IBM has launched watsonx generative AI tools to speed up translation from COBOL to Java on its IBM Z mainframes.
The watsonx Code Assistant for Z, which will be generally available in the fourth quarter, is aimed at its mainframe customer base that's looking to modernize its COBOL applications. IBM said that it will preview watsonx Code Assistant at its TechXchange conference Sept. 11-13.
IBM is rapidly building out its watsonx Code Assistant portfolio, which will include IBM watsonx Code Assistant for Red Hat Ansible Lightspeed. Coding is one of the big productivity use cases for generative AI and large language models.
- IBM Delivers AI How It Is Meant to Be With Watsonx
- IBM launches Watsonx, an AI platform with open source models, governance
According to Big Blue, the company's watsonx.ai code model will have knowledge of 115 coding languages. The aim is to automate coding as much as possible. In addition, the COBOL use case can alleviate a shortage of developers. IBM research has found that enterprises are more likely to use existing mainframe assets than rebuild applications from scratch over the next two years. Lack of resources and mainframe skills is a big challenge.
IBM's watsonx Code Assistant for Z also aims to keep its IBM Z systems current. IBM's mainframe business is strong and booms with every new Z release. By helping customers incrementally migrate COBOL code to more modern platforms, IBM also keeps the mainframe relevant.
Watsonx Code Assistant for Z is expected to include IBM's Application Discovery and Delivery Intelligence (ADDI) inventory and analysis tool that will outline the steps and opportunities in transforming COBOL code to Java code. IBM is betting generative AI will be a better option than rewriting all application code in Java, migrating to public cloud and creating code that's hard to maintain.
Here's a demo.
Constellation Research's take
Constellation Research analyst Andy Thurai provided his take on the news.
"The code generator LLMs are dime a dozen in the market now. From AWS Code Whisperer to Microsoft Github Copilot to OpenAI Codex to ChatGPT to Google T5 there are major players in the market. There are also smaller players competing in this market: Tabnine, Replit Ghostwhisperer, Codestarter, CodeWp, MutableAI, CodeSquire, Figstack, Datamaker, AirOps, What the Diff, AI query, Adrenaline, Adept, Alphacode, Debuild, Codiga, Locofy, AIXcoder, Mintlify, Maya, Codis, Durable, Bloop, Enzyme, Dhiwise, Codota, Anima, Codacy, Warp, Metabob.
The core of this initiative is to make programmers more efficient by using English as a programming language (possibly other human languages in the future). Enterprises assume the productivity of developers will improve. However, the code spewed out by these solutions can be security risks and infringe on IP and proprietary software.
The key part of any software company, outside of its core business value proposition, is its developers. Obviously, every company employs tons of software professionals and always struggles to upskill them, make them efficient, use DevOps processes, and agile methodologies, build things faster, and be more efficient. These generative AI tools can help.
IBM watsonx Code Assistant is somewhat unique. Watsonx Code Assistant is predominantly trained with code for the purpose of code generation unlike the other LLMs, which are general purpose. This LLM is task-specific and can be classified as a Specialized Language Model (SLM). One of the major issues with mainframe modernization is finding skilled developers. Finding skilled COBOL programmers is extremely difficult either because of workforce mass resignation (the great resignation) or retirement. Finding developers who can understand COBOL and program in Java is even more difficult. Watsonx Code Assistant could be useful to fill these gaps. These models are trained to understand mainframe COBOL applications and spit out object-oriented Java code that can be used to make mainframe modernization easier.
Because IBM trained this model with its own mainframe code, there will be no legal or ownership issues that can arise. However, the announcement and the release are still in preview mode. Let's hope IBM can execute."