- talks - lists.cs.princeton.edu

Princeton AI Alignment and Safety Seminar (PASS): Tuesday, March 19 at 2pm
by Emily C. Lawrence 18 Mar '24

18 Mar '24

Princeton AI Alignment and Safety Seminar (PASS) Catastrophic misalignment of large language models [ https://paulfchristiano.com/ | Paul Christiano, Alignment Research Center ] Tuesday, March 19 2:00 - 3:00 pm [ https://bit.ly/3IHlz4h | View the livestream ] Abstract: I’ll discuss two possible paths by which AI systems could be so misaligned that they attempt to deceive and disempower their human operators. I’ll review the current state of evidence about these risks, what we might hope to learn over the next few years, and how we could become confident that the risk is adequately managed. Bio: I run the [ http://alignmentresearchcenter.org/ | Alignment Research Center ] . I previously ran the language model alignment team at [ https://openai.com/ | OpenAI ] , and before that received my PhD from the [ http://theory.cs.berkeley.edu/ | theory group ] at UC Berkeley. You may be interested in my [ https://ai-alignment.com/ | writing about alignment ] , my [ http://sideways-view.com/ | blog ] , my [ https://scholar.google.com/citations?hl=en&user=B7oP0bIAAAAJ&view_op=list_w… | academic publications ] , or [ https://paulfchristiano.com/fun-and-games/ | fun and games ] . I am an advisor and board member at [ https://metr.org/ | METR ] , an external advisor to the [ https://www.gov.uk/government/publications/ai-safety-institute-overview/int… | UK AI Safety Institute ] , and a trustee of [ https://www.anthropic.com/index/anthropics-responsible-scaling-policy | Anthropic’s Long-Term Benefit Trust ] . Stay informed and receive seminar reminders by joining our mailing list: [ https://tinyurl.com/pass-mailing | https://tinyurl.com/pass-mailing ] Organized by: [ https://pli.princeton.edu/ | Princeton Language and Intelligence ]

1 0

CITP Lecture: Emily Tseng, Wednesday, March 20 at 12:30pm
by Emily C. Lawrence 15 Mar '24

15 Mar '24

CITP Lecture Emily Tseng '14, Cornell University Wednesday, March 20 12:30 p.m. 105 Computer Science [ https://citp.princeton.edu/event/tseng/ | https://citp.princeton.edu/event/tseng/ ] Title: Digital Safety and Security for Survivors of Technology-Mediated Harms Abstract: Platforms, devices, and algorithms are increasingly weaponized to control and harass the most vulnerable among us. Some of these harms occur at the individual and interpersonal level: for example, abusers in intimate partner violence (IPV) use smartphones and social media to surveil and stalk their victims. Others are more subtle, at the level of social structure: for example, in organizations, workplace technologies can inadvertently scaffold exploitative labor practices. This talk will discuss research (1) investigating these harms via online measurement studies, (2) building interventions to directly assist survivors with their security and privacy; and (3) instrumenting these interventions as observatories, to enable scientific research into new types of harms as attackers and technologies evolve. The talk will close by sharing a vision for centering inclusion and equity in digital safety, security and privacy, towards brighter technological futures for us all. Bio: Emily Tseng is a Ph.D. candidate in information science at Cornell University. Her research develops the systems, interventions, and design principles we need to make digital technology safe and affirming for everyone. Tseng’s work has been published at top-tier venues in human-computer interaction (ACM CHI, CSCW) and computer security and privacy (USENIX Security, IEEE Oakland). For five years, she has worked as a researcher-practitioner with the Clinic to End Tech Abuse, where her work has enabled specialized security services for over 500 survivors of intimate partner violence (IPV). She is the recipient of a Microsoft Research Ph.D. Fellowship, Rising Stars in EECS, Best Paper Awards at CHI, CSCW, and USENIX Security, and third place in the Internet Defense Prize. She previously interned with the Social Media Collective at Microsoft Research, and holds a bachelor’s degree from Princeton University. This talk is open to the public. If you need an accommodation for a disability please contact Jean Butcher at butcher(a)princeton.edu at least one week before the event.

1 0

CS and ECE SEMINAR 3/20/2024 - Max Simchowitz, MIT
by Emily C. Lawrence 14 Mar '24

14 Mar '24

ECE/CS Colloquium Speaker Speaker: Max Simchowitz, Massachusetts Institute of Technology Date: Wednesday, March 20 Time: 12:30pm EST Location: B205 E-Quad Host: Elad Hazan, Chi Jin Event page: https://ece.princeton.edu/events/mathematical-foundations-physical-agents Title: Mathematical Foundations for Physical Agents Abstract: From robotics to autonomous vehicles, machine learning agents deployed in the physical world (“physical agents”) promise to revolutionize endeavors ranging from manufacturing to agriculture to domestic labor. In this talk, we will develop mathematical foundations, from the ground up, for how to carry out this vision. We will begin our investigation by examining linear dynamical systems, a simple and fundamental model of the interaction between a physical agent and its environment. We prove mathematically that simple exploration attains optimal performance for some of both the simplest and the most complex learning problems in this class. The above finding, while powerful, strongly motivates moving past linear dynamics as a mathematical testbed for understanding learning with physical agents. Hence, we turn to providing mathematical guarantees for a setting of real-world importance that does not fit the linear mold: behavior cloning. Behavior cloning — teaching a robot to imitate from example demonstrations — lies at the heart of many of today’s most promising robot learning endeavors due to its intuitive data collection and simplicity. Though it can work incredibly well, we still do not have a clear understanding of what circumstances ensure its success. Bringing together the flexibility of generative models with key intuitions arising from the study of linear control, we introduce a framework for behavior cloning that enables an agent to imitate nearly arbitrary behavior with provable guarantees, even when the dynamics governing the agent and environments interaction are nonlinear. We conclude by outlining ongoing work and future steps towards building out the mathematical and conceptual tooling for understanding the next steps towards general, capable and flexible physical agents. Bio: Max Simchowitz is a postdoctoral researcher in the Robot Locomotion Group at MIT CSAIL. He studies the theoretical foundations of machine learning problems with a sequential or dynamical component; he currently focuses on robotics and out-of-distribution learning, and with past work ranging broadly across control, reinforcement learning, optimization and algorithmic fairness. He received his PhD from University of California, Berkeley in 2021 under Ben Recht and Michael I. Jordan, and his work has been recognized with an ICML 2018 Best Paper Award, ICML 2022 Outstanding Paper Award, and RSS 2023 Best Paper Finalist designation.

1 0

Princeton Robotics Seminar: Marin Kobilarov, Friday Mar 8 at 11am
by Emily C. Lawrence 05 Mar '24

05 Mar '24

Speaker: [ https://asco.lcsr.jhu.edu/ | Marin Kobilarov ] , JHU Date: Friday, March 8th, 2024 Time: 11am ET Location: Computer Science 105 Event page: [ https://robo.princeton.edu/seminar | https://robo.princeton.edu/seminar ] Title: High-confidence Robot Motion Planning under Uncertainty Abstract: This talk will provide an overview of research activities at the ASCO lab, currently including robot-assisted surgical micro-manipulation and navigation of autonomous vehicles (aerial, underwater, or ground) under state and perception uncertainty. We will specifically focus on motion planning with built-in robustness guarantees, i.e. by aiming to certify expected performance before actual deployment. The core idea is to employ probably-approximately-correct (PAC) bounds on performance which are used as an objective function in control policy optimization. Such robust policies could then provide high-confidence performance guarantees, such as “with 99% chance the robot will reach its goal, while avoiding collisions with 99.9% chance”, and result in improved safety and reliability. Bio: Marin Kobilarov is an Associate Professor at the Johns Hopkins University and a Principal Engineer at Zoox/Amazon. At JHU he leads the Autonomous Systems, Control and Optimization (ASCO) lab which develops algorithms and software for planning, learning, and control of autonomous robotic systems. Their focus is on computational theory at the intersection of planning and learning, and on the system integration and deployment of robots that can operate safely and efficiently in challenging environments. PhD students & faculty: If you are interested in meeting with Marin on 3/8, please reach out to [ mailto:nsimon@princeton.edu | nsimon(a)princeton.edu ] . PhD students: If you are interested in joining the speaker for lunch on 3/8 from 12:15 - 1:30 PM, please [ https://forms.gle/r6F5hMMUdEHawor87 | fill out this form ] .

