RelationalAI Previews AI Coprocessor, Reshapes Knowledge Graph
RelationalAI announced the preview availability of what it calls its AI coprocessor for data clouds and language models today at Snowflake Summit in Las Vegas. Its flagship offering enables organizations to utilize knowledge graph technologies directly inside of Snowflake.
The announcement signifies the ongoing relevance of relational technologies in a data landscape dominated by Artificial Intelligence. Including RelationalAI’s graph approach inside Snowflake enables users to employ statistical machine learning techniques alongside non-statistical techniques reflective of AI’s knowledge foundation, like symbolic reasoning.
Embedding these capabilities within Snowflake effectively democratizes them to a widening audience reliant on the cloud for its data-driven needs. According to RelationalAI CEO Molham Aref, enterprise use of AI has reached “an inflection point that demands a combination of AI techniques to generate results in the data cloud that cannot be achieved in silos.”
RelationalAI’s knowledge graph approach not only expands the type of AI available within Snowflake, which is becoming increasingly ubiquitous throughout the data ecosystem, but it does so in one of the most widely used cloud repositories for storing and analyzing data.
One of the most immediate benefits of using RelationalAI for assembling knowledge graphs and performing what Aref termed “graph analytics” pertains to how it’s architected with Snowflake. The solution plugs into the cloud data warehouse directly, which presents a number of significant gains for Snowflake users. Firstly, users are no longer responsible for moving their data to external resources to access knowledge graph capabilities, which are renowned for their relationship discernment and standardization of even disparate data types (depending on how they’re implemented).
Instead, organizations can access this functionality natively within Snowflake, which enables them to continue using their established models for “data governance, security, and access controls,” Aref commented. Since RelationalAI provides a knowledge graph atop Snowflake’s underlying data, users can still rely on traditional relational techniques to manage their data while benefiting from graph-aware analytics and graph AI techniques.
The Relational Knowledge Graph
The relational underpinnings of Snowflake for the graph environment RelationalAI provides result in what Aref called a “relational knowledge graph”. Although conceding such a term may be considered an “oxymoron”, Aref denoted the descriptor applies to RelationalAI’s graph functionality within the relational environment. Applying graph techniques to analytics and AI has always been beneficial for everything from the fundamentals of building models and generating features to the types of AI graphs support.
Certain techniques like clustering or Principal Component Analysis excel in graph environs, while others — such as Graph Neural Networks — all but require graph frameworks. Many of the non-statistical varieties of AI, which Aref mentioned involve aspects of “reasoning, knowledge representation, and solvers” are native to graph environments. Simultaneously, these surroundings can also support what’s arguably the acme of machine learning: the prescriptive analytics capabilities Aref mentioned that RelationalAI also enables.
Another of the chief advantages RelationalAI provides Snowflake users is its ability to deliver a semantic layer of understanding, in business terminology, for data of different formats, models, and applications. This benefit is particularly useful because the distributed nature of the data landscape makes it necessary for organizations to frequently “implement intelligent applications with semantic layers on a data-centric foundation,” Aref revealed.
Although the underlying graph mechanisms in RelationalAI don’t involve W3C standards, they still identify, organize, and clarify the semantics underscoring data’s value to specific business use cases. The knowledge graph solution’s semantic layer minimizes the manual efforts required to unlock data’s meaning to important business objectives. Placing this layer atop the various data types likely found in a customer’s particular Snowflake instance makes the data much more comprehensible to aid everything from data discovery to analytics while reducing the time to value for these processes.
Graph and Relational
The overall significance of RelationalAI’s inclusion in Snowflake ultimately transcends either vendor or its user base. From a broader perspective, this integration heralds the fact that relational and graph approaches don’t have to be adversarial or viewed in terms of one supplanting the other. Instead, they can productively coexist to merge the benefits that both bring into a single solution.
As such, the varieties of AI graph environments support — coupled with their semantic benefits — become more accessible to organizations because of the relational environs in which they’re accessed. The governance and security boons of not having to move data to exploit graph techniques with this pairing are also significant and can potentially provide compounding value across use cases.