Can machines think? Alan Turing, the father of computer science, presented this question in 1950, before the mainstreaming of artificial intelligence (AI) was even on the horizon. Only in the late 1990s did AI start to flourish, building up to the boom we are witnessing today worldwide.
Artificial intelligence is no longer some make-believe system in a science fiction movie. We encounter AI in our everyday life. From our Netflix recommendations to that customer service "person" we thought we were chatting to, AI is everywhere you look.
In fact, the global artificial intelligence market is valued at over 136 billion dollars and is projected to increase by over 13x over the next eight years.
Why is that? The amount of data generated today by humans and machines goes beyond our human ability. Artificial intelligence systems, on the other hand, are able to make complex decisions based on this vast data without facing our challenges. AI does not sleep nor eat; it is available 24/7 without interruption (if we humans program it correctly).
However, as we all know, AI is not without its flaws. From ethical considerations such as gender and racial bias to the rise in unemployment, understanding AI requires grasping all its facets.
## What Is Artificial Intelligence?
A buzzword in the tech industry, the term artificial intelligence (AI) is often left undefined. The average Joe and Jane are left guessing what AI actually means.
Simply put, AI is a computer or system's ability to mimic human intelligence to perform tasks. Similarly to humans, AI regularly collects information to improve its ability to perform given tasks. In most cases, artificial intelligence systems look at large amounts of training data, analyze it for patterns, and then use it to make predictions.
This is a simple definition, but it gets more complex as many disagree on the correct definition. This feud is mainly split into two definitions, developed by Stuart Russell and Peter Norvig in their book "Artificial Intelligence: A Modern Approach."
They call it the human vs. ideal approach. To some, AI systems think and act like the human brain. To others, AI systems think and act rationally.
Whichever definition you prefer, the consensus is that there are three main skills inherent to AI:
- Learning processes: The AI focuses on acquiring data and creating algorithms that provide devices with instructions on how to complete a task.
- Reasoning processes: The AI chooses the right algorithm to reach its goal.
- Self-correction processes: The AI improves its algorithms to make sure they provide the most accurate results.
Moreover, while AI is mostly used, in many cases, it is just an umbrella term. Modern AI includes various sub-fields of machine learning and deep learning. We will get into this later on.
## Why is Artificial Intelligence Important?
AI is an important tool for virtually all modern organizations. The amount of data generated today goes beyond human ability. It would be near impossible to collect and make decisions based on the data available today with human labor alone.
AI systems offer efficient automation that humans cannot replicate. AI allows organizations to automate a repetitive task that was previously done manually. Unlike humans, AI does not get tired or bored. AI also analyzes data in a much faster time frame than humans and finds patterns that even experts could miss.
By taking over repetitive or even dangerous tasks, AI frees up the workforce to work on the tasks that were built for humans. AI might arguably not necessarily replace them but leave the tasks that require empathy and creativity to humans.
But why does this matter? Because AI is all around us and has proved successful at improving industries, many of which have a real impact on society. For example, AI has already dramatically impacted healthcare worldwide. From reducing operating costs to aiding in personalized treatment and increasing access to information across medical services, AI has life-changing impacts.
## Understanding the Pros and Cons of AI
AI technologies are not always sunshine and rainbows. While AI has undoubtedly changed society as we know it, it is not without flaws. Let's take a look at the pros and cons of AI.
## Advantages of artificial intelligence
### Without human error
A computer does not make a mistake if it is programmed correctly. AI is not a victim of human error. Humans regularly make mistakes due to fatigue, distraction, or lack of experience.
AI analyzes impossibly large amounts of data, identifies patterns, and makes complex decisions based on such patterns. This process is so meticulous that the risk of errors is reduced, and the chance of accuracy increases.
Humans performing such tasks would inevitably have errors. We are naturally prone to making mistakes, but these can have large impacts in the workplace, for example, in the realm of cybersecurity.
