Unlocking the Concepts of

Artificial Intelligence

Click Each Section to Read More!

Artificial Intelligence (AI) encompasses a range of technologies that enable machines to perform tasks requiring human intelligence, such as learning, reasoning, and adapting. This infographic unravels the core concepts of AI, including machine learning, supervised and unsupervised learning, reinforcement learning, and deep learning.

Artificial Intelligence

Artificial Intelligence (AI) is a rapidly advancing field with no universally accepted definition. Different scholars have tried to define AI by linking it with human intelligence and cognitive abilities, from machines mimicking human behaviour to systems making autonomous decisions. In general, AI refers to the development of systems capable of performing tasks that would typically require human intelligence. These tasks include learning from data, recognizing patterns, making decisions, and adapting to new situations (Thornberry, 2021; McCarthy, 2007).

While there is no single accepted definition of AI, there are certain common features that are evident across the various descriptions of AI.

While there is no single accepted definition of AI, there are certain common features that are evident across the various descriptions of AI.

  • Information processing is a fundamental aspect of AI. AI systems are capable of processing large amounts of information.
  • Environmental perception is a key feature of AI systems. AI is able to perceive and interpret data from its environment.
  • Goal Achievement: AI systems are designed to achieve specific business objectives.
  • The final stage of the process is decision-making. AI systems make decisions based on data analysis, as outlined by Samoili et al. (2020).

Importance and Scope of AI

AI has been hailed as a technology of the future and a pinnacle of human technological advancement. Several research studies have elaborated on the scope of AI in almost every aspect of human life. These include robotics, healthcare, finance, space, social media, education, surveillance and e-commerce. The practicality of AI demonstrates that it is a dynamic tool which enhances performance, and efficiency and encourages innovation across industries.

Here are some of the applications of AI in various industries.

  • Finance and Banking: Enhancing security, reducing costs, and improving customer support through chatbots (Mhlanga, 2020).
  • E-commerce: Predicting demand, adapting to market changes, and personalizing marketing strategies, as seen with Amazon (Liang & Tao, 2020).
  • Robotics: Increasing autonomy, productivity, and safety in industrial automation (Najmaei & Kermani, 2010).
  • Social Media: Shaping content and influencing user behaviour (Balaji et al., 2021).
  • Space Exploration: Analyzing data and enabling autonomous space missions (Chien & Morris, 2014).
  • Automobiles: Enabling autonomous driving, as demonstrated by Tesla (Tong et al., 2019).
  • Surveillance: Enhancing security systems for object detection and anomaly recognition (Sikora et al., 2020).
  • Education: Facilitating personalized learning and automating tasks (Chen et al., 2020).

Machine Learning

Machine learning (ML) is a core component of artificial intelligence (AI). It enables computers to learn from data and experiences in a manner similar to how humans learn. ML systems create models that can be enhanced, identify patterns, and resolve issues. By utilising historical data, they are able to address new challenges (Celik, 2018).

Key Features of Machine Learning

Learning from Data

ML systems analyse historical data in order to identify patterns and make decisions

01

02

Self-Improvement

These systems are designed to continuously improve their performance as they process more data.

Problem-Solving

ML is capable of solving complex problems that are difficult to address with traditional programming, such as recognising speech or images

03

04

Adaptability

ML systems can adapt to new information and changing environments.

Predictive Analysis

ML is used by businesses to forecast trends, manage risks, and improve efficiency.

05

06

Real World Applications

Examples include virtual assistants, recommendation systems, and fraud detection.

Machine learning encompasses three distinct learning approaches:

Supervised Learning

Supervised learning is the foundational approach in machine learning, where a machine learns to map an input variable to an output variable based on labelled data pairs. This method is divided into two main categories: classification and regression.

