August 30, 2023, G42 a leading artificial intelligence company in the United Arab Emirates (UAE), has launched an open-source Arabic language AI model called Jais.
A massive dataset of texts and code combining Arabic and English was used as the source of the model’s training, and the model contains 13 billion parameters.
In addition to translation, text generation, and question-answering, Jais can handle a variety of tasks. It can also be used to develop new applications in areas such as education, healthcare, and customer service.
Supercomputers from Silicon Valley-based Cerebras Systems were used for the development of the software, which competes with Nvidia’s (NVDA.O) powerful AI hardware. A shortage of NVIDIA chips has forced companies around the world to look for alternative chips.
This is a joint venture of Cerebras and Mohamed bin Zayed University of Artificial Intelligence, which is a unit of the Abu Dhabi-based tech conglomerate G42.
G42 specializes in all things related to artificial intelligence, and its name is derived from the tallest peak in the United Arab Emirates. Jais was created as part of the research project named Science for the Future.
Since there is insufficient Arabic data for training a model of the size of Jais, Professor Baldwin of the Mohammed bin Zayed University of Artificial Intelligence used computer code from the English language data in order to train the model.
In order to train Jai’s model, the group made use of Cerebras’ supercomputer, called the Condor Galaxy. Cerebras announced this year it had sold three of these units to G42, the first of which will arrive this year and the remainder in 2024.
The launch of Jais is a significant development for the Arabic language AI community. It provides a powerful tool for researchers and developers to build new applications that can benefit people across the Arab world.
G42 has made Jais available under the Apache 2.0 license, which means that it can be freely used, modified, and redistributed. By enabling more users to access Arabic language AI, the company hopes to accelerate its development.
Jais has the following additional details:
- There were over 100 billion words in the dataset, including text from books, articles, and social media platforms, that were used in training the algorithm.
- It can translate between Arabic and English with high accuracy.
- It can generate text in Arabic, including poems, code, and scripts.
- The algorithm can answer any question in Arabic, no matter how open-ended, difficult, or strange it may seem.
The technology behind Jais is still in the development phase, but as far as our interaction with Arabic texts is concerned, it is going to revolutionize the way we interact with them in the coming years. It can be used to improve the quality of education, healthcare, and customer service in the Arab world.
As well as being able to create new forms of entertainment and creativity, it can also be used for business purposes.
The launch of Jais is a major milestone for the development of Arabic language AI. It is a powerful tool that has the potential to make a positive impact on the lives of millions of people.
