value) and to a set called ‘Hot’ to a certain degree (large membership value).
Genetic algorithms aim to mimic the biological behavior of human beings and animals to solve problems. Artificial Neural Networks have been widely used in the computation. These networks consist of nodes that are intricately interconnected just like the human neural networks. Outputs at each layer of a network is propagated to the next layer. This system is used to solve a lot of complex problems. Ant colony optimization is one of the areas of genetic algorithms, where actual ant colonies are mimicked to solve problems such as finding the shortest paths.
Evaluation of artificial intelligence: Turing test
Alan Turing hypothesized that a computer can achieve intelligence equivalent to that of a human being, or at least, intelligence that is indistinguishable from human intelligence if it could successfully lead a human being to believe that the machine was human. The Turing test is renowned for being the basis for classifying machine as either intelligent or not intelligent.
The Turing test involves a human evaluator who is responsible for determining which of the two entities he is talking to is actually human. The evaluator is allowed to ask a series of questions to carry out this classification. These questions are asked to both the human being and the machine. If the human evaluator fails to determine which one is human, and which one is a machine, based on the responses, then, the machine is said to have successfully fooled the evaluator and is deemed to be intelligent.
The Turing test is said to have set the benchmark for evaluating systems that may be intelligent.
The CEO of SoftBank, which, is a Japanese conglomerate, Masayoshi Son, once said, “I believe this artificial intelligence is going to be our partner. If we misuse it, it will be a risk. If we use it right, it will be our partner”. It is absolutely imperative that we tread carefully in matters pertaining to artificial intelligence. It is important to ensure that humans do not lose control over artificial intelligence. The result of losing control can be catastrophic. There is a need to impose regulations that prevent detrimental outcomes due to the development of artificial intelligence. The field of artificial intelligence has immense capabilities and should be allowed to grow, under regulated circumstances. Artificial intelligence need not have sinister outcomes. It is the responsibility of humanity to monitor A.I. and let it grow.
- A Technical Overview of AI & ML (NLP, Computer Vision, Reinforcement Learning) in 2018 & Trends for 2019
- Bill Gates: These breakthrough technologies are going to profoundly change the world