An Ultimate Guide to Artificial Intelligence

A Brief Introduction into Artificial Intelligence

Artificial Intelligence, or as it is more commonly referred to, AI, refers to machine intelligence or intelligence which is demonstrated by machines which is unlike that displayed by humans or animals.

The field of study is often referred to as that of intelligent agents, which can be further defined as any device which has the ability to perceive its environment and is able to take actions which maximize the chances of successfully achieving the goals that have been set.

History

Artificial Intelligence has made most of its appearances in history through storytelling devices in antiquity, especially in fiction works such as Frankenstein by Mary Shelley and R.U.R (Rossum’s Universal Robots) by Karel Čapek.

Research in AI started from as early as 1956 when the term “Artificial Intelligence” was coined by John McCarthy at a workshop in Dartmouth College.

He used the term to distinguish the field from cybernetics and to subsequently escape from Norbert Wiener, a cyberneticist. AI research was born from the brilliant minds of McCarthy along with Allen Newell, Herbert Simon, Marvin Minsky, and Arthur Samuel.

It started with a simple concept where computers were able to solve algebra problems and learning to speak English, along with starting to learn different checkers strategies which resulted in the computers playing better than the average human by 1959.

The United States Department of Defence heavily funded the research by the middle of the 1960s and laboratories had been established right around the world by this time as well.

AI’s founders were extremely optimistic about the future thereof and predicted that machines would be able to do any work that man can do within 20-years, although there were some difficulties and obstacles that could not have been foreseen.

There was a substantial decrease in progress during the 1974’s due to criticism and continuous pressure from the US Congress which led to the halting of exploratory research in AI, due to a decision by both the US and British Governments.

In the early 1980s, what was known as the “AI Winter” was ended when AI achieved commercial success through expert systems which is a form of AI program which stimulated both the knowledge and analytical skills of human experts.

The field of AI expanded and grew, and by the 1990s and early 21st century, AI was used for logistics, data mining, medical diagnosis, and an array of other areas.

By 2015, AI reached what was referred to as a landmark year when there was an increase in software projects which utilize AI within Google from merely sporadic usage in 2012 to over 2,700 projects.

There is still a substantial amount of research and experimentation ongoing in the field of AI and a lot of observers believe that with the increasing research output in China, it may be leading to what is referred to as an “AI Superpower”.

A Basic Understanding of Artificial Intelligence

As previously mentioned, AI is the study of intelligent agents and these agents can be defined as any device which has the ability to perceive the environment that it is in and can take actions to maximize the chances of successfully achieving its goals.

Another definition also characterizes AI as the ability of a system to interpret external data correctly, to be able to learn from such data, and to use what it has learned to achieve specific goals and tasks through being able to adapt flexibly.

Algorithms are what is behind AI as it is a set of unambiguous instructions which can be executed by a mechanical computer. These algorithms allow an AI to learn from data and subsequently enhance themselves by learning new strategies, or “rules of thumb”.

These learning algorithms are based on the belief that strategies, algorithms, and inferences which have worked well in the past, are likely to continue working in the future.

There is, however, one factor which has caused a lot of controversy towards AI, and that is that when compared to humans, AI lacks several features of common-sense reasoning which humans possess.

As humans, we can reason about space, time, and physical interactions, which is referred to as “naïve physics”. Humans can anticipate the consequences of actions or interpret folk psychology, or “common language” which AI’s have great difficulty with.

The challenges in Artificial Intelligence

The concept of Artificial Intelligence offers a vast amount of opportunities in numerous fields, but it is imperative to consider the challenges that this field of study and research is faced with.

When holistically viewed, the most common challenges faced include, but is not limited to:

  • Reasoning and problem solving
  • Knowledge representation
  • Planning
  • Learning
  • Natural Language processing

1.   Reasoning and problem solving

Algorithms were developed to imitate the step-by-step reasoning as used by humans where puzzle-solving and logical deductions were concerned.

Between the late 1980s and 1990s AI research had developed methods in dealing with uncertain or complete information by employing concepts based on probability and economics.

However, these algorithms were proven to be quite insufficient where the solving of large reasoning problems were concerned as the AI became slower as the problems grew larger.

2.   Knowledge representation

This is very classic where AI is concerned when gathering both explicit and common-sense knowledge regarding objects, properties, categories and the relation between objects, situations, events, time, and more.

Where the representation of knowledge is concerned, AI often struggle with the following:

  • Default reasoning as well as the qualification problem
  • The breadth of common-sense knowledge, and
  • The sub-symbolic form of common-sense knowledge

3.   Planning

For an intelligent agent to function properly and be classified as AI, it must be able to set goals and achieve them. This includes being able to visualize the future and make subsequent choices which will maximize the utility of choices available.

This has led to the development of multi-agent planning and it makes use of the cooperation and the competition of numerous agents to achieve a specific goal.

As result, emergent behaviour came into being, which is used by evolutionary algorithms and swarm intelligence.

4.   Learning

As a fundamental concept of AI, machine learning has been one of the foundations upon which AI was developed as it involves the study of computer algorithms which can improve automatically through experience.

There are two types of learning concerned, unsupervised learning and supervised learning. Unsupervised learning allows for an AI to find patterns in a stream of input without the inputs being labelled by a human first.

