Table Of Contents
In a pioneering move to advance Arabic language artificial intelligence (AI) technologies, Inception, a subsidiary of the G42 group, announced today the launch of a new large language model, “Jais 70B,” along with 20 additional models to support natural language processing (NLP) in Arabic. This initiative is part of the company’s efforts to foster innovation in AI and expand its applications within the Arab world.
Specifications and Objectives of “Jais 70B”
The “Jais 70B” model boasts 70 billion parameters, making it one of the largest language models specifically developed to support the Arabic language. The model aims to accelerate the integration of generative AI services across various sectors, including customer service, content creation, and data analysis. It is specifically designed for developers of natural language processing solutions, enhancing technological capabilities and improving performance in multiple fields.
Comprehensive Range of “Jais” Models
In addition to “Jais 70B,” Inception unveiled a suite of meticulously trained “Jais” models, comprising 20 models of varying sizes ranging from 590 million to 70 billion parameters. These models have been trained on up to 1.6 trillion linguistic units in Arabic, English, and programming data, making them ideal for chat and interactive language applications.
Fostering Innovation and Research in the Arab World
The “Jais” model suite aims to accelerate opportunities for innovation, development, and research within the Arabic-speaking and bilingual communities. Dr. Andrew Jackson, CEO of Inception, emphasized that AI has become a valuable asset and that large language models are at the forefront of adopting this technology. Jackson stated, “Jais was created to preserve the heritage, culture, and language of the Arab world, making AI accessible to everyone.”
Commitment to Advanced Technologies
Jackson added that the launch of “Jais 70B” and the new model family underscores the company’s commitment to providing high-quality AI models for Arab nations. He noted that Inception’s training and adaptation technologies for Arabic language models are scalable to include other underserved languages.
Model Development and Training
Neha Singhopta, a Principal Applied Scientist at Inception, pointed out that models reaching up to 30 billion parameters were trained from scratch, consistently outperforming modified models in the community. She explained that the computational complexities and environmental impact of training models with 70 billion parameters and above were significant, reflecting the challenges the company faces in developing these large-scale models.
Conclusion
With this new initiative, G42, through its subsidiary Inception, reaffirms its leadership in the AI field and strengthens its efforts to support Arabic language and culture by providing advanced technological solutions. This initiative reflects the company’s commitment to a more innovative and progressive future for the Arab community.
Source: Emirates News Agency (WAM)
