GS Paper - III
G42, an AI firm based in the United Arab Emirates (UAE), has unveiled a new large language model (LLM) called NANDA that will specifically cater to Hindi-speaking users. Named after one of India’s highest peaks, NANDA has been jointly developed by G42 subsidiary Inception and Cerebras Systems along with researchers from the Mohamed bin Zayed University of Artificial Intelligence in UAE.
More about NANDA
- NANDA will have a parameter count of 13 billion, and has been trained on approximately 2.13 trillion tokens of language datasets, including Hindi.
- The startup introduced the Hindi AI model at the India-UAE business forum in Mumbai during Crown Prince of Abu Dhabi, Sheikh Khaled bin Mohammed bin Zayed Al Nahyan’s recent State visit to India.
- The model was trained on Condor Galaxy, one of the world’s most powerful AI supercomputers for training and inferencing, built by G42 and Cerebras.
- G42 has a strong track record in the development of language and domain-specific LLMs.
- With NANDA, we are heralding a new era of AI inclusivity, ensuring that the rich heritage and depth of Hindi language is represented in the digital and AI landscape.
- NANDA exemplifies G42’s unwavering commitment to excellence and fostering equitable AI.
Flashback
- In August 2023, G42 launched an AI model called JAIS that was billed as the world’s first open-source Arabic LLM.
- With models ranging from 590 million to 70 billion parameters, JAIS set a new standard for linguistic AI which G42 now seeks to replicate for other regions whose languages are still underrepresented.
- Microsoft recently invested $1.5 billion in G42, whose health tech subsidiary also rolled out two upgraded models of Med42 LLM, an AI assistant for clinicians, healthcare administrators, insurers and patients, earlier this year.