Supervised learning requires both classification and numerical regression which subsequently requires that a human input the data first.

In addition, there is computational learning which allows for the assessment of learners by computational complexity, sample complexity, or any other notions concerned with optimization.

Lastly, reinforcement learning allows for the agent to be rewarded should a good response be provided and punished should there be a bad response given. Through this, the agent can form a strategy for operating within its own problem space.

5.   Natural language processing

This allows a machine to read and understand human language. It enables natural-language user interfaces along with the acquisition of knowledge derived directly from sources that have been written by humans.

Some of these sources include:

  • Information retrieval
  • Text mining
  • Question answering, and
  • Machine translation.

Modern and current approaches to processing natural language involve word concurrence frequencies which allow for the construction of syntactic representations of text. There have been numerous strategies employed with combinations between several and they have proven to achieve acceptable accuracy.

The goal of natural language processing of AI is to embody a full understanding of common-sense reasoning and recently, in 2019, a transformer-based deep learning architecture was able to generate coherent text which paves the way forward in dealing with this challenge.

Approaches

There is a lot of controversy involved with AI research and it is faced with numerous long-standing questions which have not yet been answered.

Some include whether Artificial Intelligence should stimulate that of natural intelligence by allowing machines to study psychology or neurobiology, whether intelligent behaviour could be described in suing simple yet elegant principles, and many more.

Through the years, there have been numerous approaches in AI, some of which include:

  • Cybernetics and brain stimulation – which involves the connection between neurobiology, information theory, and cybernetics.
  • Symbolic – which involved the exploration of the possibility that human intelligence could be reduced to merely symbol manipulation and included cognitive simulation, logic-based, anti-logic, and knowledge-based symbol manipulation.
  • Sub-symbolic which included embodied intelligence and computational intelligence as well as soft computing.
  • Statistical approaches which adopted sophisticated mathematical tools including hidden Markov models, or HMM, information theory and normative Bayesian decision theory which provided high levels of accuracy in data mining.
  • The integration of approaches such as intelligent agent paradigm, agent architectures, and cognitive architectures.

Applications of AI

AI has relevance towards any intellectual task with modern Artificial Intelligence techniques substantially pervasive. Some examples of AI include, but are not limited to, the following:

  • Manufacturing robots
  • Smart assistants
  • Proactive health care management
  • Disease mapping
  • Automated financial investing
  • Social media monitoring
  • Conversational marketing, and numerous others.

Without realizing it, Artificial Intelligence forms an important part of the world that we live in and it plays an active role in our daily lives. Something as simple as opening Facebook to read a newsfeed or performing a Google search involves AI in some way.

The breadth of AI’s applications is spread across numerous industries with endless opportunities in the research and development of such as technological innovations improve and large strides are taken in integrating AI more into daily life.

Some examples which display the very breadth of AI technology include:

  • Home robots such as robotic vacuums.
  • Humanoid robots for both commercial and consumer markets.
  • AI Assistants such as Olly, Alexa, and others.
  • PathAI which provides pathologists with AI-powered technology.
  • Proactive healthcare management through Pager, which helps patients with minor aches, pains, and illnesses.
  • Atomwise, which makes use of AI along with deep learning to facilitate the discovery of drugs.
  • AI-powered machines which detect diseases, can form, and provide a diagnosis, provide treatment, and manage processes.
  • Robo-advisors that use algorithms to automate tax loss harvesting, trade, transaction and manage portfolios.
  • AI-powered financial search engines to help investment firms.
  • AI-powered crowdsourced hedge funding.
  • Google Smart Maps.
  • Travel assistants which can provide booking prices for flights, hotels, excursions, and more.
  • Image recognition breakthroughs such as used by Facebook.
  • The monitoring and categorizing of video feeds such as on Twitter, YouTube, and other social platforms.
  • Applications as used in e-commerce by Amazon and Twiggle, which makes use of NLP for e-commerce searches.
  • Uses in building customer relationships by making use of marketing tools along with using it in conversational marketing.

Final Thoughts – in which way will AI change the world that we live in

Great strides have been made since Artificial Intelligence was merely a theory and a concept and it is currently impacting the future of virtually every industry in the world along with the lives of nearly every human on earth.

It has acted as the main driver in emerging technologies and it is set to act as the technical innovator for the foreseeable future as more technologies are developed, launched, improved, and adapted.

AI is present in transportation, manufacturing, healthcare, education, media, and customer services, and the current innovations are merely the beginning with numerous others that will emerge in time.

There are substantial amounts of funding that go into AI research annually, especially when considering technology giants such as Google, Apple, Microsoft, and Amazon, amidst a few who spend billions of dollars in creating products and services that are based on AI.

It is also important to consider society, and the impact that AI has when considering employment and the possibility of having more robotics in the workplace performing work at a faster, more accurate, and more productive and profitable level than humans can.

As the world steadily moves further into a digital era, the future of AI remains hopeful, vast, and filled with possibilities despite the controversy, and moving into such a future and co-existing with robots seems a long way off, but it also seems inevitable.

Big Data

Business Intelligence

Data Automation

data-automation

Machine Learning