Abu Dhabi has announced the development of Falcon LLM, a large language model designed to compete with existing models from OpenAI, DeepMind and Google. The model is still in development for a range of applications including language translation, chatbots, content generation and sentiment analysis. The team at the Technology Innovation Institute (TII), a government-backed research hub, has its sights set on OpenAI’s platform and aims to offer the Falcon LLM to government entities, start-ups and the private sector as an alternative to the models from the major tech players.
Foxconn to build AirPods factory in India, diversifying beyond China
The TII is a critical part of Abu Dhabi’s efforts to develop a knowledge-based economy and diversify from a reliance on oil exports. Falcon is a cheaper and faster model than GPT-3 and models from DeepMind and Google, according to an outside performance evaluation by Stanford University. The team touts the quality of its training data, which comprises a final dataset with more than one trillion collected tokens of web data.
Falcon is a 40 billion-parameter language model that is still not commercially available and no timeline has been disclosed. The TII team aims to provide a local alternative to the major tech players in the increasingly competitive and secretive artificial intelligence space. Stanford University’s evaluation will test Falcon’s accuracy, bias and ability to reason, with results expected to be made public in the coming weeks.