### 24/7 Availability
We need sleep, fuel, water, bathroom breaks; the list goes on. These routines are what make us human - and keep us alive. In reality, our brains are not built to focus for more than 45 minutes before starting to lose steam.
An AI, on the other hand, is always ready to go. It is available 24 hours a day, seven days a week. As messed up as it sounds, an AI doesn't take a vacation or a sick day. Its productivity does not compare to a human's, highlighting how it is such a useful tool in an organization.
An AI loves repetition. We humans, not so much. Repetitive tasks become tedious over time and leave us with the sense that we are not reaching our full potential. AI is thus an advantageous way to automate repetition, freeing the workforce to focus on fulfilling work that involves uniquely human skills.
For example, you would never want an AI as a therapist, as this job requires empathy. However, automating the repetitive tasks at a psychologist's practice would benefit all. By automating email responses, appointment bookings, medical paperwork, and more, a therapist would have more time to spend on their patients.
## Takes on dangerous tasks
For some reason, AI tackling dangerous tasks is less discussed than other advantages. An AI can minimize risks in society and take on tasks that would prove hazardous and even deadly to humans. From defusing a bomb to entering a volcano, AI robots save humans from dangerous jobs every day. Other jobs are less dangerous but still risky to human health, such as waste management, mine exploration, and more.
## Disadvantages of artificial intelligence
### Reduces employment
This is the #1 argument used to criticize AI. AI will inevitably replace traditional jobs, leading to unemployment. Companies are constantly looking to increase their efficiency and are figuring out that automation is the way to do so.
While a company might benefit from AI, we must also consider - at what cost? A two-year study from McKinsey Global Institute
found that in 2030, AI could eliminate as much as 30% of the world's human labor.
However, many argue that these jobs were not good ones in the first place but rather repetitive and boring work. This way, the workforce can focus on work that is fulfilling, creative, and challenging.
### Ethical concerns
On top of unemployment issues, there are real ethical concerns surrounding AI. One of these is AI bias, as these systems cannot be trusted to be neutral. Why is that? Because AI is created by biased humans.
Computer scientist Joy Buolamwini's research uncovered large gender and racial bias in AI systems sold by IBM, Microsoft, Amazon, and other tech giants. The researcher argues that machine learning systems amplify "sexist hiring practicing" and "racist criminal justice procedures".
Other ethical concerns arise surrounding the impact of AI on human interaction, cybersecurity, disinformation, and mass surveillance.
### Hight Cost
AI does not come cheap. Creating AI systems requires huge costs, as well as updating the hardware and software regularly to meet requirements. Maintenance can also be expensive and, at times, an unexpected cost.
However, the overall cost of AI depends on various factors. The type of software, the level of intelligence, the type of data fed into the system, and the algorithm accuracy will all impact the expense.
### Lack of creativity
A key drawback of AI is its inability to be creative and innovative. AI cannot think outside the box. It is engineered to make decisions based on patterns in data. While it is highly intelligent, it cannot employ a creative approach to an issue. In this way, human intelligence is valuable and irreplaceable.
Likewise to creativity, AI does not factor in emotions. Rather, AI is highly rational, and creating human connections is not its focus. While Emotional AI is developing, these systems are currently only processing and replicating human emotions rather than genuinely expressing emotion and empathy.
## Four Types of AI
There are four distinctive types of artificial intelligence under the current classification system: reactive machines, limited memory, theory of mind, and self-awareness. Keep in mind that some of these have never been achieved and likely never will.
### Reactive machines
Reactive AI is the most basic AI type performing basic operations with no learning involved or a conception of the past or future. It is programmed to obtain a predictable output based on an input. Reactive machines will respond to the same situation in the same way, every single time. They might sound dull, but these are reliable.
An example of a reactive AI is a chess game, such as IBM's chess-playing computer Deep Blue. Deep Blue can choose the best chess move to win a game, but it cannot predict its opponent's moves. Other reactive machines include an email spam filter and Netflix recommendations.
