How model fitting brings productive AI to business

There is a lot of hype and activity around generative AI models, especially in the short time since ChatGPT was first released.

ChatGPT — and the GPT-3 Large Language Model (LLM) it’s based on — are trained on public data, which serves as an excellent foundation for consumer applications, but lacks the customization, privacy, and security that an enterprise requires. That’s where Rodrigo Liang, co-founder and CEO of SambaNova Systems, wants to make a difference with today’s launch of his company’s SambaNova Suite, which aims to help businesses build and deploy custom AI models.

>>Follow VentureBeat’s ongoing genetic AI coverage<

Launched in 2017, SambaNova has focused heavily on AI hardware, raising a staggering $676 million in April 2021 to support its efforts. In recent years, the company has expanded beyond its original hardware focus to build support for both machine learning training and inference on different models, with its streaming data-as-a-service offering. The new SambaNova Suite extends the streaming service with a collection of capabilities that allow organizations to customize both open source and proprietary AI models to meet their specific requirements.

“The focus of SambaNova is how to bring more AI production capabilities to the enterprise,” Liang told VentureBeat. “There are certain things you need for businesses to be able to be successful, and we’re doing that today for them.”

Nvidia isn’t the only supplier of AI hardware for genetic artificial intelligence

There is a growing list of vendors that are built on top of AI models created to help enable enterprise use cases.

Content creation is a particularly vibrant field of AI customization created for businesses. Jasper AI, for example, recently announced its Jasper for Business offering designed to help customize AI for a specific business. Typeface came out of stealth on Monday with its enterprise content creation platform for productive artificial intelligence, along with $65 million in funding. Quantive last week announced its foray into genetic AI, helping organizations with their business strategy.

SambaNova’s key differentiator from others in the AI ​​production space is that it has its own hardware that helps optimize enterprise use cases. Rather than simply relying on Nvidia GPUs like many in the industry, SambaNova has developed its own custom silicon that is optimized for both machine learning and inference.

“What we’ve done is … take the AI ​​software stack that people really want to use with PyTorch, TensorFlow and complex models like GPT, and we’re bringing them down to the silicon,” Liang said. “We have silicon that is custom-built to run these large models for the enterprise, and vice versa.”

The team behind SambaNova, including Liang, has a background in making microprocessors for Sun Microsystems and Oracle. Liang said SambaNova has created an AI microprocessor that is highly efficient and effective in order to run these enterprise AI production applications.

Liang emphasized that custom silicon also enables a continuous training and inference capability, so that the data that feeds generative AI models can be kept up-to-date.

“In business, you need real-time information, and so you don’t want your models to fall behind,” he said.

Privacy, responsible artificial intelligence and the business

With the new SambaNova suite, Liang said his company seeks to solve the primary challenges businesses have with genetic AI. Among these challenges is tailoring for a company’s specific data, as well as the ability to limit bias and deliver accountable and explainable AI.

With its platform, SambaNova enables an organization to run personalized training in a private environment on any data the organization has, including unstructured data that might be found in a company’s Slack chat channels. Taking it a step further, Liang said the platform provides transparency to organizations on how a particular AI model actually works.

“SambaNova is built to be able to give you exposure to exactly how the model came to a certain conclusion,” Liang said. “We store all the processes around how to train and optimize the model so that when a tester comes in or someone wants to check for bias or why something happened a certain way, you can actually work through the flow and verify that the results they were done right for you.”