Hello guys, do you have any opinion about active inference? Lately there were some interesting things going on related to use of bayesian techniques to tackle the non-reasoning part of current AI structure. This topic is not publicly discussed yet, but its been doing some leaps in robotics and overall integration with LLM. Furthermore, lately there seems to be more public attention to the fact that current models are non-reasonable and "do not learn" - their thought process is just trained from the data they use. Bayesian theory/active inference tackles this problem by updating its beliefs based on the environment. For some context, I am attaching articles to get a grasp of what this is about.
https://www.nature.com/articles/s41746-025-01516-2
https://arxiv.org/abs/1909.10863
https://arxiv.org/html/2312.07547v2
https://arxiv.org/abs/2407.20292
https://arxiv.org/html/2410.10653v1
https://arxiv.org/abs/2112.01871
https://medium.com/@solopchuk/tutorial-on-active-inference-30edcf50f5dc
[link] [comments]