November 2021 - Ai - Engineering Mailing Lists

Mentors for Local HS Robotics Team
by Fern, Alan Paul 30 Nov '21

30 Nov '21

Hi Everyone, I recently talked with the coach of the robotics team at our local Crescent Valley High School. This team takes part in the FIRST Robotics Competition and will be in competition development/preparation mode starting in early January. The coach thought that the team could really benefit by having some grad student mentors in robotics and AI. Due to COVID the mentoring this year would need to be virtual, but it would still be very much appreciated. If you are interested in this, please contact me and we can figure things out from there. Given the OSU strength in robotics and AI it would be great if we could help our local high schools be world class in these competitions. Thanks, -Alan

1 0

[CoRIS-List] New Course (W'22): AI539 St / Safe and Reliable Autonomy
by Fern, Alan Paul 27 Nov '21

27 Nov '21

Hi Everyone, I wanted to make sure that everyone was aware that one of our newest AI faculty, Professor Sandhya Saisubramanian will be offering a new course this Winter 2022. It was added to the catalog late, so you may have missed it when considering courses. The course is AI 539 ST/Safe and Reliable Autonomy which is a very important topic these days. See below for more details. Regards, -Alan -------------- AI 539 ST/Safe and Reliable Autonomy Course description: This course will highlight the challenges and recent trends in intelligent decision-making to improve the safety and reliability of autonomous systems, and identify important open research questions. This is a seminar-style course in which each student will summarize the papers in the reading list, lead the discussions on analyzing the strengths and weaknesses of each paper, and apply the learned concepts in a final project. The reading list will contain a mix of papers that cover both theory and applications of safe and reliable autonomy, published at leading AI/ robotics conferences and journals. Application areas include robotics, autonomous cars, and healthcare, among others. The topics covered will broadly include techniques for safe sim-to-real transfer; using human assistance to improve system reliability, including human feedback and imitation learning; explainability to improve reliability; and safe exploration. More details about the course are available at https://www.sandhyasai.com/ai-539-st-safe-and-reliable-autonomy<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.sandh…>.

1 0

No AI seminar today
by Prasad Tadepalli 24 Nov '21

24 Nov '21

Dear All, Please note that there is no AI seminar today. Next week we are going to have a seminar by Noam Brown from Facebook AI Research on the topic of imperfect information games. Happy Thanksgiving! Prasad

1 0

FW: Tech Talk Tuesday: Building Secure and Reliable Deep Learning Systems from a Systems Security Perspective
by Fern, Alan Paul 22 Nov '21

