Motivation and Scope
“Software for all and by all” is the future of humanity. AIware, i.e., AI-powered software, has the potential to democratize software creation. The definition of software along with many Software Engineering (SE) aspects, processes, tools, platforms, and techniques will need to be either reimagined, reformulated or redesigned, enabling individuals of all backgrounds to participate in its creation with higher reliability and quality. Over the past decade, software has evolved from human-driven Codeware to the first generation of AIware, known as Neuralware, developed by AI experts. Foundation Models (FMs, including Large Language Models or LLMs), ushered in software’s next generation, Promptware, led by domain and prompt experts. However, this Promptware merely scratches the surface of software’s future. We are already witnessing the emergence of the next generation of software, Agentware, in which humans and intelligent agents jointly lead the creation of software. With the advent of brain-like World Models and brain-computer interfaces, we anticipate the arrival of Mindware, representing the 5th generation of software. Agentware and Mindware promise greater autonomy and widespread accessibility, with non-expert individuals, known as Software Makers, offering oversight to autonomous agents.
The SE community will need to develop fundamentally new approaches and evolve existing ones, so they are suitable for a world in which software creation is within the reach of Software Makers of all levels of SE expertise, as opposed to solely expert developers. We must recognize a shift in where expertise lies in software creation and start making the needed changes in the type of research that is being conducted, the ways that SE is being taught, and the support that is offered to software makers.
A foundation model (FM) is a machine learning model that is trained on broad data such that it can be adapted to a wide range of downstream tasks. We are already witnessing the current popular form of FM (i.e., Large Language Models (LLMs)) evolving to encompass other forms of data beyond text (e.g., images, audio, and video data). Today’s FM-powered software (FMware) has many limitations that hinder our abilities to develop safe, trustworthy, and high-quality software. For example, hallucination, which is the problem of FMs sometimes generating incorrect or purely fictional texts, has already led to issues with the generation of incorrect information in several court cases. Instead of waiting for better FM technologies, innovative engineering solutions are being developed to mitigate or resolve these issues. For instance, recognizing the tendency of LLMs to hallucinate and their limited mathematical capabilities, engineering solutions have been proposed that enable LLMs to invoke legacy codeware for mathematical computations and to ground themselves using legacy data sources.
This special issue on AIware aims to explore the state of Software and SE in the FM Era. The issue seeks articles that examine the developer productivity, system quality/trustworthiness, SE education, SE platforms, open/inner source collaborations and the whole SE lifecycle (developing, delivering, debugging, evolving, and monitoring complex AIware) in the FM Era.
We invite papers covering any software engineering aspects of AIware in the FM era (a.k.a, AgentWare and PromptWare), which includes but not limited to some of the topics below. Papers must have a practical perspective and emphasize their value in supporting software practitioners and/or documenting current industrial best practices:
For more information about the focus, contact the guest editors:
For author information and guidelines on submission criteria, visit the Author’s Information page. Please submit papers through the ScholarOne system, and be sure to select the special issue or special section name. Manuscripts should not be published or currently submitted for publication elsewhere. Please submit only full papers intended for review, not abstracts, to the ScholarOne portal. If requested, abstracts should be sent by email to the guest editors directly.