Classification categorizes data, such as spam detection in emails (Han, Pei & Tong, 2022). Regression, on the other hand, predicts continuous values based on existing data, such as the prediction of property prices (Sarker, 2021). Supervised learning requires structured data, as unstructured data can significantly reduce efficiency and accuracy (Kotsiantis, 2007). The process involves training the machine with a training set, which contains known input-output pairs, and then validating its performance on a test set, which evaluates the machine's ability to generalize to new, unseen data.

Structured vs. Unstructured Data

Structured data is highly organised and easily searchable, typically stored in relational databases and spreadsheets, with a fixed format of rows and columns. Financial records and sensor data are two examples of structured data. Unstructured data lacks a predefined format, making it more complex to analyse. It encompasses text documents, multimedia files, and social media content, necessitating the utilisation of advanced processing techniques such as natural language processing and machine learning for analysis.

Unsupervised Learning

Unsupervised Learning (UL) is dedicated to identifying patterns within datasets which has unstructured data points. Unlike supervised learning, UL operates independently, it does not require any external supervision and enables artificial intelligence systems to autonomously discern patterns from input data (Naeem, Ali, Anam & Ahmed, 2023). The strength of UL lies in its ability to reveal unexpected insights from data (Naeem et al., 2023). The algorithms are trained to process data which has no labels or classed data points.

Within the domain of unsupervised learning, clustering emerges as a primary learning task, focused on classifying items into cohesive groups based on similarities or patterns. Clustering methodologies encompass various forms, including partitioning, hierarchical, overlapping, and probabilistic clustering (Naeem et al., 2023).

Reinforcement Learning

Reinforcement learning (RL) empowers systems to autonomously make decisions in a dynamic environment using the principle of trial and error (Francois-Lavet, Henderson, Islam, Bellemare & Pineau, 2018). The decision which maximizes the collective rewards is reinforced based on feedback received for individual actions. The feedback can either be positive or negative, thus guiding the learning process for rational decision-making.

The RL emulates natural intelligence and mirrors human cognition. There is no explicit programming or human intervention in RL systems. These systems have algorithms which work on trial-and-error strategies to accomplish desired outcomes (Francois-Lavet et al., 2018).

Trial and Error

For example, consider a scenario where an AI-powered autonomous vehicle learns to navigate a complex urban environment. The vehicle uses its RL algorithm to learn various actions such as acceleration, braking and steering, and receives feedback in the form of rewards or penalties based on its performance. Positive rewards can be given for reaching the destination safely and efficiently, while penalties can be given for violating traffic rules or causing accidents. Over time, the car learns to optimise its driving behaviour by associating actions with favourable outcomes. As a result, the car learns the art of safe and efficient navigation without human intervention, based on its RL algorithmic system.

Deep Learning

Deep learning is a powerful type of machine learning that uses artificial neural networks (ANNs) to learn from complex data like images and speech (Kriegeskorte, 2015).

There are three fundamental layers in the architecture of an ANN. These include the input layer, hidden layers, and output layer. In image recognition tasks, for instance, the input layer receives pixel values representing the image, while the output layer produces predictions such as object labels or classifications. In between the input and output layers, the hidden layer is sandwiched. This sandwiched layer is vital in network learning, which extracts abstract features and representations from the input data through iterative training processes (Ghiassi & Saidane, 2005).

Training of ANN:
Forward Propagation and Backward Propagation

Forward Propagation

Data moves through the network to produce a prediction (Kag & Saligrama, 2021).

Backward Propagation

The network adjusts its parameters based on the prediction error to improve accuracy (Kag & Saligrama, 2021).

For example, in image recognition, an ANN is trained on labelled images. During forward propagation, it processes the pixel values and makes a prediction. During backward propagation, it adjusts its parameters based on the error between the prediction and the true label. Over time, the network improves its ability to accurately recognize objects in images (Kag & Saligrama, 2021).

