A groundbreaking advancement in the field of Arabic language artificial intelligence (AI) has been achieved by a collaborative effort between engineers, researchers, and a Silicon Valley chip company. The result is an innovative open-source software called Jais, which possesses a staggering 13 billion parameters to power generative AI applications. What sets Jais apart is its unique composition, combining both Arabic and English data, including a significant portion of computer code.
One of the primary driving forces behind this project was the scarcity of large language models that are bilingual. The team recognized the need for a sophisticated AI language model that can seamlessly process and understand both Arabic and English languages. To accomplish this ambitious goal, the project received support from Cerebras Systems, a Silicon Valley-based chip manufacturer known for its cutting-edge AI hardware designs that rival those of Nvidia.
Named after the highest peak in the United Arab Emirates, Jais is a collaborative effort between Cerebras, the Mohamed bin Zayed University of Artificial Intelligence, and Inception, a subsidiary of G42, a prominent tech conglomerate based in Abu Dhabi with a strong focus on AI.
To overcome the challenge of limited Arabic language data, the Jais model leveraged the logical reasoning abilities encapsulated within the computer code present in the English data. Professor Timothy Baldwin from the Mohamed bin Zayed University of Artificial Intelligence explained that the inclusion of code provided the model with a significant advantage in terms of reasoning capabilities.
In a groundbreaking move, the team made the Jais model available under an open source license. This decision aims to foster collaboration and encourage the development of diverse and innovative applications utilizing Jais across various industries.
The Jais model was extensively trained on the Cerebras’ supercomputer, Condor Galaxy. G42 has recently acquired three of these powerful units, addressing the global supply shortage of Nvidia chips and paving the way for further advancements in AI research and development.
With its vast parameters and bilingual capabilities, Jais holds immense potential for revolutionizing generative AI applications in the Arabic language space. This breakthrough represents a significant step forward in bridging the gap between language diversity and artificial intelligence.
FAQ
What is Jais?
Jais is an advanced open-source Arabic language AI model with 13 billion parameters, capable of powering generative AI applications.
How was Jais developed?
Jais was developed through a collaborative effort between engineers, researchers, and a Silicon Valley chip company. It made use of a combination of Arabic and English data, including a significant portion of computer code.
What is the significance of Jais being bilingual?
Jais is unique as it possesses the ability to process and understand both Arabic and English languages, addressing the scarcity of large bilingual language models.
How did the inclusion of code benefit the Jais model?
The presence of computer code within the English data used to train the Jais model enhanced its reasoning capabilities, providing it with a substantial advantage.
What is the availability of Jais?
Jais is available under an open source license, enabling collaboration and encouraging the development of diverse and innovative applications.
How was the Jais model trained?
The Jais model underwent extensive training on the Cerebras’ supercomputer, Condor Galaxy, which addresses the shortage of AI hardware chips, such as Nvidia’s.