Another GenAI-Powered Testing Tool Comes to IBM i
November 6, 2024 Alex Woodie
IBM says it’s bringing a new AI accelerator to the Power line to drive compute-hungry generative AI workloads, such as the new RPG coding assistant that it’s also developing. But IBM won’t be the only source of such GenAI-powered coding assistants, as a company called Abstracta recently announced another one.
There’s a lot of interest in the Spyre accelerators that IBM announced earlier this year for the System z mainframe and which IBM announced last month during its TechExchange conference will be coming to Power. Considering the massive demand for Nvidia GPUs, which could plug into past generations of Power chips but aren’t compatible with Power10, it’s clear that IBM wants and needs to bring some extra oomph to emerging GenAI workloads on midrange and mainframe boxes.
IBM’s biggest enterprise customers undoubtedly want to use their massive stockpiles of proprietary data to train and run their own LLMs to power customer service chatbots and other GenAI applications. Nvidia will gladly sell them DGX systems to run these next-gen workloads, which are nearly as compute-hungry in inference mode as they are to train in the first place. But with the upcoming Spyre accelerators and the recently announced Granite 3.0 models in Watsonx, IBM may have found a way to keep those big accounts from wandering over to a competitor.
In addition to announcing Spyre on Power, IBM also announced in a statement of direction that it’s developing a GenAI co-pilot for RPG. We already knew that was in the works by the IBM i development team in Rochester, Minnesota, since IBM i CTO Steve Will first unveiled the news at the COMMON POWERUp 2024 conference in Fort Worth, Texas in May.
Back in May, Will told us the IBM i coding assistant, which doesn’t appear yet to have a name, would focus on three use cases, including helping RPG developers understand, write, and maintain code; helping operators stay on top of monitors and alerts; and helping Db2 for i database engineers with trend analysis and anomaly detection. It’s not clear if IBM will offer other coding assistants for the operational and database workloads, or if they’ll be included with the code assistant for RPG.
Every GenAI product needs an LLM under the covers, including co-pilots like the RPG assistant that IBM is developing, which Will said would use the language models that IBM ships in its Watsonx product. But there are many other LLMs out there, including OpenAI’s GPT-4, which was trained on the Microsoft Azure cloud and is still one of the world’s biggest and most powerful LLMs.
Last month, a San Francisco-based company by the name of Abstracta announced the launch of its own coding co-pilot that it developed in concert with Microsoft Co-Innovation Labs. Dubbed the Abstracta Copilot, it’s designed to automate many of the routine tasks involved with software testing, including on the IBM i server.
According to Abstracta, the new offering uses natural language processing (NLP) technology to allows testers, business analysts, and product owners to access databases without SQL expertise, to “interface with APIs without Postman, and work with mainframes without AS400 commands” (sic).
The software, which runs in a Web browser, also “enables users to quickly generate user stories, test cases, and technical documentation, reducing the time needed for these tasks from days to mere seconds,” the company says.
“We created the Abstracta Copilot with one clear mission: to streamline processes, increase quality, dramatically enhance productivity, and reduce software development costs,” Sofia Palamarchuk, the co-CEO of Abstracta, said in a press release. “We are excited to share this powerful tool with our clients as part of our services and maximize the quality of the software we co-create.”
Abstracta boasts 15 years of experience building software to help automate software testing. The company, which has locations in California, Uruguay, and the UK, has customers across e-commerce, financial services, healthcare, and the tech industries, and runs one of the largest software testing conferences in South America. It has worked with the IBM i server for years, according to its website.
It’s worth noting the software testing use cases that Abstracta is chasing are some of the same uses for GenAI that IBM is targeting with its RPG co-pilot. It’s also worth noting that neither of these are the first commercially available GenAI copilots available for IBM i. That distinction would go to Remain Software, which launched its ChatGPT-based AI Chat offering back in April 2023.
Coding copilots are growing in popularity. While the early successes, such as Microsoft’s Github Copilot, are designed to work with the most popular languages, like Java and SQL, it’s just a matter of time before they’re trained to read and write more esoteric languages, like RPG.
These are early days, and it’s unclear whether these copilots will need onboard AI accelerators located in the Power box, or whether a remotely located LLM running on a GPU in a cloud data center will suffice.
Certainly, there will be a time and a place for massive computational horsepower. Code modernization projects that involve tens of millions of lines of spaghetti code may be just the sort of job that requires having ready access to large amounts of computational power.
And then there will other code maintenance or testing tasks that can be done in the cloud on commonly available LLMs, such as Remain Software’s ChatGPT-based offering or Abstracta Copilot. The world of GenAI-infused business software is still in its earliest days, and there is a lot to hash out.
RELATED STORIES
RPG Code Assist Is The Killer App For AI-Enhanced Power Systems
IBM Developing AI Coding Assistant for IBM i
Remain Launches AI Chatbot to Assist with Development on RDi