Although the first artificial intelligences date back to the 1950s, they have permeated society in recent years and are no longer a futuristic concept but rather have become an essential tool in many people’s daily lives. However, not all AIs are the same, and this is particularly important when choosing the right AI for the right purpose. Depending on their capacity, functionality and application, different types are available to suit specific needs.
Understanding the different types of artificial intelligence has practical implications for businesses and entrepreneurs. For example, machine learning can optimise marketing campaigns or resource management, while computer vision improves security through facial recognition (Zhang et al., 2022; Li et al., 2023).
Classification by Capacity and Functionality
One of the most common ways of categorising artificial intelligence is according to its ability to perform tasks, which includes three types: narrow AI, general AI and superintelligence. However, only the first is common today (Zhang et al., 2022), while the second, which would equal the human brain in any domain, is under development and still requires progress in several areas, and the third, which would surpass it, is only theoretical (Wang et al., 2023).
Functionality, i.e. how it processes information and makes decisions, is another way of classifying AIs. There are three main systems: reactive, with no memory or learning capabilities, such as the classic Deep Blue that defeated Garry Kasparov in 1997 (Russell & Norvig, 2021), with limited memory, which do use historical data to improve their performance and are used, for example, in autonomous vehicles to gather information and make real-time decisions, and mind theory, which should in theory be able to understand human emotions, beliefs and intentions and are expected to have great potential in customer service or education (Huang et al. 2022).
Classification by Technology
AI can also be categorised according to its main technology:
1. Machine learning, which can be supervised, when trained on previously labelled data, unsupervised for the identification of patterns in unlabelled data and very useful in market segmentation and identification of microtrends and by reinforcement, referring to systems that learn via trial and error, such as algorithms that play video games (Goodfellow et al., 2016).
2. Neural networks (deep learning), inspired by the human brain to process information in layers. They are used in facial recognition and virtual assistants (LeCun et al., 2015).
3. Natural language processing (NLP), to understand and generate human language, such as chatbots and machine translators (Zhang et al., 2022).
4. Computer vision, which teaches machines to interpret images and videos and is used in medical diagnostics and autonomous vehicles (Li et al., 2023).
5. Expert systems, which emulate the decision-making of a human expert in specific areas, such as medical diagnosis (Russell & Norvig, 2021).
6. Intelligent robotics, which combines AI with hardware to perform physical tasks, such as robots in warehouses (Huang et al., 2022).
Which AI to choose?
Most modern artificial intelligences combine several technologies, integrating multiple approaches to achieve more robust and accurate results. This allows them to tackle complex problems by applying each technology’s specific strengths. This brings us to the classification according to function, which can offer great advantages in a variety of industries:
1. Descriptive AI: focuses on analysing historical data to understand what has happened. It is ideal for generating reports and visualisations summarising past performance. Its applications include analysing monthly sales and identifying seasonal trends, as well as displaying key metrics such as web traffic and operational efficiency. It is also useful in logistics, to assess delivery routes and identify bottlenecks (Russell & Norvig, 2021).
2. Diagnostic IA: identifies the causes behind an event or problem. In the health sector, for example, it is used to analyse symptoms and clinical data to diagnose diseases (Topol, 2019). In business, it is also used in campaign analysis and sales results, and in the manufacturing industry it is used to detect equipment failure causes and improve production efficiency (Li et al., 2023).
3. Predictive AI: uses historical data to predict future outcomes. It is used to forecast market fluctuations or identify credit risks, predict product demand and adjust inventories or anticipate consumption peaks and optimise resource allocation (Huang et al., 2022).
4. Prescriptive AI: not only predicts outcomes, but also recommends actions to optimise them. In the supply chain, it can suggest more efficient delivery routes or adjustments to inventory levels. In business, it helps make strategic decisions or allocate budgets between different channels to maximise return on investment; and in healthcare, it can recommend personalised treatments based on medical history (Wang et al., 2023).
5. Cognitive AI: combines data analytics with advanced contextual understanding capabilities. In the education sector, it is used to create intelligent tutoring systems that adapt to student needs, and in business, virtual assistants improve customer service and satisfaction. It is also key in the entertainment industry, where recommendation systems personalise user experience based on their preferences (LeCun et al., 2015).
Conclusion
Understanding the differences between various types of AI is key to making the most of this technology and staying competitive in an increasingly digitised market.
Choosing the right type of AI depends on your goals. Descriptive AI is the best ally for understanding the past, while predictive AI is the best ally for anticipating the future. To optimise real-time decisions, prescriptive is the best option. In any case, combining these tools can transform the way we analyse data, interact with customers and make decisions. The key is to align technology with one’s specific needs.
References
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
Huang, M. H., Rust, R. T., & Maksimovic, V. (2022). The feeling economy: How artificial intelligence is creating the era of empathy. Journal of Business Research, 139, 1-10. https://doi.org/10.1016/j.jbusres.2021.09.045
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539
Li, J., Zhang, Y., & Chen, X. (2023). Artificial intelligence in supply chain management: A review and future directions. International Journal of Production Economics, 245, 108405. https://doi.org/10.1016/j.ijpe.2022.108405
Russell, S., & Norvig, P. (2021). Artificial Intelligence: A Modern Approach (4th ed.). Pearson.
Topol, E. J. (2019). Deep medicine: How artificial intelligence can make healthcare human again. Basic Books.
Wang, Y., Zhang, W., & Liu, H. (2023). Towards artificial general intelligence: Challenges and opportunities. Nature Machine Intelligence, 5(3), 123-135. https://doi.org/10.1038/s42256-023-00612-8
Zhang, X., Li, Y., & Wang, Z. (2022). Machine learning in e-commerce: A comprehensive review. Journal of Retailing and Consumer Services, 64, 102742. https://doi.org/10.1016/j.jretconser.2021.102742