References

  1. AI, H. (2019). High-level expert group on artificial intelligence. Ethics guidelines for trustworthy AI, 6.
  2. Balaji, T. K., Annavarapu, C. S. R., & Bablani, A. (2021). Machine learning algorithms for social media analysis: A survey. Computer Science Review, 40, 100395.
  3. Çelik, Ö. (2018). A research on machine learning methods and its applications. Journal of Educational Technology and Online Learning, 1(3), 25-40.
  4. Chen, L., Chen, P., & Lin, Z. (2020). Artificial intelligence in education: A review. Ieee Access, 8, 75264 - 75278
  5. Chien, S., & Morris, R. (2014). Space applications of artificial intelligence. Ai Magazine, 35(4), 3-6.
  6. Dawson, D., Schleiger, E., Horton, J., McLaughlin, J., Robinson, C., Quezada, G., ... & Hajkowicz, S. (2019). Artificial intelligence: Australia’s ethics framework-a discussion paper.
  7. François-Lavet, V., Henderson, P., Islam, R., Bellemare, M. G., & Pineau, J. (2018). An introduction to deep reinforcement learning. Foundations and Trends® in Machine Learning, 11 (3-4), 219-354..
  8. Ghiassi, M., & Saidane, H. (2005). A dynamic architecture for artificial neural networks. Neurocomputing, 63, 397-413.

References

  1. Han, J., Pei, J., & Tong, H. (2022). Data mining: concepts and techniques. Morgan kaufmann.
  2. Kag, A., & Saligrama, V. (2021, July). Training recurrent neural networks via forward propagation through time. In International Conference on Machine Learning (pp. 5189-5200). PMLR.
  3. Kotsiantis, S. B., Zaharakis, I., & Pintelas, P. (2007). Supervised machine learning: A review of classification techniques. Emerging artificial intelligence applications in computer engineering, 160(1), 3-24.
  4. Kriegeskorte, N. (2015). Deep neural networks: a new framework for modeling biological vision and brain information processing. Annual review of vision science, 1, 417-446.
  5. Liang, Z., & Tao, F. (2020, December). Research on the Application of Artificial Intelligence in E-commerce Design. In 2020 International Conference on Innovation Design and Digital Technology (ICIDDT) (pp. 455-458). IEEE.
  6. McCarthy, J. (2007). What is artificial intelligence.
  7. Mhlanga, D. (2020). Industry 4.0 in finance: the impact of artificial intelligence (ai) on digital financial inclusion. International Journal of Financial Studies, 8(3), 45.

References

  1. Naeem, S., Ali, A., Anam, S., & Ahmed, M. M. (2023). An unsupervised machine learning algorithms: Comprehensive review. International Journal of Computing and Digital Systems
  2. Najmaei, N., & Kermani, M. R. (2010). Applications of artificial intelligence in safe human–robot interactions. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 41 (2), 448-459.
  3. Samoili, S., Cobo, M. L., Gómez, E., De Prato, G., Martínez-Plumed, F., & Delipetrev, B. (2020). AI Watch. Defining Artificial Intelligence. Towards an operational definition and taxonomy of artificial intelligence.
  4. Sarker, I. H. (2021). Machine learning: Algorithms, real-world applications and research directions. SN computer science, 2(3), 160.
  5. Sikora, P., Malina, L., Kiac, M., Martinasek, Z., Riha, K., Prinosil, J., ... & Srivastava, G. (2020). Artificial intelligence-based surveillance system for railway crossing traffic. IEEE Sensors Journal, 21 (14), 15515-15526.
  6. Susar, D., & Aquaro, V. (2019, April). Artificial intelligence: Opportunities and challenges for the public sector. In Proceedings of the 12th international conference on theory and practice of electronic governance (pp. 418-426).
  7. Thornberry, W. M. M. (2021). National Defense authorization act for fiscal year 2021. Public Law, 116, 283.
  8. Tong, W., Hussain, A., Bo, W. X., & Maharjan, S. (2019). Artificial intelligence for vehicle-to-everything: A survey. IEEE Access, 7, 10823-10843.

Thank you

AUTHOR

Christian Meier-Staude