Thursday June 6, 2024

Bridging the AI Language Gap: An Interview with the Barcelona Supercomputing Center

Resource thumbnail

The GSMA Foundry has embarked on an exciting initiative to showcase innovation in the telecommunications sector, aiming to counter the perception that the sector lacks rapid innovation.

Through a series of Innovation Spotlight interviews, the GSMA Foundry aims to highlight groundbreaking developments and investment opportunities, with over 700 mobile operators globally seeking new innovations. This initiative serves as a pipeline to demonstrate that the telco sector is indeed innovative and vibrant.

A video conference screen with four participants: Rich Cockle from GSMA, Faisal Zia, Technical Director at GSMA Foundry, and Albert Cañigueral and Mariona Sanz from Barcelona Supercomputing Center. There's a "Previously live" label and the GSMA Foundry logo.

The Power of Generative AI and Multilingual Challenges

The first Innovation Spotlight interview dives into the world of Artificial Intelligence (AI) and high-performance computing (HPC) with a special focus on language technology. The fascination with this technology has grown significantly, especially with models like ChatGPT becoming mainstream.

However, there’s a critical challenge: the generative AI models excel predominantly in a few languages, leaving over 7000 other languages underserved. This language gap is a key issue the GSMA Foundry aims to address by leveraging the capabilities of partners like the Barcelona Supercomputing Center (BSC).

Introducing the Barcelona Supercomputing Center (BSC)

Mariona Sanz Ausàs, Director of Innovation and Business Development at BSC, and Albert Cañigueral Bagó, AI and Language Technology Manager, provided insights into their work. BSC, established in 2005, is a leading HPC (High-Performance Computing) center in Europe. It hosts Marenostrum 5, one of the world’s most powerful supercomputers, used for various research applications, including climate change, disease curing, and training large language models (LLMs).

Bridging the AI Language Gap: Challenges and Opportunities

The AI language gap refers to the disparity in AI technology development and application across different languages. While English and a few other high-resource languages see significant AI advancements, most of the world’s 7000 languages lack sufficient digital data, hindering AI system performance. This gap impacts system accuracy, cultural representation, cybersecurity, and overall digital inclusivity.

By working together, BSC and GSMA Foundry aim to create scalable solutions that can be replicated in different contexts, ensuring that AI technologies are inclusive and accessible.

Project AINA: Promoting Catalan in the Digital World

Project AINA is a notable initiative by BSC, funded by the regional government, aimed at promoting the Catalan language in the digital space. The project seeks to develop an open infrastructure, making models and tools accessible to the public, fostering an AI-driven language economy. The collaboration involves not only the government but also civil society and various organisations dedicated to the Catalan language.

Partnership with VEON and Beeline Kazakhstan: Developing Tools for Under-Represented Languages

BSC and the GSMA Foundry have partnered with VEON, a global digital operator that provides converged connectivity and online services, and Beeline Kazakhstan, the leading player in the Kazakhstan’s telecoms industry, to explore collaboration on the development of AI tools and language technologies for languages with low digital resources, including the local languages of the countries where VEON operates.

The Role of Open Solutions in Democratising AI

A significant takeaway from this collaboration is the emphasis on open solutions. Open models and data infrastructures ensure that AI technologies are accessible and adaptable across different languages and regions. This openness is crucial for democratising AI, making it a tool for many rather than a few.

The collaboration between BSC and GSMA Foundry exemplifies how open solutions can address the AI language gap effectively.

Looking Ahead: Integration of HPC and Telecoms

The integration of high-performance computing and telecommunications is a promising frontier. BSC’s expertise in training AI models combined with the telecommunications sector’s data management capabilities can lead to innovative applications.

This collaboration is a win-win, as telecom operators benefit from advanced AI tools, and HPC centres get real-world use cases to refine their models.

Conclusion

The GSMA Foundry and Barcelona Supercomputing Center’s initiative is a testament to the power of collaboration in addressing global challenges. By focusing on open solutions and fostering innovation, they are paving the way for a more inclusive and advanced AI landscape.

The journey to bridge the AI language gap is long, but with sustained efforts and partnerships, significant progress can be made.

Key Takeaways from the interview

  • The AI language gap is a global issue with widespread implications.
  • Collaborative efforts, like those between BSC and GSMA, are essential in addressing this gap.
  • Open solutions and transparent AI development are key to ethical and effective technological progress.
  • The AI language gap also presents significant business opportunities, fostering new startups and economic growth.

Watch the full interview here https://www.linkedin.com/events/gsmafoundryspotlightseries-your7198321939826778113

Stay tuned for further developments from the GSMA Foundry. If you have any questions or want to join this exciting journey, feel free to reach out.

Together, we can drive forward the AI revolution, ensuring it benefits all languages and communities worldwide.