Sri Lanka needs to develop a Sinhala large language model (LLM) in order to incorporate AI (artificial intelligence), and organisations should critically equip employees with AI-use training, to develop process-oriented AI solutions or face counter-productive AI investments, SLT-Mobitel Chairperson Dr. Mothilal de Silva and Dialog Axiata PLC Group Chief Analytics AI Officer Romesh Ranawana said on Monday (29).
“From Sri Lanka’s perspective, chat voice bots are more important than chat bots,” de Silva said, referring to the Sri Lankan population’s unique preference for engagement with call centres through voice, over engaging with chat bots, at the Sri Lanka AI expo.
SLT also has its own Sinhala chat bot called KITO, however the chatbot, operated through text-based chat on WhatsApp, does not currently support voice interactions.
“So unfortunately we don’t have a Sinhala LLM. Around 92% of call centre traffic comes from voice calls,” de Silva continued, stressing on the high preference among the populace.
“Especially in messaging, people don’t use Sinhala fonts much. It's important to have Sinhala voice bots. That's another barrier that Sri Lanka is facing.”
Another critical local organisational obstacle was raised by Dialog Axiata PLC Group Chief Analytics AI Officer Romesh Ranawana, during a panel discussion, who reiterated that investment in AI has now taken a discouraging turn due to the failure of organisations to incorporate AI within its processes beyond that of technical use.
According to a report released by MIT’s Media Lab/Project Nanda in August, it was found that 95% of investments made in gen AI produced zero returns for organisations.
“The best companies which use AI, the ideas and the systems come from the business, from people on the ground, from people (involved) in the operations,” Ranawana reasoned.
“And that is where it’s really savvy; how do you train your employees to a level where they understand enough about AI to be able to link it to what they do on a day-to-day basis?” Ranawana queried, further explaining that beyond building AI pilots, on-boarding the right infrastructure, data, compute (resources) and governance, there is a critical need for organisations to explore where AI may fit in operational processes, beyond those which are purely technical.
“The idea of where we should fit AI should come from the rest of the business, not the 5% of your technical team.”
Speaking to Fortune Magazine, lead author of the MIT report Aditya Challapally noted that younger demographics of entrepreneurs have seen more success.
According to the report, startups led by 19- or 20-year-olds, “have seen revenues jump from zero to $ 20 million in a year.”
“It’s because they pick one pain point, execute well, and partner smartly with companies who use their tools,” Challapally said.
The findings of the report point out that the 95% failure rate for enterprise AI solutions evinces a “GenAI divide”, whereby the issue is not the quality of AI models, but the learning gap in understanding how to use tools, and hurdles experienced within organisations in incorporating knowledge of tools in processes.
Sri Lanka is to launch a Rs. 100 million artificial intelligence (AI) fund, in a bid to invest in domestic AI-driven startups and research.
Though slated to launch alongside the expo, the proceeds of Sri Lanka’s AI expo is said to go towards the national AI fund, de Silva told media in August. Within Sri Lanka’s digital economy target of growing to $ 15 billion by 2030, AI is to contribute 10-12% of its growth.