In 2023, the Advanced Technology Research Council (ATRC), an entity of the Abu Dhabi government, invested substantial funds in developing a series of large language models. Faisal Al Bannai, the ATRC’s secretary general, made the strategic decision to make these models freely available online. His reasoning was that if these models were as robust as internal testing indicated, it would enhance the UAE’s reputation in the field of AI and attract talented individuals. The strategy proved effective. The models, aptly named Falcon after the UAE’s national bird, were considered among the most advanced open-source models upon their release, surpassing offerings from tech giants like Meta and Google. “Two years ago, how many people would place the UAE on the AI map? Not many,” Al Bannai remarked to TIME in February. Falcon dramatically changed that perception.
Al Bannai emphasizes that the focus now lies in multimodality and achieving enhanced performance with reduced resource demands. To that end, the ATRC unveiled two smaller models earlier this year: one with visual capabilities and another built on a novel architectural framework. Al Bannai shares his vision of developing a larger, multimodal model that will compete head-to-head with OpenAI’s GPT-4o.
He asserts the importance of countries having a compelling open-source alternative to the “proprietary AI solutions offered by major players.” In the UAE, a version of Falcon is being integrated into the nation’s healthcare system. Furthermore, several countries, including Serbia, Uzbekistan, and the Brazilian state of São Paulo, have embraced Al Bannai’s open-source philosophy, entering into agreements with the UAE to utilize its models.
Al Bannai’s journey in the tech industry began in the early 2000s, importing smartphones to the Middle East. At that time, he recognized that the country was “not yet ready to fully capitalize” on the internet boom. However, having dedicated the past two decades to digitizing its government, Al Bannai believes the UAE is well-positioned to leverage AI’s potential. He attributes part of this advantage to the country’s openness to using private data, such as hospital records and data held by state-backed industries, to train future proprietary models for government and enterprise clients. “We can all access data from public sources. The true value lies in the private data,” he concludes.