Scalability has played an instrumental role in enhancing breakthroughs in the artificial intelligence field – think ChatGPT.

Nevertheless, for valuable use cases to be realized, especially in regulated industries, such as healthcare and finance, explainability is needed to understand why a certain prediction or decision was reached, because solutions go beyond right or wrong answers.

Hewlett Packard Enterprise Co. has partnered with Aleph Alpha GmbH to make explainability in AI a reality, according to Jonas Andrulis (pictured, left), founder and chief executive officer of Aleph Alpha.

“We now have basically built something for explainability,” he said. “If you are trying to figure out how to orchestrate different perspectives … you need something else than just a chatbot that gives you a reply that’s either wrong or right. We’re focusing on these highly regulated industries, really complex workloads that rely on proprietary data. Both companies are very much focused on bringing engineering excellence in like a sovereign way to the customer.”

Andrulis and Eng Lim Goh (right), senior vice president and chief technology officer of AI at HPE, spoke with theCUBE industry analysts Lisa Martin and Dave Vellante at the recent HPE Discover, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed why explainability is a game-changer in AI and how HPE and Aleph Alpha are making it a reality, especially in regulated sectors. (* Disclosure below.)

Cloud service for large language models is different

Given that LLMs require heavy compute, their execution takes a different approach because they require a distinctive cloud service. As a result, HPE uses its supercomputing knowledge to deal with the single workload presented on many compute servers, according to Goh.

“It is very different when we build a cloud service for large language models,” he explained. “The traditional cloud service model is where you have many, many workloads running on many compute servers. But with a large language model, you have one workload running on many compute servers, and therefore the scalability part is very different.”

Since Aleph Alpha’s AI models have the ability to deal with highly regulated industries through explainability, HPE incorporates tools on top of them for optimality purposes. These models also have the capability of accepting images, Goh pointed out.

“What we do … is to build tools in front of what Aleph Alpha is using, especially in the regulated industry — for example, when a regulator comes in and asks for an audit,” he said. “Aleph Alpha has explainability, but sometimes they also want to ask, ‘What data do you feed that model with?’ Because that can influence the model.”

By not limiting visual or multimodal data to predefined classes, Aleph Alpha is able to boost the accuracy and validity of AI models. Building on language does not only enhance the learning of speech patterns and grammar, but also makes understanding human intelligence better, according to Andrulis.

“We found a way to combine the continuous space of images with the symbolic space that has reasoning capabilities, or at least the reasoning capabilities of a stochastic parrot,” he stated. “Now we’re able to combine these two worlds in like one embedding … our model is multilingual and multimodal, and it has a shared embedding.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of HPE Discover:

(* Disclosure: Hewlett Packard Enterprise Co. and Intel Corp. sponsored this segment of theCUBE. Neither HPE and Intel nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE