Analysis of the Current State of Research on Artificial General Intelligence (AGI)

Bachelor's thesis is focused on the topic of Artificial General Intelligence (AGI). Given the rapid advancement of generative artificial intelligence in recent years, AGI has become a subject of intense discussion, with the term often being overused. Therefore, the thesis first defines the concept of AGI considering its current interpretations among experts and provides an insight into the definition of intelligence and its generality to better understand the motivation behind the development of AGI. The thesis primarily focuses on advances in contemporary artificial intelligence that are relevant to the future achievement of AGI. It analyses classical machine learning methods (supervised learning, self-supervised learning, and reinforcement learning), none of which are currently considered sufficient to create AGI on their own. More promising approaches appear to be open-ended learning, learning in virtual environments, and especially the combination of multiple approaches. Key discoveries from recent years that have significantly influenced the current state of AI are summarized and assesses whether they bring us closer to AGI. There are discussed advances in AI's ability to reason about complex problems, its adaptability after training, memory, autonomous problem-solving, and the potential for combining models together. The thesis also includes an analysis of benchmark methods currently used to evaluate progress in AI development in relation to the pursuit of AGI. Finally, it presents an overview of the evolution of predictions made by leading experts regarding AGI and the most recent forecasts for the development of this field.