22 Nov '21

Hi Everyone, I think many of you might find the EECS Tech Talk tomorrow (Tuesday) at 11am interesting. It is from one of our new faculty at he intersection of security and ML. Details below. -Alan From: Tina Batten <tina.batten(a)oregonstate.edu> Sent: Wednesday, November 17, 2021 8:15 AM To: Fern, Alan Paul <Alan.Fern(a)oregonstate.edu> Subject: Tech Talk Tuesday: Building Secure and Reliable Deep Learning Systems from a Systems Security Perspective View this email in your browser<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailchi.m…> [https://mcusercontent.com/4ed5c72a58ff81119658f624f/images/5aee7181-8630-4c…] TECH TALK TUESDAY [https://mcusercontent.com/4ed5c72a58ff81119658f624f/images/856130cc-f3f1-fc…] Building Secure and Reliable Deep Learning Systems from a Systems Security Perspective Sanghyun Hong Assistant Professor Computer Science Oregon State University Tuesday, November 23, 2021 Talk: 11:00-11:30 a.m. PST | Q/A: 11:30-11:45 Zoom link: https://beav.es/tech-talk<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Info: https://eecs.oregonstate.edu/tech-talk-tuesday<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Abstract How can we build secure and reliable deep learning systems, e.g., autonomous cars or AI-assisted robotic surgery, for tomorrow? We cannot answer this question without understanding the worst-case behaviors of deep neural networks (DNNs), i.e., the core component of those systems. Recent work studied the worst-case behaviors, such as mispredictions caused by adversarial examples or models altered by data poisoning. However, most of the prior work narrowly considers DNNs as an isolated mathematical concept, and it overlooks a holistic picture, e.g., leaving out the threats caused by hardware-level attacks. In this talk, I will discuss my work, studying computational properties of DNNs from a systems security perspective, that has exposed critical security threats and has steered the industrial practices. First, I will present my work exposing the false sense of security: DNNs are not resilient to parameter perturbations. An adversary can inflict an accuracy drop up to 100% with a single bit-flip in its memory representation. Second, I will show how brittle the computational savings provided by efficient deep learning techniques are in adversarial settings. By adding human-imperceptible input perturbations, an attacker can completely offset a multi-exit network's computational savings on an input. Third, I will show how privacy-protection mechanisms, offered without a holistic picture, can put millions of users under a serious privacy threat. These mechanisms do not admit any arms race and eventually give the upper hand to an adversary. Finally, I will conclude by discussing how these results opened up new research directions and have steered industrial practices. Speaker Bio Sanghyun Hong is an Assistant Professor in Computer Science at Oregon State University. His research interest lies at the intersection of computer security, privacy, and machine learning. His current research focus is to study the computational properties of DNNs from a systems security perspective. He also works on identifying distinct internal behaviors of DNNs, such as network confusion or gradient-level disparity, whose quantification led to defenses against backdooring or data poisoning. You can find more about Sanghyun at https://sanghyun-hong.com<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>. Copyright (c) 2021 Oregon State University, All rights reserved. School of Electrical Engineering and Computer Science Oregon State University 1148 Kelley Engineering Center 110 SW Park Terrace Corvallis, OR 97331-5501 Phone: (541) 737-3617 | eecs.oregonstate.edu<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Want to change how you receive these emails? You can update your preferences<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> or unsubscribe from this list<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…>.

1 0

Alignment Reading Group: Objective Robustness in Deep Reinforcement Learning
by Pope, Quintin 22 Nov '21

22 Nov '21

Hello everyone, The next alignment reading group meeting will again be at 2 PM PST this Friday. We will discuss Objective Robustness in Deep Reinforcement Learning <https://arxiv.org/abs/2105.14111>. Anyone interested in the reliability and safety of current reinforcement learning systems should feel free to attend. Paper abstract: > We study objective robustness failures, a type of out-of-distribution > robustness failure in reinforcement learning (RL). Objective robustness > failures occur when an RL agent retains its capabilities > out-of-distribution yet pursues the wrong objective. This kind of failure > presents different risks than the robustness problems usually considered in > the literature, since it involves agents that leverage their capabilities > to pursue the wrong objective rather than simply failing to do anything > useful. We provide the first explicit empirical demonstrations of objective > robustness failures and present a partial characterization of its causes. > All the best, Quintin

1 0

[CoRIS-List] Slides from Hello Robot talk
by Ross L. Hatton 17 Nov '21

17 Nov '21

Hi All, Charlie Kemp and Aaron Edsinger sent out a link to their slide deck from last Friday's talk: https://docs.google.com/presentation/d/1csH7mxA_iV1ft2opNRjldRgpzEYuVbYQymR… =========== Ross L. Hatton Associate Professor, Robotics and Mechanical Engineering Collaborative Robotics and Intelligent Systems Institute Oregon State University coris.oregonstate.edu research.engr.oregonstate.edu/lram/ rosslhatton.com ross.hatton(a)oregonstate.edu

