Next-Generation Artificial Intelligence for Emergent Semantic Communications

Hello, welcome. It is May 22nd, 2025. We are in Active Inference Guest Stream 108.1 discussing Next Generation Artificial Intelligence for Emergent Semantic Communications with Christo K. Thomas. So Christo, thank you very much for joining. Looking forward to hearing about the work and also to everyone's comments and questions from the live chat. Thank you. To you. Thanks, Daniel, for the introduction. Yeah. And thank you for inviting me for this talk. So I'll be talking about Next Generation Artificial Intelligence for Emergent Semantic Communication and also I would like to express my gratitude to my collaborators, my postdoc advisor, Dr. Valit Saad and PhD student in our group, Omar Hashash. Yeah. So I would also like to advertise a bit on the upcoming research group I will be establishing as a tenure track assistant professor at Worcester Polytechnic Institute in Massachusetts. So therein, my focus will be to work at the intersection of Next Generation Artificial Intelligence and Wireless Systems with specific focus on resilient and interoperable semantic communication and integrated sensing and communication frameworks, explainable and generalizable AI for Next Generation Wireless Systems, foundation models for networking applications and developing strong mathematical foundations of cognitive and reasoning driven AI architectures. So yeah, so this is an outline of the talk. So first I will go over a motivation of Next Generation AI Native Wireless Networks. Next Generation AI Native Language Networks. So then further I will talk about two of my recent research work on semantic communication and further I will provide some future perspectives which involve emerging language design and using active inference. So yeah, so if you look at the wireless evolution right now we are at the fifth generation of wireless systems, which promised apart from you know cellular enhancing the cellular communication aspects like massive machine type of communication or ultra reliable low latency communication. However, 5G could not meet all the promises it actually made. So yeah, as like the you know saying goes usually the old generation or numbered generations are not very successful at their what they promise. So if that is true, so usually you know the even number of generation so next one will be the 6G so we can expect more exciting developments in 6th generation of wireless systems. So in 6G so we are not okay as we move towards making our environments more smarter through you know making the network intelligent and be able to control the actions of distant autonomous agents in the environment. So in order to support those applications 6G brings in a lot of different more distant set of challenges like you know trustworthiness, security and resilience of underlying AI models, immersivity, sensing, AI nativity and sustainable computing. So here AI is going to play a crucial role in the you know design management or maintenance of different network and device functions. So yeah, so such a transition from the fifth generation to AI native sixth generation is expected as you move to sixth generation of technologies. So here the question is why AI native because it can brings in the aspects of you know adaptability to dynamically varying wireless environment or tasks. It can handle non-linear signals. It can also bring in a lot of resource efficiency compared to classical systems. So these are some of the advantages that AI nativity provides. However, if you look at the state of the art of AI native networks. So currently the literature is predominantly data driven AI algorithms like you know so yeah meta learning or autoencodes, CNNs and right now you know there is a huge high point using large language models for wireless network optimization. However, these data driven machine learning algorithms are plagued by the following disadvantages like large data size which may not be always available readily in a wireless system. Second is…