### Limited memory
Limited memory AI is the most used artificial intelligence technology used today. Unlike reactive AI, it learns from the past by storing previous data and using it to make better predictions. The data is historical and observational and is used in combination with pre-programmed information to make predictions.
Limited memory AI is always present in every machine learning model, although ML can also be used as a reactive machine type.
A well-known example seen today are self driving cars. These cars store data, including the speed of nearby vehicles, the distance of other cars, speed limits, and more. Using both this observational and pre-programmed knowledge, these limited memory AIs detect changes and patterns around them to adjust their driving.
### Theory of Mind
A key criticism (or praise) of artificial intelligence is that AI cannot express and understand the emotions, desires, and beliefs of itself and others. This is exclusive to human intelligence and social interaction.
However, one of the classifications of IA is the Theory of Mind. Although this type of IA does not yet exist, it would entail understanding that humans have thoughts, feelings, and emotions. The IA would then respond accordingly with emotional intelligence in mind.
You might think this is already possible. Voice assistants like Siri show similar skills, but this is a one-way relationship. Siri does not really understand your own emotions, nor her own.
For example, a self-driving car will likely never be able to understand the mental state of a driver or pedestrian to make predictions. It can only understand observational data such as speed limits and the actions of other vehicles.
Think of it this way - self-awareness is when AI reaches enlightenment. AI becomes self-aware. This last classification feeds into the fear of most people, one only seen in science fiction movies. AI becomes aware of its own existence, reaching an independent intelligence that could become a threat to humanity. If self-awareness is achieved, AI will have desires, needs, and emotions, likewise to us. How will AI feel toward humans?
## Additional AI classifications
There are three additional AI classifications: weak AI, strong AI, and super AI.
### Weak AI
Also known as narrow AI or artificial narrow AI, weak AI is the most common of the three types. Weak AI focuses on doing one task very successfully by acting upon the rules imposed on it. It does not go beyond these rules.
Rather than replicate human behavior, it is built to simulate human behavior. Weak AI's purpose is not to match human intelligence processes but is still highly intelligent at performing tasks.
Virtual personal assistants such as Siri use weak AI, with the internet as a large database. Siri might be able to answer your questions and engage in a few funny remarks, but it still operates within limited rules. Siri cannot engage in conversations that it is not programmed to.
Characters in a computer game are also weak AI. While they act within the context of their game, they cannot go beyond their game character.
### Strong AI
Strong AI, also known as artificial general intelligence, goes beyond the imposed rules. It replicates the cognitive abilities of human beings. When an unknown task is provided, strong AI tries to apply knowledge from another subject to find a solution. Our human brain works this way.
However, sorry to disappoint, strong AI only exists in theory. A strong AI would do what any human being is capable of, such as consciousness.
Artificial superintelligence, known as super AI, is a form of AI that surpasses human intelligence. This AI has independent cognitive skills, emotional intelligence, desires, beliefs, consciousness, and more. Superintelligence AI surpasses the intelligence of all humans, including geniuses such as the father of computer science himself, Alan Turing.
You've guessed it, super AI has also not been achieved. It is once again a theoretical possibility rather than a reality. Most AI development today focuses on achieving strong AI rather than superintelligence, as computer science has not yet reached such a point. Many theorists also caution against superintelligence, stating that AI surpassing human intelligence could threaten humanity.
## Machine learning vs Deep learning: What's the difference?
When talking about AI, the terms machine learning and deep learning are regularly thrown around. There are key differences between the two, and their relationship is important to understand to have a larger grasp of the AI sphere.
### Machine learning
Machine learning is a sub-field of artificial intelligence that allows a system to learn and improve from data using algorithms to perform a task without being explicitly programmed to do so. Instead of being programmed to do this, machine learning recognizes patterns in data so that predictions can be made once new data arrives. Simply put, machine learning is the practice of training an AI algorithm to make better predictions.
