Artificial Intelligence - What's the Hype and Should You Care?
I am sure we have all heard of the word 'Artificial Intelligence' at least once in this very month. So, in this post I will briefly touch upon what it is, its types, applications and future, which will help you decide whether or not this topic is something you want to learn more about in the future.
Table of Contents:
What is Artificial Intelligence?
I imagine a world in which AI is going to make us work more productively, live longer and have clean energy.
- Fei-Fei Li (Computer Science Professor at Stanford)
Artificial Intelligence is an accelerating field of Computer Science concerned with producing artificially intelligent computers, i.e. machines that exhibit intelligence or sense just like a human being. The goal is to produce machines which are independently capable of performing tasks which would otherwise require human supervision or expertise. Currently, this is essentially done using data sets, computational structures and algorithms which allow computers to perform tasks over-time without explicitly being programmed to do so. I'm sure whenever we all hear this term we picture C3PO from Star Wars, and funnily enough that is exactly what Artificial Intelligence is trying to achieve.
The Purpose of Artificial Intelligence
I am sure many of us at some point in our lives thought about what would happen if Robots took over the world or something. If that's you, then you probably don't find Artificial Intelligence a very wise discipline to pursue. It is for that reason, that it is essential for us to identify why Artificial Intelligence is crucial in today's world.
The main goal of Artificial Intelligence is to aid human beings in complex decision making and in optimise survival conditions in the most efficient yet interdependent manner. It also seeks to reduce human effort required in tasks and allow us to divert our focus on other more meaningful problems. Speculation also suggests that an invention like this will allow for a more equal, free, united and liberal society where human suffering is at its minimum.
Brief History of Artificial Intelligence
Before we dive deeper into the field, it might be ideal for us to discover its roots. Although there had been theorised as early as the Greeks and the Egyptian civilisation, this term was first officially used to specify an academic field in 1956 by scientist John McCarthy, in at a conference in Dartmouth College.
The idea first came around that Artificial Intelligence was any machine that could pass the Turing Test. This was a test designed by a British Mathematician Alan Turing who, in his paper 'Computing Machinery and Intelligence' suggested the development of machines that mimic a human being's reasoning ability. Conversely, he also proposed this conjecture as basis to distinguish between AI and Human Intelligence.
The very first instance of artificially intelligent was presented in the aforementioned Dartmouth Summer Research Conference on Artificial Intelligence, where researchers from far and wide gathered to admire the Logic Theorist program(developed by Alan Newell, Cliff Shaw, and Herbert Simon), which is said to be the first attempt at replicating the human decision making system in software.
And from that years onwards, the field Artificial Intelligence blew wide open with extensive research and investment into Computer Science. The search for faster and higher-capacity storage was fuelled by the requirement for efficiency in pursuing AI and Machine Learning. These Initial stages of the field experienced some hinderances, with phases where progress was stagnant due to criticism and lack of advanced equipment. It even underwent stages which are referred to as the 'Artificial Intelligence Winters'. Nevertheless, record-breaking research into Artificial Intelligence has managed continue into today's world.
Present-Day Artificial Intelligence
It would be an understatement to say that Artificial Intelligence has come a long way from the Dartmouth Summer Conference. With the invention of efficient supercomputers, advanced algorithms, and computational techniques, not only computer science professionals, but even businesses from other fields are vastly inclining toward Machine Learning - the most commonly-used form of AI in today's time. Big multi-national corporations like Google, Netflix, Amazon are all making use of ML algorithms to predict their users choices and provide recommendations.
Speech Recognition has also seen a sudden boom with virtual assistants like Google Assistant and Siri. Companies like IBM have also seen very recent breakthroughs in improving this technology. This concept has even been combined with Robotics to create appliances like Alexa. Advancements in Robotics like Paro (therapeutic robot designed to lower stress levels in patients), Pepper, and Saul Robot (robot designed to fight the Ebola virus) have also proven the utility of Artificial Intelligence in general healthcare.
Presently one of the biggest challenges which researchers are trying to overcome in terms of Artificial Intelligence, is its inability to comprehend deep human emotions and non-verbal communicational conventions. As our dependence on AI increase exponentially every single day, so does the unpredictability associated with its future. Here's what Stephen Hawking has to say about the future of AI, "AI is likely to be either the best or worst thing to happen to humanity," and we couldn't agree more.