1 0

AI seminar: Nov 17 2021
by Mangannavar, Rajesh Devaraddi 17 Nov '21

17 Nov '21

Dear all, Our next AI seminar on *"Learning Transferable Visiolingustic Representations" *by professor Stefan Lee is scheduled to be on November 17th (Tomorrow), 1-2 PM PST. It will be followed by a 30 minute Q&A session by the graduate students. Note that this is an in-person/zoom hybrid event and will be held at *CORD 1109*. Zoom Link: https://oregonstate.zoom.us/j/93591935144?pwd=YjZaSjBYS0NmNUtjQzBEdzhPeDZ5U… *Learning Transferable Visiolingustic Representations* Stefan Lee Assistant Professor Computer Science Oregon State University *Abstract:* Vision-and-language research consists of a diverse set of tasks. While underlying image datasets, input/output APIs, and model architectures vary across tasks, there exists a common need to associate imagery and text -- i.e. to perform visual grounding. Despite this, the standard paradigm has been to treat each task in isolation -- starting from separately pretrained vision and language models and then learning to associate their outputs as part of task training. This siloed approach fails to leverage grounding supervision between tasks and can result in myopic groundings when datasets are small or biased. In this talk, I'll discuss a line of work focused on learning task-agnostic visiolinguistic representations that can serve as a common foundation for many vision-and-language tasks. First, I'll cover recent work on learning a generic multimodal encoder (ViLBERT) from large-scale web data and transferring this pretrained model to a range of vision-and-language tasks. Second, I'll show how multitask training from this base architecture further improves task performance while unifying twelve vision-and-language tasks in a single model. *Speaker Bio:*Stefan Lee is an assistant professor in the School of Electrical Engineering and Computer Science at Oregon State University and a member of the Collaborative Robotics and Intelligent Systems (CoRIS) Institute there. His work addresses problems at the intersection of computer vision and natural language processing. He is the recipient of the DARPA Rising Research Plenary Speaker Selection (DARPA 2019), two best paper awards (EMNLP 2017, CVPR 2014 Workshop on Egocentric Vision), and multiple awards for review quality (CVPR 2017,2019,2020; ICCV 2017; NeurIPS 2017-2018; ICLR 2018-2019, ECCV 2020). Also, everyone is encouraged to attend the talk in person if you can. Hopefully, it helps bring us closer to normalcy and prepares us for more in-person talks for the next term. *Please watch this space for future AI Seminars :* * https://eecs.oregonstate.edu/ai-events <https://eecs.oregonstate.edu/ai-events>* Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> The AI Seminar has a strict "no electronics" and "no recordings" policy. -> For graduate students in the AI program, attendance is strongly encouraged.

2 1

Meeting time for AI alignment reading group
by Pope, Quintin 12 Nov '21

12 Nov '21

Hello everyone, The survey results are in. Our first meeting will be at 2 PM (PST) on Friday, November 19th, and will repeat weekly. For our first discussion, I thought we'd start with something short and optimistic: Alignment by Default <https://www.lesswrong.com/posts/Nwgdq6kHke5LY692J/alignment-by-default>. It argues that powerful models may have internal "natural abstractions" that straightforwardly represent human values. Some combination of finetuning, interpretability tools and luck may be enough to "wire up" the human values representation to the model's output and thereby get an aligned model. I'm interested in how plausible alignment by default seems to everyone, whether there are any architecture or training modifications we can make that raise the odds of alignment by default, and in how to best manage the "wire up" step where we get a globally aligned model out of a model with a human values subcomponent. Anyone who's at all interested is welcome to join! Join Zoom Meeting https://oregonstate.zoom.us/j/95843260079?pwd=TzZTN0xPaFZrazRGTElud0J1cnJLU… Password: 961594 Phone Dial-In Information +1 971 247 1195 US (Portland) +1 253 215 8782 US (Tacoma) +1 301 715 8592 US (Washington DC) Meeting ID: 958 4326 0079

1 0

[CoRIS-List] OSU Robotics Seminar, Charlie Kemp and Aaron Edsinger and Friday November 12
by Ross L. Hatton 12 Nov '21

