IMAGE: Telenor
In the rapidly evolving landscape of AI, one of the most pressing challenges is mitigating biases, reducing hallucinations and improving inference accuracy. As global tech giants race to develop ever more powerful AI models, a crucial advantage may lie not just in data volume but in its diversity.
According to Ieva Martinkenaite, head of research and innovation at Telenor Group, the APAC region has a unique opportunity to lead the way in responsible AI development by leveraging its rich cultural, linguistic and demographic diversity.
Telenor, a Norwegian telecom giant with a 57-year legacy in research, has undergone a significant transformation in recent years.
In 2022, under the leadership of Martinkenaite, the company rebranded its research arm to research and innovation, signalling a strategic pivot toward applied innovation. With operations spanning the Nordics and Southeast Asia, Telenor is leveraging its global footprint to tackle deep tech challenges, particularly in AI and machine learning.
Martinkenaite, who has led this unit for two years, emphasised that the pace of technological change demands a shift from pure research to practical deployment—a move that aligns with Telenor’s ambition to become a “data-first, cloud-native AI company”.
Diversity as an AI strength
Unlike monolithic linguistic and cultural regions, APAC boasts a vibrant mix of languages, dialects and cultural contexts. This presents an opportunity rather than a challenge, according to Martinkenaite.
AI models trained on a wide array of linguistic and cultural inputs are inherently more resilient to biases and less prone to hallucinations—a common issue where AI generates false or misleading content due to gaps in its training data.
“Any large language model that will be trained on multiple languages, multiple texts and multiple cultural features will have the power to hallucinate less or discriminate less,” she told Deeptech Times.
Countries with a single dominant language and a relatively uniform culture face greater difficulties in addressing these issues, whereas APAC’s diversity provides a built-in advantage.
Responsible and context-aware AI
Rather than competing with tech giants in developing proprietary LLMs, Telenor has positioned itself as a responsible deployer of AI that is focused on data integrity, ethical considerations and practical applications.
The company has identified the following key pillars for its AI strategy:
- Responsible AI – Ensuring AI follows ethical guidelines that prioritise human-centric outcomes, mitigate biases and protect user privacy.
- Upskilling and education – Investing in AI literacy and training for employees and partners.
- Collaborative partnerships – Working closely with global tech leaders like Microsoft and Google to co-create responsible AI applications.
- Data governance – Establishing rigorous policies to ensure high-quality, well-structured data.
- Thought leadership – Engaging with regulators to help shape smart, practical AI policies.
- Execution and scalability – Developing AI applications that can be scaled across diverse markets from Norway to Bangladesh.
Addressing hallucinations and biases with real-world applications
One of the most notable challenges with GenAI models is their tendency to “hallucinate” information—that is, to generate text or images that are plausible but factually incorrect. This is particularly concerning in applications like customer service, legal documentation and financial decision-making, where misinformation can have significant consequences.
By incorporating diverse datasets from the APAC region, AI models can develop more robust inference mechanisms, according to Martinkenaite.
“Singapore, for example, has pioneered a LLM incorporating multiple languages and dialects,” she said. This approach ensures that AI systems are not only more inclusive but also better at distinguishing nuances across different linguistic and cultural contexts.
Additionally, Telenor is actively testing AI applications that can dynamically allocate network traffic based on real-time predictions of user demand. By leveraging advanced AI and machine learning techniques, the company can optimise network efficiency, reduce energy consumption and enhance the overall customer experience.
Collaborating with regulators for ethical AI
Governments across APAC have expressed increasing concern over AI’s potential risks, particularly in areas such as cybersecurity, privacy and the spread of misinformation. Martinkenaite acknowledged these concerns and highlighted the importance of collaboration between the private sector and regulators to address them.
“We are bringing practical examples and bridging best practices between Asia and Europe,” she said. For instance, Telenor has played a role in facilitating discussions between Southeast Asian regulators and their European counterparts, ensuring that AI governance frameworks are informed by real-world applications rather than just theoretical concerns.
Telenor’s proactive approach to ethical AI extends beyond compliance; it actively seeks to upskill employees, customers and SMEs on AI safety and security. The company has launched initiatives to train thousands of SMEs on cybersecurity best practices, ensuring that AI-driven threats can be mitigated effectively.
The future of AI in APAC
As AI technology continues to advance, its success will depend not only on computational power but also on the quality and diversity of data used to train it. APAC’s rich and varied cultural landscape provides a unique opportunity to create more ethical, unbiased and accurate AI systems.
Telenor’s approach—which prioritises responsible AI, collaboration with regulators and leveraging the region’s diversity—offers a model for other enterprises looking to navigate the AI revolution thoughtfully and effectively.
“The more diverse the training data, the stronger the AI model becomes. And in that regard, APAC is uniquely positioned to lead,” said Martinkenaite.
By harnessing its diversity, APAC can not only address key AI challenges like hallucinations and inference errors but also shape the future of AI in a way that is ethical, inclusive and transformative.
