(MENAFN- The Peninsula)
Oussama Abbassi
|
The Peninsula
Doha: Qatar's first Arabic-centric Large Language Model AI platform 'Fanar' takes the stage in the first day of Al Jazeera's second AI in media conference held on Satuday, Januarry 11, 2025.
The session brought together experts and innovators in the technology from Qatar Computing Research Institute, where they held a rich dialogue that gave exposure to the locally developed platform.
"We're not competing with large language models like ChatGPT or Gemini," explained Dr Ahmed K Elmagarmid, Executive Director of Qatar Computing Research Institute.
Dr Ahmed K Elmagarmid, Executive Director of Qatar Computing Research Institute during the second AI in Media Conference organized by Al Jazeera Media Institute on Saturday, January 11, 2025. Photo by Oussama Abbassi
"Our goal is to fill a crucial gap - the lack of attention to Arabic language in AI development."
The Arabic language AI platform represents a significant shift for the Gulf region, from being technology consumers to becoming innovators in the artificial intelligence field.
The project, launched in early 2024, emerged from collaborative efforts between Qatar Computing Research Institute, Hamad Bin Khalifa University, and the Ministry of Communications and Information Technology, with support from various Qatari institutions, including Qatar National Library, Arab Center for research and Policy Studies, Al Jazeera, and the Ministry of Endowments and Islamic Affairs (Awqaf).
Dr. Mohamed Eltabakh, Principal Scientist at Qatar Computing Research Institute, highlighted Fanar's distinctive features.
Dr. Mohamed Eltabakh, Principal Scientist at Qatar Computing Research Institute during the second AI in Media Conference organized by Al Jazeera Media Institute on Saturday, January 11, 2025. Photo by Oussama Abbassi
"What sets Fanar apart is that it's a Qatari local technology that properly reflects Arabic societies and culture."
He demonstrated how the platform generates content using Arabic names and cultural references, unlike Western AI models that default to Western-centric perspectives.
A key challenge the team addressed was the scarcity of Arabic content online.
Despite Arabic speakers representing up to 6% of internet users, Arabic content makes up less than 0.5% of online content.
To overcome this, Fanar was trained on a dataset comprising 40% Arabic content, significantly higher than other AI models.
The platform includes specialized components for religious content, image generation, dialect processing, and fact verification.
"We developed our own tokenizer specifically designed for Arabic language, rather than using off-the-shelf technology," Dr. Elmagarmid added, emphasizing their commitment to building Arabic-native AI solutions.
Fanar represents Qatar's broader vision to enhance Arabic digital content while maintaining cultural authenticity and addressing the specific needs of Arabic-speaking users and organizations.
MENAFN12012025000063011010ID1109081170
Legal Disclaimer:
MENAFN provides the information “as is” without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the provider above.