Types of Artificial Intelligence
By far the greatest danger of Artificial Intelligence is that people conclude too early that they understand it.
- Eliezer Yudkowsky
There are two distinct divisions of Artificial intelligence: Type-1 and Type-2. Broadly categorised on the basis of their ability to mimic human abilities and their level of capability, there are three types of Artificial Intelligence as per Type-1:
Narrow Artificial Intelligence
This is the most basic form of Artificial Intelligence, in which the machine or computer specialises in performing only one task and constantly keeps improving at it. This is the form of AI that humanity is currently capable of producing. Such AI, also known as Weak AI, is incapable of performing tasks beyond its predetermined boundaries. This type is also termed as the 'learning' phase of Artificial Intelligence. An example could be the GO robot which is designed only to play and improve at the game of Chess, or an automated driverless car.
General Artificial intelligence
This form of Artificial Intelligence is multi-purpose in terms of capability and can successfully perform all the tasks that human beings can, replicating them exactly. Such systems do not yet exist, however extensive research and investment is being dedicated to its discovery. The closest that humans have come to creating General AI is Fujitsu's K supercomputer which managed to stimulate human neural activity for an entire second. This type is termed as the 'intelligence' phase of Artificial Intelligence.
Super Artificial Intelligence
At this level, Artificial Intelligence will be capable of outdo human beings and surpass their level of intelligence. Such AI can perform all tasks that human beings can, but perform them better and more efficiently, including making decisions, making judgements, planning, etc. This is termed as the 'conscious' phase of Artificial Intelligence. This is the form of AI which hypothetically has the capability for a Matrix or Terminator like invasion of the world. But don't lose any sleep over this though just yet, we are still decades away from reaching this stage, what with this still being nothing more than a conjecture.
However, there is another categorisation of Artificial Intelligence referred to as Type-2, which divides it into four based on purpose and functionality:
Reactive Machines
As the name suggests, this type of Artificial Intelligence has the ability only to react to its stimuli, and does not store memories or learn for improvement. These machines are also programmed to respond to a set of predetermined input, and are thus suitable for scenarios where all input parameters are known to us. This is the most basic form of Artificial Intelligence and it thus has limited utility. An existing example of this is Google's AlphaGo which uses Neural Networks and has managed to beat top human Go experts.
Limited Memory
Such Artificial Intelligence is capable of storing memories and learn from experience. This form of AI experiences stimuli, collects data, analyses it and then makes decisions and improves its performance based on it. In a way, these machines have the ability to 'remember' things. Most of Machine Learning and Deep Learning products fall under this category, where they are fed large amounts of data in order to train them to excel in a particular task. Virtual Assistants and Automated cars are a great example of this form of Artificial Intelligence.
Theory of Mind
This refers to Artificial Intelligence which is capable of understanding human emotions and social norms. Such understanding also allows this form of AI to communicate in a way almost identical to humans. This form of AI is still just a conjecture, and we still have a long way to go before we can reach this stage.
Self Awareness
As the name suggests this type of Artificial Intelligence refers to the AI which aware of itself and has its own consciousness. The closest replication of AI to an actual human brain, including our ability to reason, remember, communicate and feel things limitlessly. This corresponds to Super AI form the Type-1 classification, and so this also purely hypothetical at the moment.
Branches of Artificial Intelligence
There are different branches of Artificial Intelligence which humanity is currently pursuing, and these can broadly be categorised into 6 fields:
Machine Learning
As introduced briefly before, Machine Learning is the scientific discipline which makes of data analysis and algorithmic computation as a way to make computers learn from experience and existing information.
This concept relies on computers interpreting solutions, making decisions and predictions based on past responses and pre-provided information by the creator. This knowledge also allows these machines to improve efficiency with each cycle. This technology is used widely in softwares to provide user recommendations and improve search results. It also has applications in Finance to analyse data and make predictions about future trends.
Machine Learning is further divided into three types, depending how the ML algorithm is trained to perform its job:
Supervised Learning: Where algorithms learn based off previously identified and labelled data.
