Select a calendar:
Filter September Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for September
-
Towards Trustworthy Physical AI Generalists
Thu, Sep 12, 2024 @ 10:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Ding Zhao , Associate Professor & Dean's Early Career Fellow - Carnegie Mellon University
Talk Title: Towards Trustworthy Physical AI Generalists
Abstract: Large language models like ChatGPT have shown that generalist foundation models can effectively tackle long-horizon tasks by training on extensive text data from the internet. It is anticipated that larger-scale data from the physical world, such as those generated by autonomous vehicles and the healthcare industry, could drive the next wave of AI development. A common challenge in deploying highly intelligent agents at scale in the physical world is ensuring their safety. In this talk, I will present our efforts to establish Trustworthy Physical AI Generalists to support this crucial transformation. I will explore the challenges of ensuring safety and generalization in the development of trustworthy AI, and discuss potential solutions, including rare event analysis, safe reinforcement learning, hierarchical generative models for task identification and transferability, and causal reasoning to improve generalizability. Additionally, I will discuss the advantages and challenges of using LLMs to develop physical AI generalists. I will introduce applications of our work in heart attack detection and acute care, self-driving technology, and robotic autonomy for assisting seniors and conducting safety-critical tasks related to climate change resilience.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
https://usc.zoom.us/j/99488778795?pwd=oXg76V89VYG9b5I0CIEcn2E2Fz7d6z.1
Meeting ID: 994 8877 8795
Passcode: 868727
Biography: Ding Zhao is an Associate Professor and Dean's Early Career Fellow at Carnegie Mellon University, where he leads the Safe AI Lab. His research focuses on developing Trustworthy Physical AI Generalists for high-stakes applications at scale. Prof Zhao was invited by Uber ATG to enhance fleet safety following the world’s first fatal autonomous vehicle collision. Zhao collaborates with leading industry partners, including Google, Nvidia, Amazon, Apple, Microsoft, IBM, Ford, Uber, Bosch, Toyota, and Rolls-Royce. He has grants from NSF, DOT, DOE, and DARPA and published over 120 papers in top venues such as ICML, NeurIPS, ICLR, AISTATS, CoRL, ICRA, IROS, and Nature Communications. Zhao has mentored 20 Ph.D. students and 7 postdocs, with 7 of them becoming faculty members in academia. Zhao has received numerous awards, including CMU Dean's Early Career Fellow Professorship, Provost's Inclusive Teaching Fellows Award, National Science Foundation CAREER Award, MIT Technology Review 35 Under 35 Award in China, George N. Saridis Best IEEE Transactions Paper Award, George Tallman Ladd Research Award, Struminger Teaching Award, Ford University Collaboration Award, Qualcomm Innovation Award, Carnegie-Bosch Research Award, and various industrial fellowship awards from Google DeepMind, Adobe, Toyota, and Bosch. His work has garnered attention from media outlets such as the New York Times, Forbes, TIME, IEEE Spectrum, Popular Science, Telegraph, and Wired.
Host: Assistant Prof. Yue Wang
Webcast: https://usc.zoom.us/j/99488778795?pwd=oXg76V89VYG9b5I0CIEcn2E2Fz7d6z.1Location: Olin Hall of Engineering (OHE) - 132
WebCast Link: https://usc.zoom.us/j/99488778795?pwd=oXg76V89VYG9b5I0CIEcn2E2Fz7d6z.1
Audiences: Everyone Is Invited
Contact: Thomas Lord Department of Computer Science
-
Geometric Regularizations for 3D Shape Generation
Mon, Sep 16, 2024 @ 10:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Qixing Huang , Associate Professor, Computer Science Department - University of Texas at Austin
Talk Title: Geometric Regularizations for 3D Shape Generation
Abstract: Generative models, which map a latent parameter space to instances in an ambient space, enjoy various applications in 3D Vision and related domains. A standard scheme of these models is probabilistic, which aligns the induced ambient distribution of a generative model from a prior distribution of the latent space with the empirical ambient distribution of training instances. While this paradigm has proven to be quite successful on images, its current applications in 3D generation encounter fundamental challenges in the limited training data and generalization behavior. The key difference between image generation and shape generation is that 3D shapes possess various priors in geometry, topology, and physical properties. Existing probabilistic 3D generative approaches do not preserve these desired properties, resulting in synthesized shapes with various types of distortions. In this talk, I will discuss recent work that seeks to establish a novel geometric framework for learning shape generators. The key idea is to model various geometric, physical, and topological priors of 3D shapes as suitable regularization losses by developing computational tools in differential geometry and computational topology. We will discuss the applications in deformable shape generation, latent space design, joint shape matching, and 3D man-made shape generation. This research is supported by NSF IIS 2413161.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
https://usc.zoom.us/j/93012253116?pwd=4bJJFbbbfblFHRjabBBvvCuavDml6J.1
Meeting ID: 930 1225 3116
Passcode: 570060
Biography: Qixing Huang is an associate professor with tenure at the computer science department of the University of Texas at Austin. His research sits at the intersection of graphics, geometry, optimization, vision, and machine learning. He has published more than 100 papers at leading venues across these areas. His research has received several awards, including multiple best paper awards, the best dataset award at Symposium on Geometry Processing 2018, IJCAI 2019 early career spotlight, multiple industrial and NSF awards, and 2021 NSF Career award. He has also served as area chairs of CVPR, ECCV, ICCV and technical papers committees of SIGGRAPH and SIGGRAPH Asia, and co-chaired Symposium on Geometry Processing 2020.
