What is AI?
In order to understand the major concept of Artificial Intelligence (AI), it is more simplistic to remember the movie “The Terminator”. If you didn’t watch it I recommend you do 🙂
In this movie, the main actor Arnold Schwarzenegger acts as a cyborg who came from the future. This cyborg is a great representation of advanced AI. It looks and behaves like a human, has ability to see and can understand what humans do and talk about, also has the ability to move like a human too.
Above mentioned cyborg characteristics represent key fields of the current AI Research and Development (R&D). To be more precise, those AI disciplines are:
- Machine Learning (ML)
- Computer Vision (CV)
- Natural Language Processing (NLP)
- Robotics and Cybernetics (R&C)
What is DataScience?
Data Science (DS) and AI are not rarely confused in today’s society. Sometimes, DS and AI are viewed as two different names for the same thing. But that is not a case.
DS is focused on obtaining conclusions from the data using statistical models and procedures. It involves various underlying data operations (cleaning and verifying data, …) which leads to more accurate statistical models, and by that more accurate results. Some of the common applications of DC are marketing, advertising, business analysis, etc.
On the other hand, AI is focused on automation in a vast number of domains (healthcare, manufacturing, robotics, etc.). AI uses its advanced algorithms (like neural networks) in order to enable a machine to give and act on own decisions (similar to humans) by processing a considerable number of data samples.
As already noted, AI’s focus is on mimicking human-like behavior in a large number of situations and circumstances. As a result, it is easy to map AI domains to the human points in the following way:
- Machine Learning = Brain/Thinking
- Computer Vision = Seeing
- Natural Language Processing = Language understanding
- Robotics and Cybernetics = Movements
Further sections will explain those domains in more detail with additional examples from the industry.
Machine learning is an application of artificial intelligence (AI) that provides systems ability to automatically learn and improve from experience without being explicitly programmed. Machine learning focuses on the development of computer programs that can access data and use it to learn for themselves.
The process of learning begins with observations or data, such as examples, direct experience, or instruction, in order to look for patterns in data and make better decisions in the future based on the examples that we provide. The primary aim is to allow the computers to learn automatically without human intervention or assistance and adjust actions accordingly.
To be exact, ML is inter combined with other AI domains. It represents the capability of the AI algorithm to learn patterns from the data and to recognize them in not previously seen data. In order to learn, a machine algorithm iterates over the data and using statistics tries to understand the underlying pattern between the different data points. Depending on the data we have and strategy (algorithm) we use to recognize a pattern in the data, ML could be categorized in:
- Supervised learning
- Unsupervised learning
- Semi-supervise learning
- Reinforcement learning
Can apply what has been learned in the past to new data using labeled examples to predict future events. Starting from the analysis of a known training dataset, the learning algorithm produces an inferred function to make predictions about the output values. The system is able to provide targets for any new input after sufficient training. The learning algorithm can also compare its output with the correct, intended output and find errors in order to modify the model accordingly.
Used when the information used to train is neither classified nor labeled. Unsupervised learning studies how systems can infer a function to describe a hidden structure from unlabeled data. The system doesn’t figure out the right output, but it explores the data and can draw inferences from datasets to describe hidden structures from unlabeled data.
Fall somewhere in between supervised and unsupervised learning, since they use both labeled and unlabeled data for training – typically a small amount of labeled data and a large amount of unlabeled data. Systems that use this method are able to considerably improve learning accuracy. Usually, semi-supervised learning is chosen when the acquired labeled data requires skilled and relevant resources in order to train it / learn from it. Otherwise, acquiring unlabeled data generally doesn’t require additional resources.
Learning method that interacts with its environment by producing actions and discovers errors or rewards. Trial and error search and delayed reward are the most relevant characteristics of reinforcement learning. This method allows machines and software agents to automatically determine the ideal behavior within a specific context in order to maximize its performance. Simple reward feedback is required for the agent to learn which action is the best; this is known as the reinforcement signal.
Did you ever think about how Amazon and similar E-Commerce solutions know in which products are you interested? As you can suppose, the reason for their knowledge lies in an enormous number of data they collect from you, and advanced AI models that predict most likely products you will buy.
While you are searching and looking for a product you want to order next, Amazon in the background collects everything. For example, which product you clicked first, did you click anything from recommended products, which products did you like/rate/comment/shared, how long you stayed on a certain page and many other (+ history of all of them). Using this information (+ same information from the rest of the users) and highly efficient AI models, they can predict which products you will want to buy next and for which price. In some cases, they probably know you better than yourself.
Would it be great to know in advance what will be the stock market price? If you can predict this, you will probably become one of the richest people in the world.
AI is also commonly used in this domain. Brokers are actively relying on AI algorithms to help them to decide what to do with certain stocks. Unlike human, the machine can process an enormous amount of information in a fraction of a second. The exploitation of this potential with the addition of AI algorithms can give in real-time predictions about stock prices in the future with certain confidence (probability).
Wherever you are, and whatever you do on the internet, your every step (click, search, like, comment, etc.) is recorded. When you search using Google, like on Facebook, follow on Instagram, … Everything is recorded. Why? Well, when you have data, you can do anything. In this case, using all data stored from you hi-tech companies like Google, Facebook, and Amazon can predict what would you most like on their platforms. For example, which product to buy, which posts to see, etc.
