Human-Centered AI: Integrating Human Intelligence for Robust, Ethical, and Explainable AI Systems

This lecture introduces Human-Centered Artificial Intelligence (AI) as a crucial paradigm for the 21st century, arguing that purely learning-based approaches, while successful, will inevitably hit fundamental walls regarding safety, fairness, and explainability. The core premise is that deep integration of human beings into both the training and operational phases of AI systems is essential for them to learn effectively and function reliably in the real world. This involves a shift from traditional machine learning, where humans primarily provide brute-force data annotation, to "machine teaching," where the AI actively queries humans for the most useful, sparse subsets of data, significantly reducing annotation effort.

The framework of Human-Centered AI is broken down into five grand challenges across two main phases: the learning phase and the real-world operation phase. During the learning phase, research focuses on machine teaching for efficient supervised learning and continuous reward engineering, where humans iteratively adjust AI's loss functions to align with societal values and prevent unintended consequences. For real-world operation, the challenges include human sensing (AI perceiving and understanding human states), interaction (creating rich, collaborative experiences between humans and AI), and ensuring safety and ethics through constant human supervision and the AI's ability to quantify its own uncertainty.

The speaker elaborates on the necessity of human supervision due to the inherent uncertainty and incomplete information in AI's generalization from data, making systems not provably safe, fair, or explainable. This necessitates human oversight at every step, from objective and subjective data annotation (leveraging individual and crowd intelligence) to real-time monitoring of AI decisions, especially in critical applications like autonomous vehicles or medical diagnosis. The lecture emphasizes that AI systems must be designed to express their degree of uncertainty, prompting human intervention when confidence is low or decisions are high-stakes.

Several grand challenges are proposed to define progress in Human-Centered AI, such as training object detection models on vastly different data sources (e.g., Wikipedia for COCO), achieving high accuracy with minimal training examples (e.g., MNIST with one digit), developing AI that can replace complex human decision-making bodies like the US Congress, building robust emotion recognition systems, and creating truly collaborative human-robot interactions over billions of miles or prolonged conversations (Turing test). Ultimately, the vision is for AI systems that not only perform tasks but also understand their limitations, communicate their uncertainty, and continuously learn and adapt under human guidance, embodying human values and preventing catastrophic outcomes.

"The underlying first prediction under the idea of human-centered AI in this century is that the learning-based approaches have been successful over the past two decades, like deep learning, machine learning approaches that learn from data, are going to continue to become better and dominate the real-world applications."

"The selection of data based on which to learn, I believe, is the critical direction of research where we have to solve in order to create truly intelligent systems and ones that are able to work in the real world."

"No matter how much we want to, these systems will not be provably safe... will not provably fair... And it will not be explainable."

"The machine queries a human with questions, and therefore the task is, and this is a wide open research field, the task is to minimize in several orders of magnitude the amount of data that needs to be annotated."

"This is what I believe is going to be the defining mode of operation for AI systems in the 21st century, is we won't be able to as much as we'd like to escape, to create perfect AI systems that escape the need to work together with human beings at every step."

"Systems that have to operate in the real world have to understand what our society deems as good and bad, and we're not always good at injecting that in the very beginning, that has to be a continuous process of adjusting the rewards, of reward re-engineering by humans so that we can encode human values into the learning process."

"If you build a system that has a high accuracy of doing real emotion recognition you can think of it, as stated here, an AI system that classifies, binary classification problem with 95% accuracy of whether you want to be left alone or not."

"The grand challenge there, really it all boils down to the ability of an AI system to say that it's uncertain about something. And that measure of uncertainty has to be good."