東北大学人工知能エレクトロニクス卓越大学院プログラム

Abstract

The potential of quantum information science is fueling demand for the design and generation of new qubits and devices operating at the single-particle level. One promising approach uses photons as information carriers. Their long decoherence lengths and fast operation speeds offer natural benefits for rapid and safe quantum communication; however, current state-of-the-art single photon sources are either bulky and poor choices for circuit integration or operate at cryogenic temperatures. On the other hand, color centers, such as defects in two-dimensional (2D) materials, are suitable for small-scale devices that can operate at room temperature, but these sources have their own limitations, particularly their restricted brightness due to luminescence lifetimes and intermediate dark states, and the control of excitation repetition rate. Moreover, we still lack some of the most fundamental understanding of these defects, and how structural properties correlate to their optical ones. To address the first set of challenges, we work to controllably and reproducibly couple these defects to plasmonic nano-antennas, which can confine light below the diffraction limit, and thus can be used to strongly manipulate quantum emitters and couple them to existing electronic circuits. For the second, we use machine learning to better understand the material properties that are relevant to our applications.

Abstract

In this speech, we explore the technical advancements in Document Intelligence, a pivotal AI technology driving enterprise digital transformation. Document Intelligence leverages state-of-the-art machine learning models to automatically and accurately extract and interpret unstructured data from documents, including text, key-value pairs, tables, and structural information.

The presentation traces our journey from optical character recognition (OCR) capabilities to fine-tuned models tailored for specific document types, culminating in the recent integration of generative AI and large language models (LLMs). This paradigm shift enhances contextual understanding and scalability, driving significant advancements in efficiency and precision.

Through a discussion of methodology, technological breakthroughs, and real-world applications, we provide insights into how these innovations are addressing complex challenges and enabling new possibilities.

Abstract

The year 2020 will be remembered by many generations as one of the most disastrous years in global history. We witnessed life-altering impacts of a pandemic during the warmest year on record along with wildfires, tropical cyclones, and other extreme hazards. The U.S. sustained historically the largest number of weather/climate disaster events with losses exceeding $1 billion in 2020 until 2023 broke this record. 2023 also set a new warmest-year record. Globally, we see an increasing trend of disasters. In response to this pressing challenge, what can we do? Can artificial intelligence (AI) that continually innovate private sectors revolutionize public sectors’ disaster management? This talk will highlight challenges and opportunities for AI in disaster management.

Abstract

We are witnessing a rise in generative AI technologies that make the creation of digital avatars automated and accessible to anyone. While it is clear that such AI technologies will benefit numerous digital human applications, from video conferencing to AR/VR, their successful adoption hinges on their ability to generalize to real-world data. In this talk, I discuss our efforts to build 3D digital human foundation models that can be used to create many downstream applications. First, I will share our recent efforts on learning to generate photorealistic 3D faces from a collection of in-the-wild 2D images. Furthermore, I will show how such photorealistic 3D synthetic data can be used to train another AI model for downstream applications, such as a single-view reconstruction of photorealistic faces. Lastly, I will address the emerging challenge of verifying the authenticity of synthetic avatars in a world where their legitimate use becomes ubiquitous.

Abstract

In both academia and industry, the exploration of 6G—a further advanced cellular communication technology—has already begun. As our society becomes increasingly digitalized, hyper-connected, and globally data-driven, future services are anticipated to rely heavily on virtually instantaneous, unlimited wireless connectivity. This presentation will outline the key drivers, challenges, and essential research topics associated with 6G. In particular, I will provide an overview of the cutting-edge technologies in this field, spanning aspects such as the physical layer, networking, and new use cases and service enablers.

Abstract

We are developing innovative methods for monitoring physiological conditions over extended periods of time, with the goal of improving healthcare and personal well-being. In this talk, we will introduce remote PPG (rPPG), an advanced technique that allows for the measurement of plethysmograms without any physical contact. We will also present our proposed methods for estimating physiological parameters, such as blood pressure and oxygen saturation, based on rPPG signals, and discuss their potential applications in both clinical and everyday settings.

Abstract

The Talent versus Luck model, awarded the 2022 Ig Nobel Prize, explores the roles of talent and luck in an individual’s career success through computer simulations. The model demonstrates that capital distribution has a heavy-tailed pattern, indicating that the most successful individuals are not necessarily the most talented. While the original model's findings are intriguing, they rely solely on numerical calculations, raising questions about their general validity. In this talk, I present a reformulated, simplified model and demonstrate that the resulting capital distribution follows a lognormal pattern, even when talent is normally distributed. Additionally, I analyze data from the Global Talent Competitiveness Index to examine changes in the rate of return on talent to GDP worldwide from 2013 to 2023.