Unsupervised Learning: Where algorithms learn based off abstract data, make their own connections, and identify hidden relationships.
Reinforcement Learning: Where algorithms learn based off past mistakes through a trial-and-error method.
Expert Systems
This branch of AI focuses on developing machines or computers which withhold complex problem solving and decision making ability. Such machines are focused only on one particular field or domain, and are meant to exhibit extreme human-level expertise on the matter. These systems are provided with something called a 'knowledge base' which contains a set of information, and instructions or rules. This knowledge acts like experience for the system to learn from.
The first Expert System ever was developed by researchers at the Stanford University, called Pathfinder - a system developed for the diagnosis of lymph-node diseases. In fact, his technology has managed even to beat the world's finest pathologist when it comes to symptoms identification and effective diagnosis.
Neural Networks
As the name suggests, this field of Artificial Intelligence focuses on the Cognitive side of things and aims to artificially replicate human neurons and neural networks into a machine to stimulate human-like thinking. Such form of Artificial Intelligence depends on making connections between different types of information and parameters to reach to conclusions, just like a human brain does. There are many types of neural networks, the most common ones including:
Feedforward Neural Networks (also known as Artificial Neural Networks | ANN)
Recurrent Neural Networks (RNN)
Convolutional Neural Networks (CNN)
Neural Networks are also referred to as Deep Learning. This technology is used widely in facial recognition software, such as that which is used in mobile phone security systems these days. Sales and Marketing is another field which extensively utilises this form of AI, in order to make predictions about the possible customer based on the demographic and financial information provided to it.
Robotics
Robotics is the interdisciplinary field of Engineering and Computer Science, which aims on transferring Artificial Intelligence into physical machines to increase practical output. As is common knowledge, the main goal of Robotics is to reduce human effort and manual labour required in tedious and physical tasks by developing machines who can perform these jobs instead. Thus Robotics focuses on designing artificially-intelligent machinery which can endlessly perform its job by determining the most-efficient and profitable method of doing it.
Assembly lines in many industries these days use Robots, especially heavy equipment like Automobiles. There have also been many domestic applications of Robotics, such as the Roomba Vacuum cleaning Robot, which is capable of navigating through the house and vacuuming in places where it deems necessary.
Fuzzy Logic
Fuzzy Logic focuses on determining the extent to which a conjecture or a hypothesis is correct. It aims to solve the problems of uncertainty associated with particular situations and information, by assigning a numerical value to the 'amount' of truth to any particular logic between 0 and 1, 0 being completely false and 1 begin completely true. As you might be able to tell, the concept works similar to mathematical probability. This allows for more informed and accurate decision making, while reducing chances of consequence.
Fuzzy Logic has many practical applications in today's world. It can be thought of as a multidimensional approach to regular Boolean Logic used in programming. One example could be its usage in Elevator controls, where Fuzzy Logic is used to reduce the waiting time for elevator arrival depending on the number of passengers.
Natural Language Processing
Natural Language processing is the field of Artificial Intelligence which focuses on improving the communication between human beings and computers. It is a collaboration between linguistics, mathematics and computer science. This field includes computers interpreting human communication and comprehending it with accuracy. The goal of this field is to make natural languages which are used by human beings, understandable to computers which only work with zeroes and ones.
The natural language can be in written or spoken form. Thus, spell check is a great example of Natural Language Processing used in daily lives, where computer understands human input. Chatbots and Spam filters are also very popular applications of Natural Language Processing. Understanding sound input however is slightly more challenging, what with the computer having to convert the sound into text first.
Applications of Artificial Intelligence
Even at such a young stage, artificial Intelligence has widespread existing applications in almost every field imaginable, whether it is Healthcare, Construction, Gaming, etc. Below is a brief summary of these.
Healthcare
AI has numerous applications in healthcare. Whether it is Diagnosis, Treatment or Post-Surgical Care, Artificial Intelligence is beneficial in each of these tasks:
More accurate and efficient diagnosis: An example of this is PathAI, a US-based start-up aiming toward more accurate cancer diagnosis.
Robotic and automated surgery: An example of this is Heartlander, a mini-robot built by the Carnegie-Mellon University for heart surgeries.
Post-treatment care and support: An example of this is CareAngel, which is a virtual nursing assistant which provides 24/7 support for patients.