Most of them are also using that data to place ads that will be viable for you. Google as a company makes income from the ads and offers most of its services for free just to gather more information about you. As the saying goes: “If you aren’t paying for it, you are the product.”.
Computer vision is an interdisciplinary scientific field that deals with how computers can be made to gain a high-level understanding of digital images or videos. From the perspective of engineering, it seeks to automate tasks that the human visual system can do.
In most cases, CV uses one specific type of AI algorithm called Neural Networks. The most popular variation of this type of network used in CV is Convolutional Neural Network. They have capabilities to look for patterns among images and videos and by that solve (or optimize) a vast number of daily problems. CV today’s focus is on object detection and recognition.
Often, people think that some domain experts are irreplaceable. One case could be a medical doctor. But, the truth is, that isn’t a case. Using the AI, you could replace almost any expert in a large number of domains.
A couple of years ago, a competition was organized in which the machine was competing with dozens of doctors on a diagnosis of a certain type of disease. By the end of the competition, the mean score was calculated for all doctors and it was compared to the machine score. The result shows that the machine scored much better then doctors did, in a fraction of time. A similar example could be applied to almost every expert of the domain and compared to human performance.
Apart from crunching financial, medical, user and rest of the data in order to give a certain prediction, AI could also be used in artistic ways. For example, compose music, create abstract art, etc.
How does this work? There exist certain AI algorithms (Neural Networks) that have capabilities to mimic an artist’s way of creating an artwork. With this approach, they are capable of creating new art, transferring one art style to another image, compose music. Mozart style, etc.
Self-driving cars are ones that don’t require a human driver in order to operate the vehicle. They are equipped with cameras, different types of sensors and advanced software which enables them to drive on their own. They are usually called and autonomous or driverless cars. Most famous company that produces self-driving cars is Tesla.
Natural Language Processing
Natural language processing (NLP) is a subfield of linguistics, computer science, information engineering, and artificial intelligence concerned with the interactions between computers and human (natural) languages, in particular how to program computers to process and analyze large amounts of natural language data.
In other words, the NLP domain tries to understand how the human process and understand language, with all complexity that comes with it (irony, sentiment, …). The main problem is that human speech is not always precise. It changes a lot depending on nationality, slang, and many more different variables that could influence how a person will construct their sentence (and speak it). Some of the NLP applications include sentiment analysis, speech recognition, natural language generation, etc.
Some applications adopted in the field of customer support. Nowadays, before you speak to the real person about your relevant question, you will be first redirected to the AI bot that will naturally answer some of your questions. If it fails to do so, you will be transferred to the regular operator.
In some cases, you will fully speak, or communicate via chat boxes with a chatbot and he/she (it) will solve your problem. From a company perspective, this cut expenses on support drastically.
Did you ever think about how will sound next track from The Rolling Stones, Bob Dylan, or perhaps Beethoven? Well using advances AI models, today is possible to generate new compositions based on previous records of one or more artists. This type of algorithm mimics musician style of composing music and then generates previously non-existing artwork. This shows that AI isn’t strictly only bind to the automation, but could also be used in creative industries.
Robotics & Cybernetics
Definition of Robotics:
Robotics is a scientific and engineering discipline that is focused on the understanding and use of artificial, embodied capabilities. The people who work in this field (roboticists) come from mechanical engineering, electronic engineering, information engineering, computer science, and other fields. On the engineering side, roboticists deal with the design, construction, operation, and use of robots, especially through computer systems for their control, sensory feedback, and information processing. On the scientific side, roboticists study how a robot’s environment and design affect how well it does its job.
Cybernetics by definition:
Cybernetics is a transdisciplinary approach for exploring regulatory systems—their structures, constraints, and possibilities. Norbert Wiener defined cybernetics in 1948 as “the scientific study of control and communication in the animal and the machine.” In other words, it is the scientific study of how humans, animals, and machines control and communicate with each other.
From the two above mentioned definitions, you can conclude that these two fields are tightly interconnected. Cybernetics focus is on understanding complex systems, especially ones with a certain type of feedback loops. Unlike robotics, it is not strictly focused on engineering. Cybernetics tries to understand processes also in economy, biology, sociology, etc.
Robotics is mostly focused on engineering domains on how to create operative machinery. It looks at how to design electronics, sensors, novel mechanisms, etc. Interconnection with cybernetics comes from that all machines have a certain type of processing (feedback) loop. Unlike cybernetics, robotics is mostly focused on pure engineering.
Robotics is used in many industries mainly for production purposes (car industry, food industry, etc). Emerging industry of our age by usage of the robotics systems is the military. By using robotics there will be no need in sending human soldiers into the war. Instead, automated machines could be remotely controlled, or be completely autonomous. One of the most popular robotic manufacturer today is Boston Dynamics.
They are creating advanced robotics systems, that are capable to adapt to rough terrains, learn and improve over time in order to complete a given mission. Another example of military usage of this field (combined with other fields) is drones. These types of military drones are equipped with massive firepower and are used to fight a war without involving humans in life-risking situations.
As you can see, AI has a number of applications in plenty of domains. And as any technology with great potential, it could be both used in a good and in a wrong way. So, the question is, will we withstand the Terminator prophecy or the largest growth of the humankind? Maybe something in between? Let’s leave this for creative mind games and imagination. Eventually, time will tell.