'How do I reset my router' isn't in LLM corpuses. An alliance of telcos wants to change that
Customer service data from five top telcos across Asia, Europe, and UAE will feed a carrier-centric multi-lingual chatbots
A joint venture comprising carriers SoftBank, Singtel, SK Telecom, Deutsche Telekom and e& Group plans to develop a large language model (LLM) they will use to automate customer service for the telco sector.
The Global Telco AI Alliance, as the entity styles itself, wants to create a specialized LLM because the participants think their effort will yield better results, faster, than working with a general-purpose LLM.
"Tariff and contract models, information on special hardware such as the router, for example (e.g. How do I do a reset?) are rarely found in the general training data of the large models," the group observed in its launch announcement.
The participating carriers have contributed existing customer service data to the LLM, which is "currently being optimized" for five major languages.
- Singtel does the 'we’re building datacenters to host Nvidia clusters’ thing
- South Korean telco SK Broadband and Netflix call a truce in network payment fight
- South Korea's biggest mobile telco says 5G has failed to deliver on its promise
- AI-generated bug reports are seriously annoying for developers
The JV will focus on the needs of its members, and will therefore work initially in five languages: Korean, English, German, Arabic and Japanese. Alliance members will add new languages after reaching mutual agreement.
Singtel operates in Indonesia and India – a pair of polyglot nations. Indonesia's 240 million speakers of Bahasa will likely prove a tempting target, in which dominant language groups have over 200 million speakers. India's main language, Hindi, is spoken by over 500 million people, while Bengali is the tongue of nearly 100 million. Marathi and Telugu are each spoken by more than 80 million souls. e& Group operates in Africa – another region with a potential need for local language LLMs.
Effective and fast customer service is the aim of the Alliance.
"This multi-lingual LLM tailored for telcos will greatly expand chatbot capabilities with relevant responses to customers' technical queries, freeing up service agents to deal with more complex customer issues," reads Singtel group CEO Yuen Kuan Moon's canned statement. "With leading telcos from three different continents working on this innovative model, this unprecedented effort to scale AI development for the telecom industry would not have been possible had we all decided to go it alone."
That last bit is a little more interesting, as it acknowledges that LLM development is not easy or cheap – and that sharing costs and expertise is needed to deliver the kind of experience customers have come to expect since the late 2022 debut of ChatGPT and the subsequent debut of many similar services.
No timeframe for deployment has been announced. Nor has the Alliance signalled any willingness to productize its wares and offer the model to telcos outside the group. ®