[ad_1]
DUBAI, May 25 (Reuters) – The emirate of Abu Dhabi is working on a large-scale artificial intelligence model “Falcon 40B” that will be open-sourced for research and commercial use, the government’s Advanced Technology Research Council (ATRC) said on Thursday. .
VentureOne, ATRC’s business investment arm, said it would also support viable ideas generated using the model.
The Falcon 40B is a basic Large Language Model (LLM) with 40 billion parameters and trained on 1 trillion tokens, developed by the Technology Innovation Institute (TII), a research center within the ATRC.
Generative AI models are the technology that powers applications like OpenAI’s bot ChatGPT.
“TII is providing access to model weights as a more comprehensive open-source package,” the ATRC said. “While most LLMs only grant exclusive licenses to non-commercial users, TII has taken a critical step in providing researchers and commercial users access to the Falcon 40B LLM.”
Abu Dhabi is the capital of the United Arab Emirates, a federation of seven emirates. The Abu Dhabi government has rapidly developed its technology industry in recent years, including the establishment of the G42 artificial intelligence and cloud computing company and the EDGE Defense Technology Group.
“We want to contribute to the community and accelerate the use of artificial intelligence,” ATRC Secretary General Faisal Al Bannai told Reuters. Bannai is also chairman of EDGE.
TII director Ebtesam Almazrouei said they want to support generative AI not just for chatbots, but also for engineering, healthcare, applicability and coding.
In recent months, as companies around the world race to bring artificial intelligence products to market, concerns have grown over how the technology could lead to privacy violations, scams and misinformation campaigns.
“Deploying these platforms to train on their own parameters means we don’t have access to the data that goes into these platforms,” ​​Bannai said when asked about privacy concerns about the Falcon model.
Reporting by Lisa Barrington; Editing by Jason Neely
Our standards: Thomson Reuters Trust Principles.
[ad_2]
Source link