Barring the growing pains of hallucination and bias, generative artificial intelligence excels at repetitive, information-based tasks.

With products such as GitHub Copilot and Amazon CodeWhisperer generally available, its capabilities are changing how developers code. How is IBM Corp. expanding the AI for coding idea from simple assistants to covering numerous facets of the software development lifecycle?

IDC’s Michele Rosen (left) and IBM’s Keri Olson (middle) discuss AI for coding with theCUBE.

“We’ve started out with coding assistants and they’re generating code, but we’re actually finding that developers are really interested in using gen AI technology for other aspects of the software development life cycle, such as testing or threat analysis,” said Michele Rosen (pictured, left), analyst and research manager at International Data Corp. “IDC has identified about 25 different gen AI use cases just within the software development life cycle — it’s a very quick-moving sector.”

Rosen and Keri Olson (right), vice president of data and AI product management at IBM, spoke with theCUBE Research’s Dave Vellante  at IBM Think, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed gen AI’s impact on software development is poised to grow, driving greater efficiency, innovation and productivity. (* Disclosure below.)

IBM’s multifaceted approach to AI for coding

At the heart of IBM’s strategy is the development and release of sophisticated Granite foundation models. These models, trained on 116 programming languages, range from 3 billion to 34 billion parameters, according to Olson.

“These models for code are performing well above their weight in terms of performance and accuracy going to our users,” she said. “That’s the first thing that we’re focused on. We have recently released these code models to the open-source community, and that is very important for us because we want to make coding as easy as possible for as many programmers as possible.”

IBM’s watsonx Code Assistant exemplifies the practical application of these models. Tailored for specific domains, such as IT automation and application modernization, it is fine-tuned to meet unique user needs. This approach enhances developer productivity and addresses skills gaps, particularly in legacy languages like COBOL, Olson added.

“IBM is really focusing on end-to-end solutions that provide organizations the ability to use AI for code for specific use cases,” she said. “One of those areas is application modernization.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of IBM Think: 

(* Disclosure: IBM Corp. sponsored this segment of theCUBE. Neither IBM nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE