SystemX

Computing systems have become ubiquitous and critical to modern business, government, military, and life. These systems are becoming wickedly complex, including monstrous amounts of poorly-understood interacting subsystems, each with their own daunting amounts of software, firmware, and/or hardware. As we trust such systems with our secrets, our money, and our lives, how can we increase our assurance that these systems will do what they are designed to do, and do nothing else malicious? Can we trust the autopilot in the airplane you hope to get on someday? Can we trust the self-driving car you hope to get in someday? Can we trust the medical device you hope to wear or implant someday? Can we trust the financial systems you hope to have substantial money in someday? Good news: there are paths toward such trustworthy systems. We will discuss some of these paths forward, including how one can build trustworthy systems from untrustworthy components.

Neurological and psychiatric disorders are dramatically increasing in prevalence due to aging population and social isolation. However, to date, there is no cure for any of the brain disorders. The goal of brain disorder treatments is to restore the brain's function. Therefore, a key challenge is to quantify the brain function underlying behavior. Once the brain function algorithms underlying behaviors of interest can be quantitatively defined, minimizing the normal and diseased brain function difference can be defined as the objective function for the brain disorder treatment. The variables then can be optimized to minimize the objective function. In order to quantify the brain function algorithms underlying behavior, cell type specific whole brain function measurements are necessary. We utilize optogenetics combined with fMRI (ofMRI) to enable such measurements. Through computational modeling of ofMRI data, the functional interactions among different regions of the brain was then quantified. In combination with electrophysiological measurements, we further model brain function at a cellular level. In order to further understand the circuit, pathology relationship, we also utilize brain clearing methods to longitudinally quantify and model pathology. Through these efforts, we aim to enable systematic design of therapeutic interventions necessary for the treatment of brain disorders.

Join Dr. Charles Chu from TSMC for a virtual information session. He will discuss Foundry business model and TSMC's unique position in the semiconductor ecosystem, and how they delight their customers through technology leadership and manufacturing excellence. He will also share the progress of the Arizona Fab as TSMC continues its investment in advanced technology in the US.

Mainstream computing has gone through multiple epochs, and the latest is now upon us. The importance of energy and space efficiency, power density, and the integration of application-driven acceleration has made mobile computing a new driver of compute architecture—and the next generation of mainstream computing.

Humans are good at engaging in conversation and understanding written text, so much so that these sorts of natural language processing (NLP) tasks are widely considered essential requirements for any truly intelligent artificial agent. The NLP community has made a great deal of progress over the last few years on certain challenges (for example, the SuperGLUE benchmarks and Winograd Schemas), thanks to the power of neural networks and pre-trained models like BERT and GPT-3. Nevertheless, existing systems still don't approach human performance for useful tasks such as answering general, open-ended questions about stories or participating in dialogue.

In this talk, I analyze why NLP's recent achievements do not generalize to true language understanding. I suggest a broad remedy – automated large-scale extraction of knowledge from text and reasoning with that knowledge, together with existing background information, to achieve deeper understanding – and propose methods to achieve this. I focus on texts that have exploitable structure, including bulleted lists and tables, and discuss the techniques I've used for automating understanding of regulatory and molecular biology texts. Finally, I discuss how comprehending the goal-action-effect structure implicit in narrative texts, including news articles and business case studies, can help extract useful general knowledge about actions and plans, and consider how these techniques could contribute to AI succeeding at the difficult tasks of story understanding and analysis.

Microwave and millimeter-wave (mmWave) Imaging Systems have witnessed huge progress during the last two decades, thanks to the advances in the semiconductor miniaturization in analog as well as digital circuits. Notwithstanding the ever-increasing demands for mmWave imaging solutions serving in the industrial, security, and medical domains, many efforts have been successfully accomplished to mature mmWave imaging technologies in order to qualify for the real world problems. Many more methods are still yet struggling to see the light away from the lab bench and the computer simulators. This talk aims to present and overview on latest mmWave multistatic imaging technologies and to help raise the awareness on the latest trends and challenges.

As we map out the frontiers of innovation, quantum computing (QC) looms large as possibly the most important scientific invention of the 21st century. QC upends what we thought to be fundamental principles of computer science, promising to execute entirely new classes of algorithms beyond the practical capabilities of conventional computers. The technology was initially explored in the 1980s when, at a series of lectures at MIT and Caltech, Berkeley Professor Paul Benioff defined the concept of a quantum circuit, and Richard Feynman floated the basic model for building such a computer. The technology had struggled to mature throughout the last few decades, but has started accelerating the recent years. In this talk, we will explore the current landscape of commercial entities pursuing the development of quantum computers and ponder on how they might change the future.

Quantum computers promise a new paradigm of computation where information is processed in a way that has no classical analogue. However, the known problems for which quantum computers offer a computational advantage, require long gate sequences and large number of qubits. Error-correction codes and fault-tolerant gate implementation require the encoding of logical qubits on a large number of physical qubits, and given the overheads, it is expected that general purpose quantum computers will have millions of physical qubits, thus requiring an underlying qubit technology that can be manufactured at scale. Photons make great qubits, they are cheap to produce, resilient to noise and the only known option for quantum networks. Most crucially, they can be efficiently manipulated with silicon photonics, an intrinsically scalable and manufacturable platform in which all the fundamental quantum gates can be implemented. In this talk, I will describe an architecture for universal fault-tolerant quantum computing based on linear optics, in the process I will explain how measurement-induced non-linearity can overcome the challenge of creating entanglement and how loss can be effectively tackled with error correcting codes.

In the last decade, we have experienced a significant leap forward in computer vision for tasks such as object recognition, reconstruction, 3D vision, detection, and tracking, where artificial systems have reached human and sometimes even "superhuman" performance. These advances result from the combination of novel machine learning algorithms, more powerful computers and large-enough curated datasets, which allow for off-line learning. This success has not yet transferred to robotics. In robotics, despite significant progress, machines are still far behind humans when it comes to physical and purposeful interaction with unstructured environments. Humans largely exploit physical interactions with the surroundings to solve complex long-horizon tasks (cooking, cleaning, tidying, assembling...) in "perceptually dirty", cluttered, uncontrolled, and often unknown environments such as homes, and offices. On the contrary, most of the state-of-the-art robotic solutions are restricted to short-horizon tasks in "perceptually clean" environments and try to minimize the interaction with the surroundings.

Despite all the challenges of physical interaction, in my research at SVL I advocate to consider interactions with the environment as part of the solution instead of the problem. In my talk I will present work in our group exploiting physical interaction to solve robotic tasks. I will present our work on Interactive Navigation, tasks where the robotic agent needs to interact with the environment (e.g. open doors, push away obstacles) to achieve a desired location. Solving this type of navigation is necessary to move in common human uncontrolled environments such as our homes. I will also present our work on Mechanical Search, where we equip robots with skills to search efficiently for target objects in piles of cluttered objects using physical interaction (pushing other objects, grasping them), a problem that is faced frequently not only in home robotics but also in logistic domains. And finally, I will present iGibson, SVL's large effort to provide interactive agents with a simulation environment to train and test interactive AI solutions, and demonstrate that our approach outperforms state-of-the-art learning-based and classical methods on real-world data while maintaining efficiency.

Over the next few years, food making robots are expected to be prevalent in our lives. The speaker describes this trend, shows video demonstrations of some commercially available food making robots and indicates what we could expect in the future.