Seven Peaks Insights

Code to culture: developing open source Thai AI language models

Written by Seven Peaks | Dec 2, 2024 1:45:05 AM

On Thursday October 24th, Seven Peaks’ Vice President of Data Analytics, Damien Velly and Kasima Tharnpipitchai, SCB 10X’s Head of AI Strategy, uncovered current potential and future possibilities for AI advancements in an open-floor discussion at our latest in-person event: “From Code to Culture: Developing Open Source Thai AI Language Models”.

The event, held at Seven Peaks Headquarters Office in Bangkok, explored the fast changing landscape of AI and the roles it plays in achieving Thailand’s long-term mission to preserve its linguistics and cultural legacies.

Damien set the stage by breaking down technical AI jargons into digestible and easy to understand terms in his talk on the data-powered digital solutions. His talk focused on practical ways to make AI more accessible for business owners and entrepreneurs as well as his proposed solutions to break down barriers limiting AI accessibility in Thailand. 

To Damien, these barriers are technical jargons and deep-rooted cultural fear towards AI and its dominance in the job market. The development of Thai Large Language Models serves as a way to dispel fear for business owners, leaders, and entrepreneurs. However, it is essential that we truly understand the benefits and limitations of AI in decision making in order to move forward with the development. 

“Without quality data, there can be no quality AI”

Damien also stressed the importance of AI in creating a tech-powered future for businesses across multiple industries. By putting business needs for data quality improvement at the center of his talk, Damien urged business owners who are eager to leverage AI in their businesses to improve on the quality and management of their data as they are crucial in the development of a quality AI

He later went on to discuss strategies to help businesses enhance data quality more effectively. One of his proposed methods was the diversification of data sources. By standardizing and diversifying data sources, said Damien, businesses would be able to draw information from various sources, learn from multiple scenarios and come up with smarter solutions.

For the second half of the event, Kasima invited the audience to explore the vast Al landscapes in the context of socio-cultural dimensions in his presentation titled The Future of Ai in Thailand. According to him, Thai LLM is more challenging to develop due to complex writing systems and rich vocal tones found in the modern Thai language. To resolve this, AI needs to be able to produce text at a high level of accuracy. 

However, Kasima remained optimistic about the advancement of Thai language models. He believed that Thai-speaking, and local developers would be able to the full potential of Thai language models. From Kasima’s point of view, local LLMs should be under the supervision and management of local organizations to keep the traditional Thai language and culture alive.

Later in his talk, Kasima discussed his current Typhoon project on AI’s language abilities at SCB 10X. The project focuses on the utilisation of a mixture of MoA and agentic workflows to leverage Typhoon’s ability to understand specific language nuances. 

According to Kasima, the current AI market in Thailand is prone to being overly dependent on less sophisticated AI technologies. However, open source AI models offer a myriad of opportunities for the country to create its own AI-powered future through collaboration and extensive exchange of knowledge. 

The most recent Open source AI application revealed that AI is able to understand local Thai dialects and capable of giving practical advice to local farmers. 

Open source AI is democratizing AI developments and can be integrated into other business sectors. Its ability to understand language can benefit the quality of public services in Thailand. Through agentic frameworks, multiple AI agents worked together to increase workflow efficiency and improve the accessibility of public healthcare and more. 

The event ended with a Q&A session which turned into a series of fascinating discussions about ethical practices in AI adoption and issues with data privacy. Despite the potential challenges that lie ahead, Damien and Kasima remained optimistic about the development of AI in Thailand with open source AI becoming more widely accepted by local talents, businesses, and entrepreneurs. 

Find more events like this here. Follow us on Facebook and LinkedIn to stay up to date on our upcoming events.