Artificial intelligence is changing the world, but for effective AI implementation, businesses must transition from proof-of-concept to full-scale production.
Organizations are eagerly exploring this potential, but many need help with the path to realizing meaningful returns on investment. The complexity of transitioning AI initiatives from POC to full-scale production is a significant factor contributing to these struggles, according to Martin Willcox (pictured), senior vice president of worldwide industry and analytics at Teradata Corp.
Martin Willcox, senior VP of worldwide industry and analytics at Teradata Corp., talks with theCUBE about AI implementation.
“There are a lot of organizations that are struggling to know where to start and a lot of organizations struggling to know really how to deliver return on investment from their investments in AI and machine learning,” Willcox said. “So, really, the thrust of [this morning’s ‘Load the Dice’] presentation was [that] right now, failure rates are pretty high. We can debate exactly how high, depending on which study you trust, but they’re pretty high.”
Willcox spoke with theCUBE Research’s Rob Strechay and co-host Savannah Peterson at Teradata Possible during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the challenges and strategies for effectively implementing AI and machine learning in businesses. (* Disclosure below.)
Overcoming challenges in AI implementation for real business impact
Many businesses invest heavily in innovation labs and advanced prototypes only to see these efforts stall before delivering tangible outcomes. The key to overcoming this challenge lies in starting with a clear end goal and integrating AI outputs into operational workflows, Willcox explained.
“There’s no return on investment if we don’t actually get to production [and] if we don’t change the way we do business,” Wilcox said. “There’s a lot to unpack there. There’s a variety of reasons why that happens …We don’t actually change anything in the business by building a model and updating a table of predictions; we’ve got to actually embed those predictions in an operational business process and change the way we do something.”
Enterprises must strategically focus on data infrastructure and production readiness to succeed in their AI endeavors. It’s not just about building sophisticated models but about making those models work in real-world scenarios, Willcox concluded.
“We kind of miss the fact that actually what matters is what works,” Willcox said. “It’s been a truism in AI machine learning for as long as I’ve been involved in the field — the best model is the simplest model that’s sufficiently accurate for your use case. There is a time and a place when only one and a half trillion parameter large language model will do.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of Teradata Possible:
(* Disclosure: TheCUBE is a paid media partner for Teradata Possible. Neither Teradata Corp., the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)