Cooperation vital if region is to fulfill its AI potential
https://arab.news/jae53
There is a growing focus on digital growth across the Gulf, driven by the region’s young and dynamic population and a desire to utilize artificial intelligence to drive economic transformation. As the region is creatively incorporating AI into its societies, generative AI technologies such as Arabic large language models have emerged as an area of focus. The rise of Arabic large language models is critical not only for language processing but also as a tool of economic competitiveness and to support enduring sociocultural relevance.
Studies show that the Gulf Cooperation Council states alone could gain nearly $10 of economic growth for every $1 invested in generative AI by 2030, leading to an overall economic impact of at least $23.5 billion per year by the end of the decade. Saudi Arabia, the UAE, and Qatar are expected to be the main beneficiaries of this growth, with a yearly estimated impact of $12.2 billion, $5.3 billion and $2.6 billion, respectively. This can be attributed to generative AI’s potential to significantly improve business efficiency, enhance research and development and drive forward digital and linguistic sovereignty.
This complements the region’s ongoing economic diversification agenda. The transformation program has spurred business growth in the Gulf, with the domestic private sector and international investors being more prominent in the region than before. Growing privatization has made the Gulf states more receptive to digitalization across sectors, including education technology, digital governance and digital wallets. It is worth noting that, in their move toward adopting digital technologies and generative AI, the Gulf states seem not to be relying only on established international technology providers, but rather on developing competitive domestic platforms.
Saudi Arabia, the UAE, and Qatar have been particularly successful in developing local Arabic large language models that are fit to compete with international alternatives. The open-source Falcon 40B large language model developed by the Abu Dhabi Advanced Technology Research Council has received international acclaim for its cutting-edge technology and sophistication. Similarly, Jais, developed by the UAE-based technology leader G42, has been touted as the world’s most advanced Arabic large language model.
In September, the Riyadh-based conversational AI provider WideBot AI launched AQL-7B, which is known for its compact and high-performance model. Earlier that month, the Saudi Data and AI Authority launched ALLaM-2-7B, in collaboration with Microsoft Azure, to facilitate natural language understanding in both Arabic and English. These followed the Saudi Data and AI Authority and IBM’s partnership in May to launch ALLaM and Saudi technology company Watad’s release in March of Mulhem, the first large language model trained exclusively on Saudi data sets. Qatar has also prioritized generative AI and in May launched Fanar, an Arabic large language model aiming to advance scientific cooperation in the Arab world and drive innovation.
Beyond its economic benefits, incorporating Arabic in the generative AI framework is also key for the survival of the language.
Zaid M. Belbagi
The commercial potential of Arabic large language models is immense, particularly in fintech and e-commerce. Companies like Al-Rajhi Bank, Amazon MENA and Jumia are already using these models to deliver personalized marketing and improve localized customer support. Tonomus, a subsidiary of Saudi Arabia’s NEOM, has partnered with Nvidia and Oracle to enhance AI adoption in the smart city. While plans for communication between citizens and city services through Arabic large language models, as well as for an Arabic dialect cloud, are in the development stage, this strategy is a step in the right direction. The media industry can also benefit by using these models for automated content creation, audience interaction and subtitling, allowing it to reach broader audiences while maintaining cultural authenticity.
Beyond its economic benefits, incorporating Arabic in the generative AI framework is also key for the survival of the language. Concerns in the Arab world about the survival of Arabic are primarily driven by the growing popularity of English as the language of business and entertainment in the region. Moreover, children and youths in the region have an unprecedented exposure to online content, the majority of which is in English. Even generative AI platforms like ChatGPT, which are today commonly used in the education space, are based on the structure of English. Thus, if the future of Arabic is in digitization, then the region must commit to the future of digitization, which lies in generative AI and large language models.
The region’s growth in the generative AI space has not been without its challenges. The diversity of Arabic dialects poses a notable challenge to the standardization of Arabic large language models. Arabic is spoken in diverse forms across the region, with the Gulf and Egyptian dialects and Moroccan Darija arguably being the three most popular versions. Differences in phonetics, pronunciation and vocabulary demand the creation of context-specific Arabic large language models. While the use of Modern Standard Arabic in large language models is fruitful in the business context, incorporating dialects will increase generative AI’s relevance in the media, culture and customer service industries.
Generative AI’s success in the region will also depend on how willing governments are to adopt digital technologies. State funds in the Gulf have indeed made impressive investments in generative AI, but success will only be realized at the stage of implementation. The Gulf states must also commit to partnering with academia and the private sector, as the development of this technology lies at the intersection of investment and infrastructure, research and commercial scalability. As the field grows, there is also the need for comprehensive regulatory and legal support to those developing and implementing generative AI.
Further, an element of competition has appeared between the various Arab countries that are actively investing in and developing domestic large language models, particularly the Gulf states. For Arabic large language models to be truly successful, however, it is crucial that the Arabic-speaking world works together. There is strong potential for collaborative efforts, given that generative AI is currently dominated by English and significant work remains to be done to expand its reach to a complex language like Arabic. This will require steady investment and collaboration in research and development.
Such challenges are to be expected in the developmental phase of any technological transformation and the region’s continued commitment to generative AI is promising. The various large language models across the Middle East and North Africa will need to learn from and build on each other’s achievements to collectively create a digital future for Arabic.
• Zaid M. Belbagi is a political commentator and an adviser to private clients between London and the Gulf Cooperation Council.
X: @Moulay_Zaid
