Ashwani Saigal, SVP Product Management at Vantiva, delves into the transformative impact of AI-powered home gateways and how edge AI applications empower Network Service Providers to boost service efficiency and enhance subscriber satisfaction.
Q1 – 00:25 – Will the gateway become the intelligence point within the home by hosting the household A.I?
It’s a very good question let me first explain the nature of the gateway in the home. The Gateway as you all know is a focal point in the home that is carrying the traffic to and from the home. With the very singular nature of this gateway, it provides the ideal point in the home for service providers and the end users to use that as a gateway for AI management in the home. In addition to that the AI gateways that we are going to be putting on our road map have the capabilities in the NPUs as well as the CPBU capabilities that can help in having an edge AI at home.
Q2 – 01:17 – What are some of the most emblematic applications of edge A.I in the broadband CPE that are going to be deployed in the future?
We can divide the use cases and applications for edge AI into two main buckets: the first bucket is focused more on the OPEX saving for the operators. The second bucket of applications are more focused towards the end user experience. Let’s dive deeper into the first bucket. Because of the focal nature of the gateway in the home with the NPU capabilities it can do a very good job in analyzing the streams destined to and from the gateway and help operators using the inference models, to predict and avoid any helpdesk calls or costly truck roads and manage their networks and CPE in a much more efficient way than they would be without the help of the AI. The second bucket of the user experience is more focused on security, privacy, and enhanced user experience. This can be enabled with a smartphone app which is used on the Gateway and that can help manage all the AI endpoints in the home like smartphones, smart PCs, IoT devices and multimodal set-top boxes. The gateway can provide the user the capability to monitor and manage all the AI endpoints in the home.
Q3 – 03:00 – Is edge A.I. enough or do we need a combination of cloud and edge A.I?
Edge AI is a critical piece of the puzzle but it’s not sufficient in itself, the reason being that the edge AI has a limited amount of CPU as well as memory and NPU, while the cloud has enormous capabilities of processing power. But the balance to be had is the cloud AI uses a lot of power and is not very efficient as far as the accessibility to the AI capabilities which are needed only on the edge. And in that regard the edge AI is sufficient to have some applications addressed locally in the home that does not need a cloud AI. Now we have seen use cases where the edge AI is tiered on top with a cloud AI of a service provider especially for the services that the service provider has full control over like streaming video, and other content which they do not have to go to the public AI cloud. And if you layer it up with edge AI the service provider AI cloud and then the general availability AI Cloud where you have models like Chad-GPT and other AI models running so we see use cases depending on again the need of the accessibility of information it can go from edge AI, to to operator AI to Cloud AI.
Q4 – 04:44 – Why and how Vantiva is best positioned to assist NSPs in successfully implementing A.I. to enhance their services and operations?
So Vantiva is promoting our next generation gateways that have the NPU the Neural Processing Unit capabilities in the gateways in addition to the CPUs that are needed to have the packet processing that is needed to offer a reliable broadband service. These NPU capable gateways have different level of TOPS [Tera Operations Per Second] that are needed to offer the services both for the end user as well as an efficient network management. We are working very closely with multiple SOC vendors and operators to define these AI use cases, and in some cases we are working closely even to define the models the large language models or the small language models that will be adapted for use cases in the home not only as an end device but also as a network piece of equipment at the home.