Humanity has been striving for thousands of years to understand its own cognitive processes: How can the brain, as a biological structure, perceive, interpret and manipulate an environment that is vastly more complex than it is? However, the field of artificial intelligence (AI) goes beyond epistemological analysis. Its purpose is not only to model the intellect theoretically, but also to produce and reproduce it artificially.<\/p>\n\n\n\n
In the current scientific discourse, AI is, in addition to cellular biology, the most dynamic and intensively researched sector. As a young discipline with only a few decades of history, it carries many open questions, ontological uncertainties and technological challenges. This five-part series of articles guides the reader under the hood, exploring the theoretical foundations of AI, the logic of rational decision-making and the structure of modern architectures.<\/p>\n\n\n\n
The term \u2018artificial intelligence\u2019 is to some extent misleading, as it is often combined with anthropomorphised meaning in everyday language. In technical terms, most modern systems do not \"think\" in the sense of human consciousness, but rely on pattern recognition and statistical rationality derived from complex data sets.<\/p>\n\n\n\n
Statistical Probability vs. Cognitive Content<\/strong> If an AI model complements the initial sequence \u2018Apple\u2026\u2019, it does not evoke the physical attributes (taste, texture) of the object. The model derived schemas (maximum likelihood estimation) from the data during the learning phase. If seven times \u2018red\u2019, twice \u2018green\u2019 and once \u2018yellow\u2019 are used in the input corpus, the system will generate the \u2018red\u2019 output based on the statistical probability.<\/p>\n\n\n\n Today, AI is best described by the metaphor of a highly effective personal assistant. Models such as ChatGPT are \u2018good at all times\u2019, giving the appearance of General AI, when in fact they are complex networks of specific algorithms. As an intellectual discipline, AI ranges from chess and proof of mathematical theorems to complex diagnostics and autonomous driving.<\/p>\n\n\n\n The development of the discipline can be divided into distinct logical and technological stages:<\/p>\n\n\n\n Alan Turing\u2019s method proposed in 1950, the \u2018imitation game\u2019, is the cornerstone of behavioural intelligence determination. During the test, the investigator communicates with the subject through a terminal; If you cannot distinguish machine responses from human responses, the machine is considered intelligent.<\/p>\n\n\n\n For functional intelligence, AI must have four basic capabilities:<\/p>\n\n\n\n The modern extended Turing test goes beyond text interaction and includes machine vision<\/strong>, a understanding of speech<\/strong> and a robotics<\/strong> (physical manipulation). Although the $100,000 Loebner Prize is yet to be awarded, technology is rapidly approaching completion.<\/p>\n\n\n\n At the heart of AI research is rationality, which can be broken down into two components: decision and action.<\/p>\n\n\n\n A formal logic<\/strong> in principle, it is the perfect toolbox for rational decision-making, but reality data is often noisy, uncertain or incomplete. For this reason, modern AI uses probability as a bridge between formal logic and uncertain reality.\u00a0A Rational action<\/strong> An agent must function independently in its environment. It is important to note that rationality does not always mean \"perfect\" calculation: the Human Race for unconditional reflexes<\/strong> similarly, in some situations, a quick response is more rational than a time-consuming, deep analysis.<\/p>\n\n\n\n An agent is an entity that senses and acts in its environment. The hierarchy of agents reflects an increase in cognitive complexity.<\/p>\n\n\n\n Reflex-like agents<\/p>\n\n\n\n The simplest structures, which operate without internal state and memory, are based on direct stimulus-action rules. They only respond to the currently detected data.<\/p>\n\n\n\n Internal state agents (Model-based)<\/p>\n\n\n\n These agents maintain an internal \u2018world model\u2019. This also allows them to track changes in the environment that they are not directly perceiving, taking into account past interactions.<\/p>\n\n\n\n Targeted agents<\/p>\n\n\n\n Their operation is guided by a defined end state (target). They are able to plan future states, choosing between different sequences of action leading to the goal.