Another exciting development is Adobe's project "Stardust". Announced at their recent conference, Adobe MAX, Stardust is a new AI-powered capability allowing users to move, remove or change objects in an image, like the colour or type of clothing a person wears, all in seconds. This is the power of generative AI. It's democratising access to advanced technology, enabling everyone to become adept at many skills with a basic understanding of the information.
And, of course, Southeast Asia (SEA) is catching up in this AI race. Kearney and EDBI predict a significant impact on SEA's economy, with AI contributing US$1 trillion across the region in the next decade. AI is enabling SEA countries to compete on the global stage.
In Thailand, the Bangkok Metropolitan Administration has started using AI to manage the city's infamous traffic. In Vietnam, authorities are deploying AI to identify and flag instances of tax evasion, reducing human error and increasing tax compliance and government revenues. Last but not least, in Indonesia, an aid agency utilised DALL-E to visualise an aspirational image of an inclusive forest city for their future capital city, Nusantara.
However, the path to AI adoption has obstacles. The term "AI" often brings up negative connotations, such as job loss and potential misuse for scams or electoral fraud. The environmental cost of AI tools like ChatGPT, Midjourney, and Stable Diffusion is significant. To train GPT-3, for instance, hundreds of tons of CO2 are consumed.
Despite these challenges, the pace of AI development in tech giants like Google, OpenAI, Amazon, and Meta is relentless. SEA brands and marketers are grappling with how to harness this power responsibly. They ask the ultimate question: "How do we do it correctly?"
Marketers can start with these three immediate steps to answer that question:
1)Upskilling internally and externally
First, improving your understanding of how to leverage generative AI in your day-to-day work is vital. In Singapore, 94% of employees have used generative AI in marketing campaigns, but only 31% report their employers using these tools. To bridge this gap, companies need to upskill hard and soft skills regarding generative AI.
So, starting from where I am in MediaMonks in Southeast Asia, my three fellow AI enthusiasts and colleagues are pioneering an educational series that will scale up AI-expert teams and upskill talent across Singapore, Kuala Lumpur, Japan, and India. This includes teams from different departments, be that creatives, strategists, production, and even legal know-how, on how to utilise generative AI tools. What’s essential is not just to cover the basics, but to focus on real-world use cases, ensuring legal processes are in place.
Within three months, there is momentum in the Southeast Asia team in using generative AI in their day-to-day, from using MonkGPT (our customised ChatGPT interface) to assist in translating or concept ideation to creating our own SG AI Office Tour to show how we can use multiple AI video editing tools. Externally, we invite clients to do regular "AI sessions," creating more relevant and tailored client experiences. The success of these sessions hinges on co-creation together, where everyone not only understands the tools but also how to leverage them most effectively.
2)Clear cuidelines and selective data
Next, we need to address the potential misalignment of generative AI with a brand's values and the issue of data waste. Marketers can develop guidelines for AI-generated content and create review processes to ensure consistency. This might entail crafting distinct regulations for the variety of content that can be produced, setting limits for AI-generated content, and instituting evaluation procedures to confirm that the AI-produced content aligns with a brand's ethos and communication style. Moreover, as part of reducing data waste, selective data and rigorous monitoring are essential. Garbage in equals garbage out, and, the quantity and quality of data both form the bedrock of a successful AI implementation strategy. Identifying the appropriate data sources to incorporate or discard, and determining the right campaigns to pair with each unique automated bidding type, is an essential expertise.
3)“SEA” human expert intervention
It’s not news that generative AI tools can produce biased or misaligned results, especially in Southeast Asia, where AI training datasets are often Western-centric. In her book The Atlas of AI, author Kate Crawford highlights the potential for bias in datasets like ImageNet, which initially used WordNet lexicons for image categorisation, leading to the inclusion of offensive labels that perpetuate societal stereotypes. Similarly, the UTKFace database restricts gender to male or female and race to categories such as White, Black, Asian, Indian, and others. Considering that a mere 12 institutions from culturally-dominant regions generate half of the world's training datasets, it's crucial to involve Southeast Asian human experts. These people possess not only a working knowledge of these tools, but also a fundamental understanding of the resulting outputs. A proper framework and structure that includes Southeast Asian AI experts, together with development through upskilling and clear guidelines, help create more AI experts in the region. My observation is that we need to make this ASAP; otherwise, output creation will become less culturally-specific and will lack diversity, which contradicts the points where generative AI is a creativity enabler.
In summary, the adoption of generative AI in SEA is accelerating. When used responsibly, it can unlock significant value for brands and consumers, creating tailored experiences on an unprecedented scale. Overcoming barriers such as job loss fears, misinformation, and high-cost production requires empowering teams, establishing clear guidelines, and ensuring expert oversight.
Avisenna Gusta is a digital content producer at MediaMonks.