1 0

Ksenia Sokolova will present her FPO "Deep Learning for Sequence-Based Gene Expression Prediction" on Monday, March 18, 2024 at 10:30 AM in the 252 Nassau Street Conference room.
by gradinfo＠cs.princeton.edu 04 Mar '24

04 Mar '24

Ksenia Sokolova will present her FPO "Deep Learning for Sequence-Based Gene Expression Prediction" on Monday, March 18, 2024 at 10:30 AM in the 252 Nassau Street Conference room. Location: 252 Nassau Street Conference room The members of Ksenia's committee are as follows: Examiners: Olga Troyanskaya (Adviser), Kai Li, Ellen Zhong Readers: Mona Singh, Yuri Pritykin Everyone is invited to attend her talk. Abstract follows below: Human biology is rooted in highly specialized cell types programmed by a common genome, 98% of which is outside of genes. While genetic variation in the enormous noncoding space is linked to the majority of disease risk, the impact of this variation is poorly understood. The recent advances in sequencing technology made it possible to perform whole genome sequencing of the large cohorts, uncovering many variants per individual. A crucial challenge is to understand the collective impact of these variants on gene expression across varied human cell types and their subsequent roles in disease progression. This dissertation begins by tackling the challenge of associating noncoding genetic variants with changes in gene expression in primary human cell types. We introduce ExPectoSC, an atlas of modular deep-learning-based models for predicting cell-type-specific gene expression directly from sequence. With models spanning 105 primary human cell types across seven organ systems, it offers a detailed insight into the effect of variation. The resulting atlas of sequence-based gene expression and variant effects is publicly available in a user-friendly interface and readily extensible to any primary cell types. We follow this work with an example application of the ExpectoSC to the study of glomerular diseases, a major cause of end stage renal disease in the US. Despite having similar clinical presentations, these diseases are known for their heterogeneity and variable patient outcomes. By integrating whole-genome sequencing data with ExPectoSC's predictions, we construct comprehensive gene expression disruption profiles for patients. 4 Finally, we developed a new method for genomic-centered contrastive pre-training, called cGen, to improve training of the models from sequence alone in limited-data contexts. Utilizing sequence augmentations, after pre-training cGen generates unsupervised embeddings that highlight functional clusters and are informative of gene expression in the absence of any labeled information. Together, these contributions highlight the power of computational approaches to decode the noncoding genome, offering new avenues for the diagnosis, prognosis, and treatment of human diseases.

1 0

DeCenter Seminar: Giulia Fanti, March 6 at 12:30 pm
by Emily C. Lawrence 29 Feb '24

29 Feb '24

DeCenter Seminar: Accountability in Central Bank Digital Currency March 6, 2024 12:30 pm Friend Center, Convocation Room Speaker: Giulia Fanti, Carnegie Mellon University Abstract: Today, over 110 central banks worldwide are actively exploring the deployment of a centrally-banked digital currency (CBDC). A (retail) CBDC is central bank money that is (a) digital in nature, and (b) widely accessible to the public. Some of the proposed benefits of CBDCs include greater flexibility in implementing monetary policy, improved financial inclusion, and more efficient money transfers at lower fees. However, in parallel with the momentum surrounding CBDCs is a growing fear that CBDCs will significantly increase the attack surface for the financial system at large. In this talk, we will explore how to provide accountability for CBDCs. In other words, instead of preventing malicious activity outright, we want to design protocols that are rich enough to provably attribute malicious behavior to the faulty party. An essential question is how to provide accountability without compromising the efficiency of the system. We will discuss challenges and proposed solutions to the accountability problem at the consensus layer, the payment verification layer, and at the interface with end users. Notably, each of these problems is challenging in part because of partial decentralization that arises naturally in certain CBDC designs, despite their ostensibly centralized nature. Bio : Giulia Fanti is an Assistant Professor of Electrical and Computer Engineering at Carnegie Mellon University. Her research interests span the security, privacy, and efficiency of distributed systems. She is a two-time fellow of the World Economic Forum’s Global Future Council on Cybersecurity and a member of NIST’s Information Security and Privacy Advisory Board. Her work has been recognized with several awards, including best paper awards, a Sloan Fellowship, an Intel Rising Star Faculty Award, and an ACM SIGMETRICS Rising Star Award. She obtained her Ph.D. in EECS from U.C. Berkeley and her B.S. in ECE from Olin College of Engineering.

1 0

PPH upcoming events: March 8 seminar kickoff and April 5 Tech & Mental Health workshop
by Emily C. Lawrence 27 Feb '24

