BARCELONA: Veon (VON.AS), a prominent telecom company, has joined forces with mobile operator Beeline Kazakhstan, the Barcelona Supercomputing Center, and the GSMA lobby group to tackle the issue of an “AI language gap” prevalent in under-represented languages.
In a collaborative effort announced on Wednesday, the partners highlighted the challenge posed by the limited availability of online data and resources for numerous languages worldwide. They emphasized that while large language models, such as those powering ChatGPT bots, rely heavily on abundant online data to learn and generate human-like responses, such resources are lacking for many languages.
According to the joint statement issued by the collaborating entities, only seven out of nearly 7000 languages spoken globally are considered high-resource languages in the digital realm, including English, Spanish, French, Mandarin, Arabic, German, and Japanese. This disparity in available resources leads to what they term as an “AI language gap.”
The partners outlined their commitment to addressing this gap by focusing on developing tools and language model documentation for under-represented languages. This effort includes languages spoken in the regions where Veon operates, such as Pakistan, Ukraine, Bangladesh, Kazakhstan, Uzbekistan, and Kyrgyzstan. Additionally, Catalan, spoken by approximately 10 million people, was highlighted as a language of focus.
The statement underscored the significance of bridging the AI language gap, emphasizing that the lack of resources in certain languages results in sub-optimal user experiences in AI applications, exacerbates bias in AI models, and risks widening the digital divide in AI technologies.
