- The telco industry needs access to good, diverse sources of data to realize autonomous networks
- This can be easier said than done
- We're still a long way away from achieving the dream of fully autonomous networks
FUTURENET WORLD, LONDON -- Good data is the bedrock of good telco AI. It's also a key roadblock to realizing fully autonomous networks.
This was one of the main points driven home on the second day of the FutureNet show. Accessing good data that is located in multiple systems is a major challenge for almost every AI system and telco AI is no exception.
“You don’t need perfect data,” stated Anita Tadayon, director, planning, transformation & performance at Virgin Media O2 on Thursday morning. “You need clean data, you need it to be accessible, but you also need it to be diverse."
That – it turns out – can be a big problem.
Tadayon cited an example of hard-to-reach data residing in two different ERP systems. “Those are the sort of typical execution problems you come across,” she noted. The procurement data is all over the place and the contract data is somewhere in spreadsheets. Even when you can access the data, you still have security concerns, she stated.
Various speakers throughout the day noted that many AI trials and proof-of-concepts fail because the systems can’t deal with the various sources of data and the disparate places that data is located. This will be an obstacle to enabling Level 4 or Level 5 autonomous networks, as communication service providers (CSPs) start to reach towards that objective.
As Afnan Ahmed, director of technology, strategy and architecture at Telenor, said in a panel discussion on Thursday afternoon, CSPs will need to have access to a flat “data lake” in order to move ahead on the autonomous journey.
The industry is far from that point now. Ahmed noted that legacy CSPs have many old radio sites that are not regularly updated. “This is the reality,” he said.
To deliver a minimum-touch autonomous network though will require that CSPs have easy access to data access across the organization.
“Can we have one data lake...that empowers all your processes and systems,” Ahmed asked. “We are far from that.”