Introduction to Artificial Intelligence

Original article was published on Artificial Intelligence on Medium


Introduction to Artificial Intelligence

VPN SRUS

What is artificial intelligence?
Artificial intelligence (abbreviation AI) is a branch of computer science that deals with the automation of intelligent behaviour and machine learning. In practice, AI is often used as a collective term for algorithmic systems that can solve specific tasks. The term artificial intelligence is generally used very widely. The reason for this is that it cannot be clearly delimited due to the lack of a precise definition of intelligence.

The goal of AI is to capture the collective intelligence of people and to perform a given task better than any individual human being can ever do.

AI applications in the modern world:
– Autonomous driving
– Language translation/language recognition
– Face recognition
– Medicine (for ex. Detection of lung cancer or strokes on the basis of CT scans)

What is data science?
Data Science is a science that deals with the extraction of knowledge from large amounts of data.
Techniques and methods from mathematics, statistics, stochastic, and computer science are applied.

DS applications in the modern world:
– Fraud and Risk Detection
– Healthcare
– internet search
– Airline Route Planning

Is Al taking jobs away?
Artificial intelligence does not take jobs!

Heise.de writes about a study by the IT consulting firm Capgemini, which surveyed a good 1000 companies on artificial intelligence. “The study contradicts a fear, particularly in the United States of America or even worldwide, that artificial intelligence makes human work superfluous. Algorithms may do tasks that humans have done, but companies grow economically through the use of AI and make further future-oriented jobs possible”. For companies is important to train their workers in the use of AI.

What is Machine Learning?
Machine learning is the generation of knowledge from experience. An artificial system learns from examples and can generalize them after the learning phase is over. For this purpose, algorithms in machine learning build a statistical model based on training data.

Machine Learning has 3 subfields:
– Supervised Learning
– Unsupervised Learning
– Reinforcement Learning

Supervised Learning:
In supervised learning, you have input variables and an output variable, and you use an algorithm to learn the mapping function from input to output.

The goal is to approximate the mapping function so well that you can predict the output variables for new input data.
It is called supervised learning because you can think of the process of an algorithm learning from the training data set as a teacher supervising the learning process. We know the correct answers, the algorithm makes iterative predictions about the training data and is corrected by the teacher. Learning stops when the algorithm has reached an acceptable level of performance.

Imagine: We give a person vegetables and also tell him what the vegetables are called (e.g. artichokes, beans, …). This person should identify the after that the names of the vegetables.

A current topic would be:
We now have data from a Covid-19 data set (data of infections worldwide) and therefore we can use an algorithm to predict the worldwide Covid-19 infection in 20 days.
However, it is important to note in this example which measures (mask obligation, quarantine, and other measures) the health authorities would take in 20 days, this also influences the prediction of our model.

Unsupervised Learning:
Unsupervised machine learning algorithms derive patterns from a data set without reference to known or tagged results. Unlike supervised machine learning, unsupervised machine learning algorithms cannot be applied directly to a regression or classification task because you have no idea what the values for the output data might look like, making it impossible for you to train the algorithm as you would normally do. Unattended learning can be used instead to discover the underlying structure of the data.

Imagine: We give a person vegetables and that person should sort the vegetables into different groups, but the person has no knowledge about vegetables. This is called unsupervised learning (clustering). The person will sort the cards in his own intuition by color or size.

  1. A current topic would be:
    – Biology — for genetic and species grouping;
    – Recommendation systems — giving you better Onlineshop purchase suggestions or Prime/Netflix movie matches;
    – Medical imaging — for distinguishing between different kinds of tissues;
    – Market research — for differentiating groups of customers based on different attributes.
Wikimedia

Reinforcement learning:
As a result, reinforcement learning makes a form of artificial intelligence possible that can solve complex control problems without prior human knowledge. Compared to conventional engineering, such tasks can be solved many times faster, more efficiently, and in the ideal case even optimally. Leading AI researchers describe RL as a promising method for achieving Artificial General Intelligence.

In short, it is the ability of a machine — similar to a human — to successfully perform any intellectual task. Like a human, a machine must observe and learn from various causalities in order to solve unknown problems in the future.

The goal of RL is to find an optimal policy. A policy is simply the learned behavior of a software agent. A policy specifies which action should be executed for any behavior variant (Observation) from the learning environment (Environment) to maximize the reward.

Imagine: We humans are not always lucky enough to be able to do everything in life, which is also the case with reinforcement learning. If we do something wrong, we don’t get any reward, but if we do something right, we get a reward. Through a reward we automatically learn along with it, that is also the case with reinforcement learning!

A current topic would be:
RL can be used in robotics for industrial automation.
– RL can be used in machine learning and data processing
– RL can be used to create training systems that provide custom instruction and materials according to the requirement of students.

Wikimedia