Gupshup Debuts Domain-Specific Generative AI LLMs
Enterprise conversational AI developer Gupshup has introduced a set of large language models fine-tuned for specific business uses and industries. The new ACE LLMs leverage third-party foundation models like Meta’s Llama 2 and OpenAI’s GPT-3.5 Turbo to assist enterprise development of conversational generative AI for marketing, commerce, and customer support.
LLMs can generate realistic human conversation, but Gupshup sees a deficiency in the precision and governance needed for enterprise deployment. ACE provides the bridge to improving the standard LLMs with pre-trained models sized between 7 billion and 70 billion parameters and speaking more than 100 languages. The models are adapted to different functions and verticals that incorporate safety guardrails and compliance controls suited for real-world business applications.
According to Gupshup, ACE LLM will allow companies to transform customer experiences by injecting AI into conversations across the entire customer journey. Built-in guardrails limit hallucinations or inappropriate responses and fine-tuning on a company’s knowledge base improves accuracy and transparency for domain-specific conversations. Enterprise controls grant abilities like restricting model outputs based on data sources, monitoring tone, and teaching non-AI responses.
“To harness the full power of foundation LLMs, enterprises need to fine-tune them for their domain requirements and add additional guardrails around security, compliance, and relevance, while also ensuring data residency and cost efficiency,” Gupshup CEO and co-founder Beerud Sheth said. “Gupshup is excited to launch the ACE LLM family of domain models, which are custom-built to fill this gap, thereby enabling enterprises to transform their customer experiences.”
ACE LLM is offered through Gupshup’s public cloud with options for data residency in specific geographies or private cloud deployment. Pricing models will support high scalability for enterprise needs. Use cases span enabling product discovery, generating leads, facilitating purchases, and providing conversational support, all delivered with speed, scalability, and security.
The launch builds on Gupshup’s experience with the Auto Bot Builder platform launched in January. Auto Bot Builder applies GPT-3 and fine-tuned proprietary data to generate a chatbot. ACE LLM evolves that platform with more options for foundational models and adjustment options. The varied verticals and tasks offered by the pre-trained models Gupshup provides fits with the company’s strategy in that regard. Last year, Gupshup made a series of acquisitions, buying startups with a focus on conversational AI for e-commerce and retail, finance, and cloud communication. The company raised $340 million in 2021 and has funneled that cash toward diversifying its stable of tech.