Artificial intelligence - what is it?
Artificial intelligence is a field of computer science that deals with the creation of systems and algorithms capable of performing tasks that normally require human intelligence. Such tasks can include speech recognition, decision-making, natural language understanding, data analysis and even creative thinking.
John McCarthy is considered the author of the term, but the topic of artificial intelligence has been addressed by many researchers, including Alan Turing. He developed the famous Turing test to help assess the level of intelligence of machines. A computer program passed the test the moment it managed to convince the judge that it was talking to another human being.
Today, the field of AI uses a wide variety of techniques and algorithms that vary in their approach to data, complexity, and areas of application. There is much more to artificial intelligence, or AI, than just popular chatbots in the form of apps or browser-based helpers (such as ChatGPT or Microsoft Copilot).
Types of artificial intelligence – predictive AI vs. generative AI
AI tools can most broadly be divided into two types: predictive and generative AI, which differ primarily in how they process data and generate results.
The goal of predictive AI is to predict future events based on historical data. Predictive models learn patterns and relationships from training data, and then use these patterns to predict future performance. This approach is widely used today, for example, to forecast the weather, predict machine failures or analyze credit risk before lending. Predictive models learn from historical data and, to be effective, need very large datasets to do so. It is in this context that one encounters the term “Big Data”.
Generative AI creates new data that are similar to the training data. Generative models learn the distribution of training data and can generate new data samples that are statistically similar to the data they were trained on. They can create images, texts, sounds and other forms of data.
Key technologies used in artificial intelligence are based on machine learning (ML) and deep learning (DL) algorithms, among which decision trees, random forests, and neural networks, are the most common. Another important technology for AI is natural language processing (NLP) algorithms, which are integral to the recently popular chatbots.
The use of artificial intelligence in practice
In recent years, the use of artificial intelligence has grown tremendously in popularity, and it is now hard to find an industry where AI is not present in some way. Although it often arouses extreme opinions, there is no denying the importance of this technology in today's world. Due to the fact that very different tasks are placed before AI, the current trend is to create very specialized models and programs to meet specific needs.
Predictive AI models have already been used for years in many fields, such as:
- medicine: disease diagnosis, analysis of medical images (e.g. MRI images);
- finance: credit risk analysis;
- industry: supply chain management,
- transportation: route optimization, autonomous vehicles;
- education: personalization of learning;
- administration: automatic document generation.
Predictive models are usually created with long-term use in mind. They can be created only on the basis of internal data of a company or institution - if one is not necessarily interested, for example, in how customers' buying behavior is in general, but rather how customers of that particular store behave. Especially if the model uses sensitive data, which is often necessary, for example, in banking or administrative systems, it is important that the model is created and used internally within the institution.
When implementing predictive AI solutions, advanced analytical software is used to reliably build models, having control over data quality, accuracy of predictions and interpretability of results. With these tools it is possible not only to process and analyze large data sets, but also to monitor and optimize models in real time. Some programs, in addition to the algorithms themselves, may have additional functionality. An example is PS CLEMENTINE PRO which includes the IBM SPSS Modeler engine, a tool that has a graphical user interface (GUI) that allows users to create models by dragging and dropping components. This is particularly useful for analysts and statisticians who may not be familiar with programming. Also, not every program provides visualizations or predefined, advanced algorithms such as automatic data preparation or basket analysis.
The use of artificial intelligence models for prediction and forecasting has been widely used for a long time (e.g., the Dendral system for analyzing chemical data from the 1960s), while the popularity of generative intelligence is a matter of recent years. Although the first algorithms were also developed several decades ago, it is only today's models that make it possible to create high-quality and relevant content.
Generative AI models are used in many industries, such as:
- arts and entertainment: creating images, generating music, writing literary texts;
- manufacturing: generating new product designs and patterns, simulating production processes;
- marketing and advertising: creating advertising content;
- video games: creating realistic characters and environments, generating scenarios.
Generative models are typically used for tasks requiring creativity or innovation. They are trained on huge data sets to learn structure and patterns that can be used to create new, original data. Generative AI is much more likely to use off-the-shelf tools instead of building its own models and systems.
Summary
Artificial intelligence (AI) is a rapidly growing field of technology that enables systems to perform tasks that require human intelligence, such as speech recognition, decision-making and forecasting. AI can be divided into two main types: predictive and generative.
Unlike predictive models, which focus on predicting future events, generative AI focuses on creating new content that is similar to training data, yet unique and innovative. However, regardless of the tasks for which they are created, AI systems and tools require a very large amount of input data (historical, training data) to be effective.
While today's AI tools have broad applicability, enormous potential and unprecedented capabilities, they also pose a number of challenges. They require consideration of ethical issues, such as data privacy and algorithm bias, which can lead to unfair decisions. In the case of generative models, a considerable controversy has been the use of various works for which the software developers did not have copyrights to teach the model. The rapid development of artificial intelligence requires constant adaptation of laws and industry standards to use its capabilities safely and ethically.