Tuesday, July 2, 2024

Unique: Snowflake and Reka add multimodal LLMs to information cloud

Be a part of Gen AI enterprise leaders in Boston on March 27 for an unique evening of networking, insights, and conversations surrounding information integrity. Request an invitation right here.


Snowflake Information Cloud is getting multimodal giant language fashions (LLMs). At present, the Sridhar Ramaswamy-led firm introduced it’s partnering with Reka, an AI startup based by researchers from DeepMind, Google and Meta, to convey its proprietary fashions into its information platform. 

The transfer comes mere weeks after the firm’s partnership with Mistral and can allow enterprises utilizing its information cloud to construct generative AI purposes that may work with not solely textual content inputs but in addition pictures and movies. This may open a complete new method for groups to drive worth from datasets. 

Snowflake participated in Reka’s $60 million spherical of funding final 12 months. Nonetheless, it didn’t affirm whether it is rising its funding with this partnership.

Baris Gultekin, the pinnacle of product administration at Snowflake AI, instructed VentureBeat the corporate is “at all times exploring methods” to assist companions and drive innovation for purchasers however has nothing extra to share on the funding entrance at this stage.

VB Occasion

The AI Affect Tour – Atlanta

Persevering with our tour, we’re headed to Atlanta for the AI Affect Tour cease on April tenth. This unique, invite-only occasion, in partnership with Microsoft, will function discussions on how generative AI is reworking the safety workforce. House is restricted, so request an invitation right this moment.


Request an invitation

Reka Flash and Core coming to Snowflake Cortex

Since its inception, Snowflake has targeted on constructing the info infrastructure of selection for its clients.

Nonetheless, the imaginative and prescient has not been simple to realize. The corporate initially supplied customers with a easy information warehouse however because the ecosystem developed it added assist for extra information codecs and capabilities to construct a full-fledged information cloud — enabling a variety of downstream use circumstances, together with AI and analytics.

Final 12 months, because the generative AI wave took the world by storm, it took one other step to proceed this evolution with the launch of Snowflake Cortex, a totally managed service tailor-made for constructing LLM apps.

Cortex offers enterprises utilizing information cloud with a set of AI constructing blocks, together with open-source LLMs, to investigate information they’ve on the platform – with the identical safety, privateness and governance of Snowflake – and construct purposes focusing on completely different business-specific use circumstances. The corporate began with LLMs for specialised duties resembling sentiment evaluation and is now going multimodal with the inclusion of two LLMs skilled by Reka: Flash and Core.

Reka Flash is a state-of-the-art 21-billion parameter mannequin that serves as an optimized “turbo class” providing and matches the efficiency of many bigger fashions throughout numerous language and imaginative and prescient benchmarks. The Core mannequin, alternatively, is alleged to be Reka’s largest and most performant mannequin that approaches GPT-4 and Gemini Extremely by way of efficiency. Nonetheless, not like the Flash mannequin, the Core model isn’t out there publicly at this stage. 

Snowflake stated it’s including the Flash mannequin to Cortex instantly and growing assist for the larger one to verify it turns into out there proper after launch.

Gultekin didn’t share a timeline however stated it will drop quickly. He additionally famous that the corporate may additionally add different fashions from Reka, together with the smaller Edge mannequin, primarily based on the demand. 

How will multimodal AI profit Snowflake customers?

With Cortex and Reka’s AI fashions, customers will be capable to construct generative AI purposes that may work with not solely textual content inputs but in addition pictures and video recordsdata.

This could turn out to be useful throughout numerous use circumstances, together with options for captioning movies, labeling pictures, producing e-commerce product descriptions, and answering analytical questions primarily based on charts.

“Some instance purposes our clients are enthusiastic about are chatbots that may perceive charts, or leisure companies producing advertising and marketing or promoting content material given their video or picture belongings,” Gultekin instructed VentureBeat. 

Whereas the AI product head didn’t share what number of firms need to use or utilizing Reka fashions, he did share that over 400 enterprises are leveraging Cortex and the fashions it hosts to construct gen AI apps within the information cloud.

This consists of purposes focusing on domains, proper from analyzing safety vulnerabilities from service tickets to chatbots that assist healthcare suppliers higher interface with insurer information. 

The addition of Reka’s fashions will take the overall variety of LLMs on the service to a dozen, he stated. This consists of fashions from Mistral and Google, which had been launched only a few weeks in the past.

“Our AI innovation pipeline is in overdrive. At its core, Snowflake’s imaginative and prescient is to get AI into the arms of each consumer, no matter their technical experience, to ship recognizable enterprise worth quicker. By furthering our partnership with Reka, and thru integrations with Gemma and Mistral, and with what we now have coming within the subsequent weeks and months, we’re doing simply that,” Gultekin stated whereas hinting the corporate will announce extra AI improvements within the run-up to its annual summit in June.

Nonetheless, he didn’t share specifics of what’s within the pipeline.

“Our roadmap stems from the foundational precept that there isn’t any AI technique with no information technique – Information is the gasoline for AI. We’re leaning into that and enhancing our capability to spice up productiveness, improve collaboration, and finally speed up end-to-end AI and ML workflows for our clients, all with Snowflake’s safe and trusted information basis,” Gultekin added. 

Notably, Databricks, which competes with Snowflake within the information ecosystem, has additionally had the identical play. The corporate has been introducing open fashions since buying MosaicML final 12 months whereas additionally bolstering its tooling to construct gen AI apps on the identical time. In truth, only a few days again, it acquired Lilac, an AI startup that helps groups analyze and repair unstructured information for AI coaching.

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles