Telecommunications giant Orange’s logo displayed at Mobile World Congress 2024 in Barcelona, Spain.
Joan Cross | Nurphoto via Getty Images
French telecommunications giant orange said on Tuesday that it was working with Microsoft– Supports OpenAI and Facebook owners Yuan Build customized artificial intelligence models to better understand African regional languages.
Orange said it is working with OpenAI and Meta to develop customized artificial intelligence models based on their respective Whisper and Llama open source artificial intelligence models. These models are open and available systems that can be adapted to specific needs and can understand the inability of most conversational systems. Understand West African languages.
Currently, most of the data that major AI companies train their algorithms on comes from the United States, which means their models can lose important context, such as culture and language, when it comes to regions as diverse as Europe, the Middle East, and Africa.
Steve Jarrett, chief AI officer at Orange, said this means these models may have difficulty understanding text and speech-based communications composed of poorly represented languages.
“With an open model, you can do what’s called fine-tuning, which is introducing additional information to the model that wasn’t included when it was first trained,” Jarrett told CNBC. “We’re increasing recognition of languages in West Africa that currently don’t have any Artificial intelligence cannot understand these languages.”
Orange plans to first launch an artificial intelligence model covering two West African regional languages, Wolof and Pural, which will be spoken by approximately 16 million and 6 million people respectively by early 2025.
Wolof is a language spoken in Senegal, the Gambia, and southern Mauritania, while Pular is spoken primarily in Senegal.
The company said that the open source artificial intelligence model will be provided externally by Orange and will be provided with a free license for non-commercial purposes such as public health and education. Orange plans to expand its customized AI model program to eventually cover all 18 West African countries.
“We operate in West African countries and our contact centers speak many local languages, but current AI models cannot understand what these people are typing or saying,” Jarrett told CNBC.
Orange’s head of artificial intelligence said that major large language models such as OpenAI’s GPT, Meta’s Llama and Anthropic’s Claude are not well suited to African needs because they are not specifically trained on data from the region.
“Sovereign Artificial Intelligence” Promotion
The move capitalizes on a concept that is gaining traction globally, namely “Sovereign Artificial Intelligence”.
The term refers to the idea that countries and regions should seek greater control over the core technical infrastructure on which artificial intelligence systems rely by localizing data storage and processing to ensure that they represent specific languages, cultures, and cultures. and history.
Orange also hopes to localize data processing and hosting of OpenAI models in European data centers. Orange said this will give it early access to OpenAI’s latest and most advanced artificial intelligence models and help it build new applications, such as artificial intelligence voice systems for customer service.
Jarrett said Orange is committed to using artificial intelligence “responsibly” given the environmental concerns associated with the technology’s massive energy needs, and “not always using massive large language models (LLMs) to solve every problem “.
In addition to using AI systems to improve customer service, Orange is also using the technology to improve a core part of its business: its mobile network.
“On the network side, we use artificial intelligence to not only optimize how we plan our network, but how we operate the network properly,” Jarrett told CNBC.
“The amount of data coming from all the devices on the network is huge, and with the help of artificial intelligence systems, we can help identify patterns in the data that help us identify and predict failures before our customers notice.”
About The Author