<\/p>\n\n\n\n Utility-oriented agents<\/p>\n\n\n\n The highest level of agents that maximize a utility function. What matters is not only the achievement of the goal, but also the efficiency and \u2018quality\u2019 of the road. This is where they appear The Anytime Algorithms<\/strong> also: systems that are able to provide a sub-optimal response immediately, but continuously refine and refine their decision, giving them more time.<\/p>\n\n\n\n In machine learning, the system improves its future decision-making efficiency based on observations. A critical element of the process is hypothesis space<\/strong> The choice of (possible) solutions. Here is a basic trade-off, or equilibrium constraint: between the expressiveness of the hypothesis space (e.g. Turing machines vs. linear functions) and the computational complexity of the search.<\/p>\n\n\n\n There are two main directions of inductive learning:<\/p>\n\n\n\n When selecting models, it is Ockham's Razor<\/strong> principle applies: if several hypotheses are consistent with the data, the simplest hypothesis should be preferred, as it has the best generalising ability and a lower risk of over-learning.<\/p>\n\n\n\n Artificial intelligence is not based on a mystical \u2018machine brain\u2019, but on a network of rational agents built on rigorous mathematical and logical frameworks. From the first neural models in 1943 to modern utility-oriented systems, the history of AI is about balancing computational limitations with expressive power.<\/p>\n\n\n\n My series\u00a0Part 2<\/strong>\u00a0We go into the technical depths of machine learning: We examine the logic of decision trees, the question of linear separability and the world of perceptrons, which are the direct precursors of modern neural networks.<\/p>\n\n\n\n <\/p>","protected":false},"excerpt":{"rendered":" Welcome to this series of articles! As an expert and technology educator, my goal is to help dispel the mysticism and possible anxieties surrounding artificial intelligence (AI). <\/p>","protected":false},"author":2,"featured_media":1557,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","format":"standard","meta":{"slim_seo":{"facebook_image":"https:\/\/teszarypeter.hu\/wp-content\/uploads\/2025\/10\/AI-Category-Cover.jpg","twitter_image":"https:\/\/teszarypeter.hu\/wp-content\/uploads\/2025\/10\/AI-Category-Cover.jpg","title":"1. R\u00e9sz: Mi az a mesters\u00e9ges intelligencia? \u2013 fogalmak \u00e9s szintek - Tesz\u00e1ry P\u00e9ter","description":"\u00dcdv\u00f6z\u00f6llek ebben a cikksorozatban! Szak\u00e9rt\u0151k\u00e9nt \u00e9s technol\u00f3giai eduk\u00e1tork\u00e9nt az a c\u00e9lom, hogy seg\u00edtsek eloszlatni a mesters\u00e9ges intelligenci\u00e1t (MI) \u00f6vez\u0151 miszti"},"footnotes":""},"categories":[93],"tags":[94,96,95],"class_list":["post-2112","publikacio","type-publikacio","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","tag-ai","tag-mesterseges-intelligencia","tag-mi"],"desktop_mode_lock":null,"desktop_mode_contributors":[],"desktop_mode_attached_media":[1557],"meta_box":[],"_links":{"self":[{"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/publikacio\/2112","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/publikacio"}],"about":[{"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/types\/publikacio"}],"author":[{"embeddable":true,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/comments?post=2112"}],"version-history":[{"count":5,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/publikacio\/2112\/revisions"}],"predecessor-version":[{"id":2257,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/publikacio\/2112\/revisions\/2257"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/media\/1557"}],"wp:attachment":[{"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/media?parent=2112"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/categories?post=2112"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/teszarypeter.hu\/en\/wp-json\/wp\/v2\/tags?post=2112"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}3. Milestones of AI: A Brief Overview of History<\/h3>\n\n\n\n
\n
4. Measurement of intelligence: The Turing Test<\/h3>\n\n\n\n
\n
5. Rationality: Decision and action<\/h3>\n\n\n\n
6. Types of intelligent agents<\/h3>\n\n\n\n
Agent type<\/th> Main feature<\/th><\/tr> Reflex-like agent<\/td> Immediate response to current stimulus without internal state.<\/td><\/tr> Internal state agent<\/td> Maintains a constantly updated internal world model.<\/td><\/tr> Targeted agent<\/td> Plan future states to achieve the stated goal.<\/td><\/tr> Utility-oriented agent<\/td> Maximize utility function for efficiency.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n 7. Environments and Learning Paradigms<\/h3>\n\n\n\n
\n
8. Summary and outlook<\/h3>\n\n\n\n