As Artificial Intelligence (AI) and generative AI technologies evolve, human interaction with machine intelligence systems will change. In the last decade, we observed the transition from keyboard interfaces to natural human communication platforms using speech. Examples include speech-to-text (STT) for text messages and information systems such as Siri, Google, and Alexa. With large language models (LLMs) making significant leaps in natural language understanding, we are witnessing another jump forward in the useability of AI for both technical and nontechnical users. The frontier of a scenario where AI can comprehend nuances in human language has arrived. Generative AI, the branch of AI designed to generate new data, images, code, or other types of content that humans do not explicitly program, is rapidly becoming pervasive across nearly all facets of business and technology.
Based on the previously published Dell Validated Designs for Inferencing and Model Customization, this solution is the first to address a full-stack use case, from the server, storage, and networking hardware to a cross-industry solution that implements a digital assistant solution based on generative AI.
A digital assistant is a 2D or 3D representation of a persona that acts as a “co-pilot” or assistant to human users. This AI-powered avatar extends the traditional chatbot and voice experience because it can mimic various human body languages. It interprets clients’ input and provides factual responses and appropriate nonverbal reactions. Digital assistants can connect to any domain or company-specific dataset to share knowledge. They interact using verbal and nonverbal cues such as tone of voice and facial expressions. Digital assistants are accessible 24/7, making it possible to re-create natural human interaction at scale. For customer service, digital assistants can provide client support and coaching while maintaining a human-like emotional connection. They are designed to offer the best of both AI and human conversation, enhancing the customer experience by providing personalized, real-time, and data-driven interactions.
An LLM is at the core of a digital assistant. It is an advanced type of AI model that has been trained on an extensive dataset, typically using deep learning techniques that can understand, process, and generate natural language text, voice, and images. However, AI built solely on public or generic models is not well suited for enterprises to use in their business as they have been trained on general knowledge and do not have the domain-specific knowledge needed for the customer scenario. Enterprise use cases require domain-specific knowledge to train, customize, and operate their LLMs. However, training LLMs from scratch is difficult, both from a skill and a compute capacity perspective.
Retrieval Augmented Generation (RAG) is an advanced AI technique that optimizes the output of an LLM by referencing an authoritative knowledge base outside of its training data before generating a response. It is an architectural approach that improves the efficacy of LLM applications by retrieving relevant data or documents pertinent to a task and providing them as context for the LLM. This process helps to mitigate the unpredictability of LLM responses and ensures that the generated content remains relevant, accurate, and useful in various contexts.
Dell Technologies has designed a scalable, modular, and high-performance architecture that enables enterprises to create a range of generative AI-based digital assistant solutions that apply specifically to their businesses, reinvent their industries, and give them competitive advantages.
This guide describes the Dell Validated Design for Generative AI Digital Assistants for an on-premises enterprise environment. It is based on high-performance Dell and NVIDIA infrastructure and uses cutting-edge technologies, such as RAG for secure information retrieval based on the Pryon Information Retrieval Platform and the 2D/3D rendering capabilities of the Digital Human Platform by UneeQ.
This Dell Validated Design for on-premises digital assistants is the first application-level solution in a series of designs for generative AI that focus on all facets of the generative AI life cycle, based on designs for model training, customization, and inferencing. This solution, which is implemented along with a Dell Professional Services engagement, enables the widespread industrialized rollout and adoption of next-generation workforce automation capabilities.
