# Beyond Borders: Singapore's Pioneering March Towards Inclusive AI

The growing field of artificial intelligence (AI) and, in particular, the field of generative artificial intelligence (AI) needs to be as diverse as possible. Because in fact, in the world of generative AI, the problem is the same as in the world at large: how to be truly inclusive, and truly international. Singapore, situated at the centre of Southeast Asia, is leading an effort to make sure that the world of large language models (LLMs) reflects genuinely the cultures, languages and contexts of the world.

The ROGUE Challenge of Global Representation

As we are embracing AI as the core of innovation in the digital age, we see that it is walking on a tightrope towards a great divide. There is one glaring issue – while tech is becoming increasingly diverse, the underlying LLMs are not. And without diverse LLMs, many major populations are sidelined. In Southeast Asia alone, more than 692.1 million people speak hundreds of languages and the LLMs that power today’s world are overwhelmingly rooted in Western-centrism.

Singapore's Strategic Response

This landscape is not simply being observed by Singapore, but being deliberately created by it. Aware of the critical importance of foundation models to reflect a population with diverse identities, it is at the forefront of developing a Southeast Asian-focused LLM, increasing cultural and linguistic coverage while ushering a new era of data inclusion.

SEA-LION: A Beacon of Innovation

Taking the lead, Singapore has cultivated its own open-source LLM, called SEA-LION (which stands for Southeast Asian LION), through the national-level incubator company AI Singapore (AISG). Since the work of SEA-LION is moving toward smaller, faster models, the shift will hinge on having models that are locally attuned to the little differences of human interactions in Southeast Asian cultures and languages. SEA-LION is the beginning of a rogue wave.

Cultivating a Diverse AI Ecosystem

The search for a truly diverse AI does not stop with SEA-LION. Singapore is calling for a ‘spirit of cooperation’ to invite other countries, research bodies and industry partners to augment capabilities within the model. This reflects the vital importance of regional cooperation as a guiding principle for the development of AI in the coming years, allowing regional tag-teams like SEA-LION to meet the challenges and harness the opportunities of Southeast Asia’s differences.

Embracing Asian Foundation Models

But foundation AI models power much more than linguistic representation – they’re central to ideas about the ownership and control of data, improving public services, and the growth of the economy generally. India is on this path too. Its homegrown models must be interpreter-friendly and understand the local context The broader importance of ‘homegrown’ models is that they fit into a shift: there’s a growing sense that foundation models, especially those developed in Asia, might perform better on tasks related to local languages, cultures and contexts.

Fostering Hybrid AI Platforms

The diverging story of AI in the Asia-Pacific involves a hybrid model that draws on the best features of regional and global foundation models – in a manner that doesn’t violate local data privacy laws. It also has the potential to benefit from the multilingual and multicultural nature of the region’s businesses. The ecosystem surrounding local models is accelerating. And as these models become more popular, they will enrich the assistance available to smart companies that want to use these smart new forms of AI.

The Journey Towards Responsible AI

with discussions of rogue AI models growing more common, the need to stay alert, if not to curb the wild wild West, is clear if we want innovation to avoid spilling out of controlling. Balancing innovation and responsibility in AI applications is key in Singapore, where human supervision is required from the recruitment process to credit scoring. The emphasis on needing frameworks that allow for the responsible use of AI drives towards a wider demand for putting the best application of the technology for society’s benefit into practice, facilitating more accrued benefits from AI and increasing the public’s trust.

The Future of AI: A ROGUE or a Companion?

The international community faces a make-or-break choice. An imaginative future at the nexus of AI and humanity is possible – when every element in the human AI ecosystem is inclusive and responsibly operated – but it involves precarious points and bumpy roads. Making sure that AI is used for the benefit of everyone means dealing with data from all walks of life; and making sure it does not run amok means careful consideration of potential risks and the nature of rogue behaviours. It is clear that the world, and Singapore in particular, is tackling these issues head-on.

Exploring ROGST UE: Understanding AI's Unpredictable Side

So when we are confronted with the possibilities embodied in the capabilities of AI and the range of its development as it progresses, rogue is a word that makes us think about the wild card of autonomous systems. We also see it used to describe deviant behaviour of foundational models presently or in the future that have strayed off-course from what their programmers intended. It calls out the need for robust regulatory mechanisms and ethical frameworks, and international collaboration and cooperation. It points us in the right direction for confronting rogue AI as we shape an AI future that respects our freedom to be who we are, and a world where our nations can collaborate in developing AI that we can trust.

So, what do we conclude? Singapore’s project of building more diverse and inclusive AI models is an essential first step in correcting biases inherent in current AI systems. Creating and promoting models such as SEA-LION doesn’t just ensure that AI technologies are able to speak and engage with the diversity of languages and cultures found in Southeast Asia. It sets an example for the rest of the world to follow. The road to a more inclusive AI will be full of obstacles. But, with continued innovation, collaboration and governance, AI has the potential to be a new force for good in our diverse and multicultural world.

Jun 14, 2024
<< Go Back