Khalifa University launches a linguistic model for artificial intelligence in the radio frequency domain
ABU DHABI, 6 April, 2017 (WAM) – The Institute for the Digital Future at Khalifa University of Science and Technology announced today the launch of a language model for artificial intelligence in the radio frequency band, the first-of-its-kind Generative Pre-Trained Transformer in the radio frequency band (RF-GPT), capable of interpreting wireless signals, in a way that supports the telecommunications sector.
The new model has shown steady improvements in performance on RF spectrum tasks, outperforming existing base models by up to 75.4%, demonstrating an in-depth understanding of RF on the part of its developers.
Another distinction was to correctly count the number of signals in a spectrum in 98% of cases, a feat that almost all general-purpose AI models have failed to achieve.
The new model works by converting radio signals into visual patterns that artificial intelligence can understand. Once the conversion process is complete, AI systems can analyze these patterns and answer questions about what is happening within the wireless spectrum using simple language.
This foundational model contributes directly to the UAE’s Artificial Intelligence Strategy and lays the foundation for more intelligent and autonomous wireless networks.
Researchers from Khalifa University developed this new model project, led by Professor Marwan Dabbah, Senior Director of the Digital Future Institute.
Professor Ahmed Al Durrah, Deputy Academic Director for Research and Development at Khalifa University, said that the launch of the “RF-GPT Generative Pre-Trained Switch” model reflects the university’s long-term focus on innovation in digital infrastructure to drive progress on integrating artificial intelligence into strategic sectors and the next generation of communication research, in line with national priorities. Initiatives like this model contribute to enhancing the rapidly growing human capital in the UAE and developing the research capabilities necessary to support the country’s advanced digital system.
For his part, Professor Marwan Dabbah said, “The RF-GBT Generative Pre-Trained Switch model represents a turning point in spectral intelligence, as it moves from isolated radio frequency supply lines with specific tasks to a unified interface for radio frequency languages, which is a decisive step towards sixth generation networks supported by artificial intelligence.”
The RF-GPT/Pre-Trained Generative Switch model has been trained using approximately 625,000 computer-generated radio signal examples and is designed for telecommunications network operators and network and network engineering teams involved in spectrum orchestration, supporting increasingly complex wireless networks.
The model demonstrated strong performance in tasks such as identifying signal patterns, detecting broadcast interference, identifying wireless standards, monitoring device use in Wi-Fi networks, and extracting data from 5G device signals.
- For more: Follow Khaleejion 24 Arabic, Khaleejion 24 English, Khaleejion 24 Live, and for social media follow us on Facebook and Twitter
