schwarzd Spannende Frage nach den Möglichkeiten des Einsatzes von AI.
Ich lanciere an der HSLU gerade ein Forschungsprojekt mit dem Arbeitstitel “Active Inference basiertes Community Management für automatisiertes und intelligentes ‘Managen’ sowie transparentes Lernen”. Lass uns sprechen…
Dazu muss ich noch sagen, dass “Active Inference” ein AI-Ansatz ist, der nichts mit den gehyepten Lange Language Models zu tun hat (siehe unten)
Active inference-based artificial intelligence differs significantly from traditional machine learning methods like supervised, unsupervised, and reinforcement learning. Active inference is a theoretical framework that posits intelligent agents minimize the discrepancy between their internal models of the world and sensory inputs by actively seeking information (Friston et al., 2017). In contrast, traditional machine learning methods, such as supervised learning, rely on labeled data to learn patterns and make predictions based on provided examples (Nguyen & Ngo, 2009). Active inference-based AI focuses on the agent’s interaction with the environment, incorporating beliefs, desires, and intentions to make decisions (Friston et al., 2017). This is in contrast to traditional machine learning methods like unsupervised learning, where the system aims to find the inherent structure in input data without explicit guidance from labeled examples (Nguyen & Ngo, 2009). Moreover, active inference-based AI involves continuous decision-making processes where the agent iteratively updates its beliefs based on new information, akin to reinforcement learning but with a different underlying principle (Friston et al., 2017). In reinforcement learning, agents learn through trial and error by receiving rewards or penalties based on their actions, aiming to maximize cumulative rewards over time (Vargas et al., 2014). Active inference-based AI stands out for its emphasis on the agent’s active role in information gathering and decision-making, which is distinct from the more passive learning approaches of traditional machine learning methods. While traditional methods have been successful in various applications, active inference offers a unique perspective on how intelligent agents can interact with and learn from their environment in a more dynamic and adaptive manner.
References:
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., & Pezzulo, G. (2017). Active inference: a process theory. Neural Computation, 29(1), 1-49. https://doi.org/10.1162/neco_a_00912
Nguyen, D. and Ngo, T. (2009). Supervising an unsupervised neural network.. https://doi.org/10.1109/aciids.2009.92
Vargas, D., Takano, H., & Murata, J. (2014). Novelty-organizing classifiers applied to classification and reinforcement learning.. https://doi.org/10.1145/2598394.2598429