12 Nov '21

Hi All, I'm pleased to announce that this week we will have an OSU Robotics seminar from Aaron Edsinger and Charlie Kemp of the mobile-manipulation company Hello Robot. (Dr. Kemp is also a professor at Georgia Tech). The seminar will be at 10am Pacific on Friday, November 12. ---- Drs. Edsinger and Kemp will be joining us remotely via Zoom, at https://oregonstate.zoom.us/j/93958424124?pwd=bVlhZlRCbTFOaHIwZUx1R1h6d205Z… ---- If you would like to attend the seminar in person with others at OSU, I will have LINC 302 (the seminar room for this term) set up for Prof. Koppal's presentation, using the new classroom Zoom capabilities. If you would prefer not to attend in person, but would like some unstructured interaction time with others in the OSU Robotics community, I encourage you to join the seminar's Gather space before or after the seminar, https://gather.town/app/HeC6WAB1zL6cKWp5/OSU%20Robotics%20Seminar =================== Title Hello Robot: Democratizing Mobile Manipulation Abstract Mobile manipulators have the potential to improve life for everyone, yet adoption of this emerging technology has been limited. To encourage an inclusive future, Hello Robot developed the Stretch RE1, a compact and lightweight mobile manipulator for research that achieves a new level of affordability. The Stretch RE1 and Hello Robot’s open approach are inspiring a community of researchers to explore the future of mobile manipulation. In this talk, we will present the Stretch RE1, its academic origins, and the growing community and ecosystem around it. We will also touch on connections with Oregon State University, including participation in our first summer internship program, and Hello Robot’s role in the new NSF AI Institute, AI-CARING. We will conclude with a live demo and a question and answer session. Bios Aaron Edsinger, CEO and Cofounder Aaron has a passion for building robots and robot companies. He has founded four companies focused on commercializing human collaborative robots. Two of these companies, Meka Robotics and Redwood Robotics, were acquired by Google in 2013. As Robotics Director at Google, Aaron led the business, product, and technical development of two of Google’s central investments in robotics. Aaron received his Ph.D. from MIT CSAIL in 2007. Charlie Kemp, CTO and Cofounder Charlie (https://charliekemp.com) is a recognized expert on mobile manipulation. In 2007, he founded the Healthcare Robotics Lab at Georgia Tech, where he is an associate professor in the Department of Biomedical Engineering. His lab has conducted extensive research on mobile manipulators to assist older adults and people with disabilities. Charlie earned a B.S., M.Eng., and Ph.D. from MIT. He first met Aaron while conducting research in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). =========== Ross L. Hatton Associate Professor, Robotics and Mechanical Engineering Collaborative Robotics and Intelligent Systems Institute Oregon State University coris.oregonstate.edu research.engr.oregonstate.edu/lram/ rosslhatton.com ross.hatton(a)oregonstate.edu

2 1

AI seminar: Nov 10 2021
by Mangannavar, Rajesh Devaraddi 10 Nov '21

10 Nov '21

Dear all, Our next AI seminar on *"**Estimating Long-term Rewards by Off-policy Reinforcement Learning" *by Lihong Li is scheduled to be on November 10th (Tomorrow), 1-2 PM PST. There will be *NO *graduate student Q&A session. Hence everyone is encouraged to ask questions during the seminar. Note that this is a zoom event. Zoom Link: https://oregonstate.zoom.us/j/93591935144?pwd=YjZaSjBYS0NmNUtjQzBEdzhPeDZ5U… *Estimating Long-term Rewards by Off-policy Reinforcement Learning * Lihong Li Senior Principal Scientist Amazon *Abstract*: One of the core problems in reinforcement learning (RL) is estimating the long-term reward of a given policy. In many real-world applications such as healthcare, robotics and dialogue systems, running a new policy on users or robots can be costly or risky. This gives rise to the need for off-policy, or counterfactual, estimation: estimate the long-term reward of a given policy using data previously collected by another policy (e.g., the one currently deployed). This talk will describe some recent advances in this problem, for which many standard estimators suffer an exponentially large variance (known as "the curse of horizon"). Our approach is based on a dual linear program formulation of the long-term reward, and can be extended to estimate confidence intervals. *Bio*: Lihong Li is a Senior Principal Scientist at Amazon. He obtained a PhD degree in Computer Science from Rutgers University. After that, he held research positions in Yahoo!, Microsoft and Google, before joining Amazon. His main research interests are in reinforcement learning, including contextual bandits, and other related problems in AI. His work is often inspired by applications in recommendation, advertising, Web search and conversational systems. Homepage: http://lihongli.github.io *Please watch this space for future AI Seminars :* * https://eecs.oregonstate.edu/ai-events <https://eecs.oregonstate.edu/ai-events>* Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> The AI Seminar has a strict "no electronics" and "no recordings" policy. -> For graduate students in the AI program, attendance is strongly encouraged.

2 1