Getty Images/iStockphoto
Global telcos team up on large language models
Global Telco AI Alliance comprising telcos from Asia, Europe and the Middle East will form joint venture to build multilingual LLMs that are more attuned to telco domain
A group of telcos from Asia, Europe and the Middle East is planning to build multilingual large language models (LLMs) that cater to telco-specific needs in a bid to drive adoption of generative AI (GenAI) within their businesses.
To that end, South Korea’s SK Telecom, Germany’s Deutsche Telekom, the Emirates’ e& Group, Singapore’s Singtel and Japan’s SoftBank – members of the Global Telco AI Alliance (GTAA) – will establish a joint venture later this year to develop the LLMs that will be used to power digital assistants and chatbots for a start.
Besides being more attuned to the telco domain and better at understanding intent, the domain-specific LLMs will be optimised for languages including Korean, English, German, Arabic and Japanese. The effort will be expanded to other languages such as Bahasa Indonesia so that the LLMs can be deployed in Southeast Asia as well.
Work on the LLMs has started, with customer service data already being used to fine-tune the models to help telco chatbots answer telco-specific questions from customers. The data includes tariff and contract models, as well as information on hardware routers which are rarely found in the training data of general LLMs.
“We as telcos need to develop tailored LLMs for the telco industry to make telco operations more efficient, which is a low-hanging fruit,” said Ryu Young-sang, CEO of SK Telecom. “Our ultimate goal is to discover new business models by redefining relationships with customers.”
Claudia Nemat, board member of Deutsche Telekom for technology and innovation, said that with over 100,000 customer service dialogues a month in Germany already being handled by GenAI, integrating telco-specific LLMs will enable the company’s chatbot to become more human-centric.
Dena Almansoori, e& Group’s chief AI and data officer, expects the LLMs to streamline customer support interactions and deliver personalised recommendations, while Yuen Kuan Moon, group CEO of Singtel, expects the expanded chatbot capabilities to free up service agents to deal with more complex customer issues.
Read more about artificial intelligence in APAC
- The Sea-Lion large language model was built to cater to the language and cultural diversity of Southeast Asia, which is currently underserved by existing models that mostly originate from the West.
- Google Cloud has teamed up with the Singapore government on a slew of initiatives to drive AI adoption, build an ecosystem of AI startups and expand the pool of AI talent in the city-state.
- Malaysia’s Aerodyne is running its Dronos platform on AWS to expand its footprint globally and support a variety of drone use cases, from agriculture seeding to cellular tower maintenance.
- Healthcare professionals at Singapore’s National University Health System can now summarise patient case notes and predict patient healthcare journeys using a large language model trained by a supercomputer.
Some telcos are also looking to offer GenAI capabilities for enterprises. For example, Malaysia’s Maxis plans to integrate its 5G network with Amazon Web Service’s (AWS’s) GenAI capabilities to build cloud services aimed at the retail, manufacturing, logistics and financial industries.
Maxis will also use AWS machine learning capabilities to build, train and deploy AI models at scale, including LLMs trained in the Malay language, to improve customer service and support for enterprise customers.