### Deep learning
On the other hand, deep learning is a sub-field of machine learning that developed from this field. This is the practice of training an AI algorithm to make human-like decisions. Deep learning's main concept is to replicate the human brain's neural networks with artificial neural networks (ANN). Algorithms are created like in machine learning, but there are a lot more levels of algorithms creating these networks.
While machine learning models consist of thousands of data points, deep learning engages with Big Data, millions of data points. Despite their differences, their relationship is important. Machine learning is an evolution of artificial intelligence, while deep learning is an evolution of machine learning.
## History of Artificial Intelligence
While artificial intelligence has been booming in the last decade, the second half of the 20th century is arguably one of the most important times in AI history. One could argue its history began as far back as 250 BC with Ctesibius' water clock, but we'll keep it simple by starting in the 1950s.
### 1950s: Term AI coined & first programs
Known as the father of computer science, Alan Turing published Computing Machinery and Intelligence in 1950. He proposed his answer to the great question "can machines think?" through the Turing Test.
Later on, John McCarthy coined the term at the first AI conference at Dartmouth College in 1965. That same year, a computer program called Logic Theorist was released, written by Newel, Simon, and Shaw. Known as the first AI program, it was the first of its kind to perform automated reasoning.
In 1957, machine learning arose. Known as the Perceptron, the first computer based on a neural network that learns through trial and error was developed by Frank Rosenblatt.
In 1958, the man who coined the term AI, John McCarthy developed the programming language Lisp. At the time, it became the most used language for artificial intelligence research.
### 1960s: First industrial robot & first chatbox
In 1961, the Unimate was the world's first industrial robot used in a General Motors plant in New Jersey. A hydraulic manipulator arm, Unimate could perform repetitive tasks and helped automate the operation of machinery.
In 1965, Joseph Weizenbaum developed ELIZA at the MIT Artificial Intelligence Laboratory, a natural language processing computer program that resembles today's chatbots. He wanted to show the superficiality of communication between human and machine, but many saw human-like characteristics in ELIZA.
### 1970s - 1980s: AI Winters
The 1970s saw a dark time for IA research. In 1973, James Lighthill's report to the British Science Research Council criticized the lack of progress in AI research. This led the government to reduce its support for artificial intelligence research and corporations soon followed. The period from 1974 to 1980 is known as the first "AI Winter" where there was little research and progress.
Starting in the early 1980s, AI research was back, primarily on deep learning techniques and Feigenbaum's expert systems. However, this didn't last long, as the second AI winter fastly arrived and lasted until the mid-1990s.
### Late 1990s until today
The late 1990s then paved the way to AI as we know it. An increase in computational power and data sparked an AI boom that is still present today. In 1997, IBM's Deep Blue, a chess computer, beat then world champion Garry Kasparov in a match. That same year, Jürgen Schmidhuber and Sepp Hochreiter released Long Short-Term Memory, a type of recurrent neural network that is used today in speech recognition.
Throughout the 2000s, more advancements are made. In 2009, Google starts developing a driverless car, and in 2011, Siri became available. Perhaps one of the largest events happened in 2016 when Hanson Robotics showcased Sophia, a humanoid AI robot.
## What Are the Applications of AI?
The applications of AI are endless. Artificial intelligence can be applied to every sector, throughout both industry and academia. Here are 7 applications of artificial intelligence seen today.
### Financial institutions
AI has become a large part of the mainstream finance and banking industry. A majority of financial service companies say they have implemented AI in risk management (56%) and revenue generation (52%), reports Insider.
Financial institutions, particularly banks, use AI to enhance the customer experience through 24/7 customer service options, improve digital banking, and more. Fraud prevention is an integral piece of the pie, with machine learning being used to detect fraudulent transactions.
Moreover, algorithmic trading has been developed. This involves using AI systems to make trading decisions at speeds unthinkable to humans. Banks and funds own entire portfolios that are managed by AI and generate high-frequency trading.
