Over the past twelve months, tools such as ChatGPT and similar solutions for image, video and text creation have already become mainstream. This year, they will continue to gain a foothold in the day-to-day work of developers and administrators. The possible applications are many and varied. But what many people forget: Machine learning algorithms and AI tools, which are themselves applications, go through a life cycle that is very similar to that of traditional software. AI applications also have to be developed, tested, deployed, checked and monitored - this is known as 'MLOps'. One challenge that the IT industry will therefore have to deal with in the coming year is the marriage of the software lifecycle with the ML lifecycle. This is not so easy, as machine learning applications are data science projects that are primarily written in Python and other mathematical programming languages - an administrative challenge for developers. Platforms such as 'Red Hat OpenShift Data Science' can help to solve the 'MLOps' problem with suitable Jupyter notebooks, Kubeflow deployment workflows and similar tools.

Five examples of generative AI in code development

But how will generative AI and large language models (LLMs) make developers' lives easier and affect the productivity of programmers and administrators? Five examples.

Source code on command: natural language instead of complex code. Generative AI has the ability to understand even complex issues and provide appropriate answers to questions. This also works when it comes to programming: more and more development environments will be equipped with coding assistants that are able to process natural language (natural language processing). Initially, these digital helpers will not yet be able to write highly complex programs independently without a great deal of additional input. However, AI tools are already able to define a Java class, for example. AI could also make the operation of IT automation tools such as Ansible even simpler and more intuitive, thereby increasing developer productivity.

More security through AI. The democratization of programming skills also has its downsides: Even laypeople can misuse modern chatbots and have them write malware. However, developers will also use AI tools more and more in the coming year to secure their applications, for example to find vulnerabilities and security risks in the source code. Administrators will also train algorithms and LLMs that are suitable for pentesting the systems they manage and the application landscapes they operate.

Code analysis made easy. If an application issues an error after the code has been created, the sometimes nerve-wracking error analysis begins. AI will also make this work much easier in the future and therefore increase productivity. Properly trained, LLMs can not only detect typos and missing brackets in the code, but also recognize logic errors and insufficient instructions.

Modernization of applications. The better trained an AI is and the more comparative data is available to the algorithm, the more demanding tasks it can take on. In the coming year, tools may already be able to evaluate the domain model of a monolithic application. As soon as this is the case, AI will also be able to break down the monolith into microservices. Creating APIs so that these can communicate with each other and packing them into containers will then be a rather easy exercise for developers: this will be of particular benefit to companies that want to transfer their legacy applications to a modern cloud or Kubernetes infrastructure.

Markus Eisele is Developer Strategist at Red Hat

© Red Hat

Roll backwards: AI explains the code to humans. If the AI is able to understand code, it can explain it in an understandable way using a corresponding language model. In the coming year, developers will be able to use tools with this capability to automatically create documentation for applications, for example, or even have an artificial intelligence insert comments at critical points in the source code.

2024 will therefore be the year of artificial intelligence. On the one hand, developers will have access to a variety of new tools that support them with AI capabilities and relieve them of a lot of work on a daily basis; their productivity will increase measurably as a result. At the same time, administrators will be faced with the task of operating AI tools and managing their lifecycle - no easy task, but suitable platforms are already waiting in the wings.