AI with Gaudi Powered Conversational System
How the new Inflection for Enterprise platform, built on Intel Tiber AI Cloud and Gaudi 3, aims to solve deployment and scaling challenges for corporate customers.
Intel and Inflection AI, a company known for its conversational, “empathetic” AI model Pi, have announced a strategic collaboration to deliver a comprehensive, enterprise grade AI system. This joint effort, dubbed Inflection for Enterprise, is a significant move that positions Intel’s Gaudi AI accelerators against the market dominance of Nvidia in large scale corporate AI deployments.
The core of the offering is an all encompassing AI system powered by Intel Gaudi technology and the Intel Tiber AI Cloud. According to the companies, this system is designed to provide businesses with a highly personalized, empathetic, and scalable AI coworker, specifically trained on the unique data, policies, and culture of each enterprise. The initial availability is through the AI Cloud, with a plan to deliver the platform as an industry first Gaudi 3 powered AI appliance in the first quarter of 2025.
A Response to Enterprise Frustration
The partnership addresses a pervasive point of friction in the corporate adoption of generative AI, namely the gap between generic, off the shelf AI tools and the specialized requirements of a large organization. Ted Shelton, Inflection AI’s COO, noted that many CEOs and CTOs express frustration that existing market tools are not truly enterprise grade. The complexity of model fine tuning, which is necessary to integrate a large language model into a company’s operational fabric, is often beyond the technical capabilities of most enterprises.
Inflection for Enterprise seeks to bridge this divide by offering a full stack solution that includes the foundational model, the hardware, and a streamlined deployment environment. By leveraging the performance of the Intel Gaudi 3 AI accelerator, the companies promise an industry leading price to performance ratio, which is a critical factor for organizations facing immense capital expenditure for AI infrastructure. The use of Intel’s AI Cloud further streamlines the process by offering a unified environment for building, testing, and deploying AI applications, promising a faster time to market for businesses.
The Technical and Strategic Rationale
The strategic decision to power Inflection 3.0 with Gaudi 3 accelerators marks a notable shift for Inflection AI, whose consumer application, Pi, previously ran on Nvidia GPUs. This move underscores the competitive push by Intel to capture a larger share of the burgeoning AI hardware market. Justin Hotard, executive vice president and general manager of Intel’s AI and Data Center Group, emphasized that the collaboration is setting a new standard for AI solutions, offering immediate and high impact results. He highlighted that the competitive performance per watt and support for open source models and tools make the Gaudi 3 solutions more accessible and efficient for businesses of any size.
The technical architecture of Inflection for Enterprise is centered on deep customization and performance optimization. Inflection AI fine tunes its foundational model, Inflection 3.0, to be native to each organization. This fine tuning process incorporates a company’s history, policies, content, product information, and operational data, aligning the AI’s tone and purpose with the corporate identity. This is achieved through proprietary fine tuning and Reinforcement Learning from Human Feedback (RLHF), the same methodology that refined the original Pi model. This level of personalization is intended to drive higher user adoption and utility across various use cases by ensuring the virtual AI coworker speaks the language of the business.
Beyond customization, the solution addresses crucial enterprise concerns around security and ownership. The models fine tuned for a specific client are guaranteed to be the exclusive property of that client, never shared outside their organization. Furthermore, companies retain flexibility in deployment architecture, allowing the model to be hosted on premises, in the cloud, or in a hybrid environment.
Driving Down the Cost of Generative AI
A key selling point is the optimization of the total cost of ownership. The combination of Inflection for Enterprise and Gaudi 3 is touted to deliver up to a 2x improvement in the price performance ratio compared to current competing offerings. The upcoming Gaudi 3 powered appliance will offer 128GB of high bandwidth memory capacity, a significant feature for optimizing demanding GenAI workloads.
For enterprises looking to adopt generative AI without the upfront capital expense and complexity of model building and hardware testing, the AI Cloud provides application templates. These templates allow businesses to bypass the hardware provisioning phase, enabling rapid scaling and deployment. Intel’s confidence in the offering is further demonstrated by its plan to be one of the first customers, implementing Inflection for Enterprise internally.
Looking ahead, the collaboration aims to foster a broader developer ecosystem. Inflection AI and Intel plan to empower developers to build critical enterprise software tools and applications on top of the robust, human centric Inflection 3.0 system. Interested corporate clients can find more information about the offering at the Inflection AI website.