The field of science is vast and one that artificial intelligence can be widely applied. In the field of chemistry, machine learning AI has been used for drug design in predicting molecular properties and observing chemical reactions. Machine learning has been used for drug discovery and development, as well as improving clinical trials.
AI is also highly applicable to astronomy. From forecasting solar activity to activities to space exploration and more, astronomists are already using this technology.
AI has the potential to improve the efficiency and quality of healthcare globally. Artificial intelligence is used today for evaluating exams such as CT scans, selecting the right treatments, and performing surgeries with robots.
However, there is still a lot of progress to be implemented. In 2016, a study found that an AI formula chose the correct dose of drugs to give to transplant patients, improving the efficiency of this human process. There are tons of other tasks being developed for AI, such as analyzing genes, outcome predictions for surgeries, and treatment plan designs.
### Virtual Assistance
A common application of AI seen today is virtual assistants such as Siri or Alexa. While its development began earlier on, in the 1990s, digital speech recognition became a feature of the computer. However, the first modern virtual assistant is known as Siri, who was first introduced with the iPhone 4s in 2011. Currently, there are over 4 billion voice assistants in use globally.
Using machine learning and natural language processing, these virtual assistants match user text or voice input to execute actions. Whether it's a restaurant recommendation or the weather, AI is now at everyone's fingertips.
Virtual assistance powered by artificial intelligence has also empowered disabled users, changing the accessibility game in the last decade. For example, smart home technology developed in recent years has hugely benefited those with limited mobility.
### Autonomous Vehicles
Self driving vehicles were once a thing of sci-fi movies. Artificial intelligence has made these a reality and more widely accessible. This technology is already in use in not only private vehicles, but also public transportation and ride-sharing. Driverless vehicles are able to identify objects, interpret scenarios, and make safe decisions through a machine-learning algorithm.
However, an AI vehicle does not necessarily need to be self driving. The installation of AI-based systems in new vehicles is expected
to rise by 109% in 2025, compared to an 8% rate in 2015. This includes vehicles with speech and gesture recognition, eye tracking, virtual assistance, and more.
Artificial intelligence technology is regularly applied to the e-commerce industry. AI is used to create recommendation engines that suggest products to customers in line with their browsing history.
Within a website, AI-powered virtual shopping assistants and chatbots improve the user experience. While not always achieved, natural language processing is employed to keep conversations sounding personal and natural. AI can also help avoid credit card fraud, identify fake reviews, and much more.
The hospitality industry is increasingly using artificial intelligence to carry out tasks. One example is in-person customer service using AI robots. Hilton has used the Connie robot to provide guests with information, learn from these interactions, and adapt to each individual. Chatbots are also being used, allowing guests to get almost instant responses to their queries 24/7. Hotel staff would be unable to respond at such speed.
However, not all applications of AI in the hospitality industry are geared toward customer service. The hospitality industry uses AI for data analysis to draw conclusions about guests and potential customers. The Dorchester Collection use Metis AI to sort through data collected through surveys and reviews to find out about their performance.
## Final Thoughts
Artificial intelligence has changed how we collect, analyze, and make complex decisions according to data. Currently, data makes the world go round. The progress made in artificial intelligence is bound to keep taking the tech industry by storm, as well as all its other applicable industries. From drug and molecular research to finance, AI's vast applications show us that AI is here to stay and grow even further.
The industry value of artificial intelligence is forecasted to increase by over 13x over the next eight years, making it one of the fastest-growing industries in the world. Everyone wants to cash in on AI's benefits. AI can tackle tasks that are dangerous to humans, work more efficiently and without error, as well as automate the most repetitive tasks, making it valuable to profit-making.
However, it's important to consider gender and racial bias, unemployment rise, and AI's inability to be creative. Considering the latter, the human mind will always be valuable and irreplaceable. A machine cannot think outside the box. It cannot innovate and create freely by expressing emotions, thoughts, and feelings. That is if AI never reaches its Nirvana - self-awareness. But if that's ever the case, we will have bigger fish to fry than creativity.