Category: IEEE Intelligent Systems

Much has been written and discussed in previous years about human–AI interaction. However, the debate so far has mainly concentrated on “average” decision makers, neglecting important differences when it is experts who require support. In this article, we are going to talk about expert–machine collaboration for decision-making. We investigate the current approaches for expert decision support and exemplify the inefficiency of this approach for a real clinical decision-making problem. We propose two solutions for expert–machine collaboration to overcome the shortcomings of the current state of the art. We think that the proposed approaches open new horizons for expert–machine collaborative decision-making.

The article views the forthcoming virtual societies, decentralized society (DeSoc) or metaverse, as cyber-physical social systems, and discusses the key issue of management for such human-centered hybrid systems—the prescription of human behaviors. We build parallel humans with its aggregation—parallel population—to complete that task, where each individual's mental knowledge is cognitively modeled by heterogeneous learning, analyzed by generative big data with deep evolutionary reasoning, and prescribed by knowledge convergence with an active recommendation. A case study from social security has validated that our parallel population and parallel human are a feasible and effective way to construct DeSoc or metaverse.

Automated machine learning (AutoML), in particular, neural architecture search (NAS) for deep learning, has ignited the fast-paced development of automating data science (AutoDS) and artificial intelligence. However, in the existing literature and practice, AutoML, AutoDS, and autonomous AI (AutoAI) are highly interchangeable and primarily centered on the automation engineering of data-driven analytics and learning pipelines. This challenges the realization of the full spectrum of AI paradigms and human-like to human-level intelligent and autonomous systems. Going beyond the state-of-the-art paradigm of AutoML and their automation engineering, there is an expectation that the new age of AI and autonomous AI (or AutoAI+) will incorporate mind-to-action intelligence and integrate them with autonomy. We pave the way for this new AI and AutoAI integrating mindful AI and AutoAI with AI mind and mindfulness and actionable AI and AutoAI with AI actions and actionability and translating AI mind to AI action for autonomous, all-around AI systems.

Advertisement.

In the AI evolution, a significant and lasting vision and mission has been on designing autonomous AI systems (AutoAI). AutoAI differs significantly from another set of movements on automated machine learning (AutoML) and automated data science (AutoDS), which are often deemed interchangeable with automated AI. AutoML and AutoDS aim to automate some of the analytical and learning tasks, processes, and pipelines. This issue highlights the theme on AutoAI: Autonomous AI with six feature articles. My editorial further clarifies various misconceptions, myths, and pitfalls about the three related and often confused areas: AutoAI, AutoML, and AutoDS. This issue also includes an article on parallel population and human in the column AI Expert, expert–machine collaboration in the column AI Focus, and another article on intelligent mobile spaces and metaverses for the AI and Cyber-Physical-Social Systems (AI-CPSS) department.

Presents the table of contents for this issue of the publication.

Presents a listing of the editorial board, board of governors, current staff, committee members, and/or society editors for this issue of the publication.

Presents the front cover for this issue of the publication.

Advertisement.

Presents a listing of the editorial board, board of governors, current staff, committee members, and/or society editors for this issue of the publication.