Drug Development: AI helps in research and the discovery of new drugs by accelerating the compound-testing process and aiding in predicting the protein structures of pathogens.
E-Commerce
According to Servion Global Solutions, by 2025, 95% of customer interactions will be powered by AI. This is a clear indication of the growing requirement of AI in the E-Commerce industry. Some examples include:
Customer Service and Support: Artificially Intelligent Chatbots are become increasingly popular these days with almost all world-renowned companies - Amazon, Sephora, Domino's, Nike, and many more - using them.
Personalised Recommendations: Machine Learning algorithms are new being excessively used to make recommendations for customers based on their past purchases. Companies like Amazon use this practice.
Security: Especially when it comes to E-Commerce, AI is used to authenticate users and protect their privacy, in order to reduce the risks of getting defrauded and hacked.
Advanced Search: AI has helped improve search algorithms by personalising the search keywords, enable Image search, where the product can be found with merely a picture, and improve voice search (Natural Language Processing).
Entertainment
The Entertainment industry is one of the most benefitted by the discovery of Artificial Intelligence.
Intelligent Game NPCs: An example of this is the PS4 game Last of Us, which features a teenage sidekick named 'Ellie' who is an AI-NPC and is also the key to the gameplay, which is about the survival of humanity.
Personalised Gameplay: AI games have also used Expert Systems and ML algorithms to allow players to make distinct choice in video games, and then determining the output of the choice in order to continue the game.
Promoting and Marketing Films: An example of this is 20th Century Fox's Merlin Video Neural Network to determine the success of its promotional videos and clips based on the demographic information of its audience.
Predicting the success of Films: AI can analyse the success of a Film's script based on the regional demographics and genres, and determine its possible revenue. An example is Sony Pictures, who used the app ScriptBook to analyse 62 of their movies.
Although the applications are countless and there are so many breathtaking examples of AI once you actually get into it deep, this was just a sneak-peek into what awaits us out there.
Professional Scope in the AI Field
Artificial Intelligence is the future. And the future is here.
- Dave Waters
With the Artificial Intelligence Field gaining so much popularity in the modern times, many across the world are considering pursuing it as their careers. So here is a brief glance at the occupational opportunities in the field of Artificial Intelligence:
Data Scientist:
Data Scientists help in solving problems and making decisions especially in businesses, using computational procedures, mathematics, and statistical analysis or pre-existing datasets in order to identify trends and make predictions about the future.
Data Engineer:
Data Engineers work toward providing Data Scientists with usable data for performing analysis. They basically engineer comprehensible information for consultants and researchers out of the raw data provided to them.
Business Intelligence Developer:
This is similar to a Data Scientist, but focused only to a particular business. They analyse data to determine future trends and draw conclusions regarding a company's finances, or targeted market as key to optimising the company's output.
Artificial Intelligence Researcher:
As the name suggests, AI Researchers and Engineers work toward developing AI technology and incorporating AI solutions, algorithms and logic systems into existing software. They are exposed to all aforementioned fields of Artificial Intelligence.
Machine Learning Engineer
This can also be called a sub-category of AI Research. ML Engineers essentially focus on the development and incorporation of algorithms which allow computers to learn on their own using data and experience. They work toward making these systems compatible with existing software as well.
Robotic Scientist
Robotic Scientists focus on bringing the software concept of Artificial Intelligence into the physical and work toward combining hardware with digital algorithms and AI programmes. In layman terms, these people build robots and smart automated machinery.
Software Engineer
It is not secret that someone who has studied Artificial Intelligence has to have studied Software Engineering. Doing Software Engineering after studying AI opens your doors to incorporating many interesting features to your software such as Augmented Reality, Blockchain, Machine Learning, etc.
Conclusion
Artificial Intelligence truly is a magnificent concept, one that no one would have though possible just a hundred years ago. There are countless new applications to explore in this field and what I have written in this post has barely scratched the surface of our existing knowledge of AI. I will soon be doing a post about the technical side of AI, so if you have any suggestions or requests please drop them in the comments along with your likes. I would like to extend my sincere gratitude and before I sign off, I would just like to remind you to keep exploring and keep dreaming!
Until Next Time ~