The five telcos behind the Global Telecom AI Alliance (GTAA) – SK Telecom, Deutsche Telekom, e& Group, Singtel and Softbank – took to the stage at DTW24-Ignite alongside their hyperscaler partners to explain their telco-specific business vision for LLM.
GTAA members want to use the LLM to identify and solve telecommunications company-specific business problems using local language, industry-specific terminology and their own internal business vocabulary – none of which can be done quickly or easily with a generic LLM.
“Simply adding engineering prompts to a generic AI LLM is quite a hassle. It requires a lot of tuning to get consistent outcomes and accurate results, especially for highly complex use cases,” said William Woo, group chief information officer and chief digital officer at Singtel.
According to Wu, Singtel concluded that its “customised telco LLM is clearly superior to the generic LLM for telco-related tasks”.
Another objective of the partnership is to ensure that communications service providers (CSPs) can benefit from future AI business models built using communications data.
“We discussed [in the past] “When we think about hyperscalers and the disruptive role they've played in telecom infrastructure,” said Harrison Lang, group chief strategy officer at e&. “In an AI world, how do we ensure that telcos not just have a seat at the table, but true ownership of the value pool that's created from that? We felt it would be best to work together to build this telecom LLM.” [in which] We each have our own data sets and assets.”
At the event, the GTAA founders announced a joint venture agreement to jointly develop and launch a multilingual Telco Large Language Model that will support a range of languages including Korean, English, German, Arabic and Bahasa. Together, the two entities will equally invest in the initial working capital required for the development.
The joint venture will explore the deployment of innovative AI applications tailored to the founding parties' needs in each market across its global customer base of approximately 1.3 billion people across 50 countries.
Striving for personalization
Customer care is one of the first areas where carriers have leveraged GenAI and is an initial focus for GTAA members.
“By obtaining an industry-specific Master's in Law that provides a comprehensive understanding of our diverse product lines, we will be able to present personalised offers to customers that suit their preferences,” Singtel's Wu said.
But CSPs are looking further ahead.
For example, e& is looking at how to “embed AI into all our existing operations and business models,” Lung said.
“We are a communications company and a technology company. How can we embed AI capabilities in these assets? [for example] “We want to build a recommendation engine that's better than the Netflix equivalent that we have,” Lung added. “How do we incorporate AI into Careem, the local super app, so that the next time you open it, it's customized for you and customer-centric.”
SK Telecom, meanwhile, believes the service personalization enabled by GenAI will spur the development of a third-party ecosystem that will feed into its AI data center ambitions. (SKT is working with Anthropic on LLM, investing $100 million in the startup last year.)
“With the participation of various third parties, we hope to provide a more premium and personalized experience for our users while also increasing our influence within the … ecosystem,” explained SG Chung, SK Telecom's chief AI global officer. “On the enterprise side, we are expanding our infrastructure business with AI data centers. With the sheer number of end users passing through our … platform, we believe there will be increased demand for AI data center infrastructure.”
Jung added, “SK Telecom is proactively leveraging the synergy between these two business areas. I want to emphasize that we can generate revenue through the AI data center itself. In addition, SK Telecom is [can run their] “By reducing operational costs, we will be able to drive our AI business more efficiently.”
Achieving automation
Other telco-specific uses of GenAI include improving employee productivity, field operations and network operations.
Speaking about the autonomous network level at TM Forum, Sameer Vuyyuru, Director, Telecom Global Business Development, AWS, said: “We believe that generative AI can enable Level 4 automation two to three years sooner than using traditional ML. [machine learning] That's why we're focused on improving our network operations and field operations.
Meanwhile, AWS is working with BT to provide coding assistance to around 1,200 network engineers.
“So far, we have automated about 12% of tedious, non-value-added tasks and eliminated over 100,000 jobs. [lines of] “We wrote the code in the first four months,” Vuyyuru says.
AWS predicts that the industry will develop domain-specific, language-specific, and task-specific LLMs.
“The future will be about as much interaction between models as possible,” Vuyyuru says. “So models for network operations will talk to customer operations. That's the direction we're heading.”
Telecommunications companies are not the only ones investing in customized LLMs.
“We are currently working with probably about 20 different sovereign governments to build law master's programs specific to their country or region,” Vuyyuru said.
Microsoft and Google, which participated in the presentation, are looking forward to further development of GenAI in the communications industry.
“As an industry, we're still just scratching the surface of what's possible here,” said Ankur Jain, telecom VP for distributed cloud at Google Cloud.