IBM and Red Hat - Code that Writes Code

Jon Corchis, Manager of Architecture
​Unilogik Systems Inc.

---

IT (Information Technology) application monitoring is no longer a choice for successful enterprises, nor is automation, so it makes perfect sense to lower the IT automation barriers (time, training, staffing, cost, etc.) to entry. Project Wisdom was announced at Ansible Festival in 2022; it is another project within the IBM AI (Artificial Intelligence) for Code research group in collaboration with Red Hat. This choice makes sense since Ansible is the de facto tool for IT automation. The goal of Project Wisdom is to make large-scale infrastructure automation using plain English a reality.  
 
Most AI models depend on real world data to be successful. In true Red Hat open-source practice there is currently a Beta program in progress, and IBM states that the AI model and underlying engine will be open as well..  This project uses Ansible playbooks from the participating Beta community to further refine their AI model. Project Wisdom’s interface is Visual Studio Code IDE with an Ansible YAML extension. Users describe what they want to accomplish using plain English in the name tag, and after a moment, Project Wisdom makes a proposal. At that time, an experienced Ansible developer can change the code, or someone less experienced with Ansible can take the playbook and run it against a pre-production server to compare against the desired state of the host. In both cases there is value added. The experienced playbook developer is not bogged down with mundane programming; they can use their experience to review the playbook directly, which frees them up for more value-added tasks. While someone new to Ansible can learn from the community through the AI. ​

​AI today is not perfect and never will be, just like humans. Yes, there are deterministic exceptions, but Coding is an art. Compare the differences between a streaming service recommending a movie versus one of your friends – both are hit or miss. The key point about code that writes code is that it still requires human intervention. Like all code that is released to production, there is a process that happens before code or configuration ever touches a production server or device. In general, code that writes code puts more demand on scanning tools and human reviews to verify code as part of the release process.  
 
Unilogik will be watching the progress of this project closely in order to help our customers get to a mature automation process quickly. Project Wisdom roadmap features, like redundancy identification and content explanation, will also help accelerate this progress. How is your company using AI or planning to use AI to do more with less? Add your comments below.