Host: Assistant Prof. Yue Wang
Webcast: https://usc.zoom.us/j/93012253116?pwd=4bJJFbbbfblFHRjabBBvvCuavDml6J.1Location: Olin Hall of Engineering (OHE) - 136
WebCast Link: https://usc.zoom.us/j/93012253116?pwd=4bJJFbbbfblFHRjabBBvvCuavDml6J.1
Audiences: Everyone Is Invited
Contact: Thomas Lord Department of Computer Science
-
Equalizing Hollywood: Using AI to Make Professional Grade Art & Content That Sells & Impacts
Wed, Sep 18, 2024 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Stephen Gibler, Adjunct Professor, USC School of Cinematic Arts
Talk Title: Equalizing Hollywood: Using AI to Make Professional Grade Art & Content That Sells & Impacts
Abstract: In “Equalizing Hollywood,” discover how AI is democratizing the creation of professional-grade art and content. This talk explores innovative AI tools that empower creators to produce high-quality work that not only sells but also makes a significant impact. Learn about the transformative potential of AI in leveling the playing field in the entertainment industry. Join us to see how technology is shaping the future of Hollywood and beyond.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Stephen Gibler has been a producer in Los Angeles for over a decade, running production for over 50 projects, including eight feature films, over a dozen commercials, 30 new media projects, 5 reality TV shows, and one of the largest immersive art museums in China ’The Silos’. He has collaborated with renowned figures in the film industry, such as James Ivory, Jackie Earle Haley, Haley Joel Osment, Molly Ringwald, Drake Doremus, and many others while also working with brands such as Amazon, Lancome, Head & Shoulders, Fiverr, AMC, and so forth. Stephen now serves as an adjunct professor at USC’s School of Cinematic Arts and is the founder of AI Tech Startup, Logline AI, aiming to use AI to accelerate the creative process in filmmaking.
Host: CAIS
More Info: https://cais.usc.edu/events/usc-cais-seminar-with-stephen-gibler/
Location: Montgomery Ross Fisher Building (school Of Social Work) (MRF) - 102
Audiences: Everyone Is Invited
Contact: Thomas Lord Department of Computer Science
Event Link: https://cais.usc.edu/events/usc-cais-seminar-with-stephen-gibler/
-
A Code Generation Framework To Replicate Software Design Concepts
Mon, Sep 30, 2024 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Prof. George Heineman, Associate Professor, Computer Science - Worcester Polytechnic Institute
Talk Title: A Code Generation Framework To Replicate Software Design Concepts
Abstract: In 1998, Philip Wadler used the term Expression Problem (EP) to describe a common situation that occurs when devising software that must evolve, specifically with regard to the structure of the data types and the operations over these data types. Over time, software engineers extend systems by adding new data types and/or new operations, and the goal is to avoid changing existing code as part of the extension. Dozens of researchers have investigated approaches to EP using a variety of programming languages. The papers from the research literature often contain only small code fragments and the lack of a common benchmark makes it difficult to compare different approaches with each other. We designed EpCoGen, a code generation framework, to replicate the results of numerous papers using a rich benchmark domain of mathematical expressions. While it was not our intention, in completing this project, we devised a novel CoCo approach to EP based on Covariant Conversions. This result would not have been possible without the meticulous effort in both developing a comprehensive benchmark and trying to replicate existing results from the literature. We generate fully coded solutions in Java, Scala and Haskell, with accompanying test cases, for nine different approaches to EP using a language-independent code generation framework (CoGen) that can be expanded to include additional languages and approaches.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: George Heineman is an Associate Professor of Computer Science at WPI in Worcester, Massachusetts. His research interests include component-based software engineering, modularity, code generation, and algorithms. He is the author of Algorithms in a Nutshell (2ed, O’Reilly Media, 2016), Learning Algorithms (O’Reilly Media, 2021), and an online training video course, Coding Interview Preparation: Learn to Solve Algorithms Problem to Land Your Next Software Role (O’Reilly Media, 2024). George is also an avid puzzler with a lifelong interest in logical and mathematical puzzles. He is the inventor of Sujiken® puzzles (a variation of Sudoku), Trexiken puzzles (a variation of Ken-Ken®) and the online https://wordygame.net, if you are ready to try a more challenging word puzzle than Wordle.
Host: Nenad Medvidovic, Department Chair - USC Thomas Lord Department of Computer Science
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Thomas Lord Department of Computer Science
