Dell SVP of AI Strategy talks through LLM curation, optimization for telco AI use cases
Earlier this year at Dell Tech World, company CEO, Chairman and Founder Michael Dell identified generative AI as among the five big challenges enterprises are facing. The others were the future of work, multicloud, edge computing and security. A key piece here, whether its related to telco AI use cases, or healthcare AI use cases, or manufacturing AI use cases, etc…is taking a large language model (LLM) that was built to a general purpose tool, and layering in proprietary or industry-specific data to optimize the LLM for particular tasks in particular sectors.
In an effort to simplify deployment of LLMs, Dell recently announced a partnership with Meta that will see the former combine its IT infrastructure, software and services with the latter’s Llama 2 family of LLMs. In an interview with RCR Wireless News, Dell SVP of AI Strategy Matt Baker talked through the process of curating an LLM for an individual business, and discussed the benefits of running generative AI on premise as opposed to in someone else’s centralized cloud.
Another key aspect of Dell’s approach to gen AI is the development, with Meta, of pre-tested validated designs. According to the company, “With fully documented deployment and configuration guidance, organizations can get their gen AI infrastructure up and running quickly and operate Llama 2 with more predictability…Our goal is to be the preferred on-premises infrastructure provider for customers deploying Llama 2 and bring the best-of-breed generative AI solutions to our customers.”
Looking into telco AI opportunities, Baker noted ongoing trends around operators pushing compute deeper into the networks to radio sites, customer premises and other locations, alongside the move toward hardware/software disaggregation which is currently a hot topic as it relates to the radio access network.
“A lot of AI inferencing, we believe, will occur at the edge,” Baker said. “The broader world around us has a connectivity problem. And the way to deploy some of these edge inferencing elements may in fact be to deploy them at the network edge. Edge inferencing could be the killer app for these advanced networks.”
Baker also called out overlap between AI and private 5G, which he said “go together very well…A lot of AI inferencing and automation requires a degree of connectivity, and what better way to do that than with telco infrastructure. We think that it’s a big boon to this sort of movement towards more openness in the network so you can deploy more open applications in the network.”
Click here for more interviews in Telco AI Deep Dive series. d