27 Feb '24

Hello, We are excited to announce some upcoming Princeton Precision Health events. We are kicking off our weekly seminar series on March 8, when Dan Notterman and Noreen Goldman will present a seminar entitled: Cardiovascular Health among Young Adults: a Demographic, Molecular, and Clinical Study. Our weekly series will be in our new space at 252 Nassau Street (between Tiger Noodles and the Small World outpost on Nassau). Lunch will be provided. See attached flyer for details. Check out our web site for upcoming seminars ( [ https://pph.princeton.edu/events | https://pph.princeton.edu/events ] ). We are also hosting a Tech and Mental Health Workshop on April 5, from 9:00 am - 2:00 pm at Lewis Library 120. Ine Beyens, from the University of Amsterdam, will give a keynote titled, “Beyond one-size-fits-all: the person-specific impact of social media on adolescent mental health." For more information, check out our Program page ( [ https://pph.princeton.edu/tech-and-mental-health-symposium | https://pph.princeton.edu/tech-and-mental-health-symposium ] ) and the attached flyer. You can register here ( [ https://pph.princeton.edu/form/2024-april-symposium-registratio | https://pph.princeton.edu/form/2024-april-symposium-registratio ] ). The PPH Team ************************************ Princeton Precision Health [ http://pph.princeton.edu/ | pph.princeton.edu ] [ mailto:princetonpph@princeton.edu | princetonpph(a)princeton.edu ]

1 0

CSML/CS/PSY Seminar: Brenden Lake, TODAY at 12:30pm
by Emily C. Lawrence 27 Feb '24

27 Feb '24

Departments of Psychology, Computer Science, and CSML Speaker: Brenden Lake, New York University Title: Towards more human-like learning in machines: Bridging the data and generalization gaps Day: Tuesday, February 27 Place: Computer Science Building Room 105 Time: 12:30-1:30 PM Lunch available from12:15 with RSVP to csml(a)princeton.edu Host: Tom Griffiths, Henry R. Luce Professor of Information Technology, Consciousness and Culture Event page: https://csml.princeton.edu/events/towards-more-human-learning-machines-brid… Webinar Registration: [ https://princeton.zoom.us/webinar/register/WN_aUfLfVQFSIqV2RDSYu4big | https://princeton.zoom.us/webinar/register/WN_aUfLfVQFSIqV2RDSYu4big ] Abstract: There is an enormous data gap between how AI systems and children learn: The best LLMs now learn language from text with a word count in the trillions, whereas it would take a child roughly 100K years to reach those numbers through speech (Frank, 2023, "Bridging the data gap"). There is also a clear generalization gap: whereas machines struggle with systematic generalization, children can excel. For instance, once a child learns how to "skip," they immediately know how to "skip twice" or "skip around the room with their hands up" due to their compositional skills. In this talk, I'll describe two case studies in addressing these gaps. 1) The data gap: We train deep neural networks from scratch (using DINO, CLIP, etc.), not on large-scale data from the web, but through the eyes and ears of a single child. Using head-mounted video recordings from a child as training data (<200 hours of video slices over 26 months), we show how deep neural networks can perform challenging visual tasks, acquire many word-referent mappings, generalize to novel visual referents, and achieve multi-modal alignment. Our results demonstrate how today's AI models are capable of learning key aspects of children's early knowledge from realistic input. 2) The generalization gap: Can neural networks capture human-like systematic generalization? We address a 35-year-old debate catalyzed by Fodor and Pylyshyn's classic article, which argued that standard neural networks are not viable models of the mind because they lack systematic compositionality -- the algebraic ability to understand and produce novel combinations from known components. We'll show how neural networks can achieve human-like systematic generalization when trained through meta-learning for compositionality (MLC), a new method for optimizing the compositional skills of neural networks through practice. With MLC, neural networks can match human performance and solve several machine learning benchmarks. Given these findings, we'll discuss the paths forward for building machines that learn, generalize, and interact in more human-like ways based on more natural input. For related articles visit: https://csml.princeton.edu/events/towards-more-human-learning-machines-brid… BIO: Brenden M. Lake is an Assistant Professor of Data Science and Psychology at New York University. He received his M.S. and B.S. in Symbolic Systems from Stanford University in 2009, and his Ph.D. in Cognitive Science from MIT in 2014. He was a postdoctoral Data Science Fellow at NYU from 2014-2017. Brenden is a recipient of the Robert J. Glushko Prize for Outstanding Doctoral Dissertation in Cognitive Science and the MIT Technology Review 35 Innovators Under 35. His research was also selected by Scientific American as one of the 10 most important advances of 2016. Brenden's research focuses on computational problems that are easier for people than they are for machines, such as learning new concepts from just a few examples, learning by asking questions, learning by generating new goals, and learning by producing novel combinations of known components.

1 0

Howard Yen will present his MSE talk "Long-Context Language Modeling with Parallel Context Encoding" at 1pm on March 7th, 2024 in the 1st floor conference room of CSML (26 Prospect Ave)
by Gradinfo 27 Feb '24

27 Feb '24

Howard Yen will present his MSE talk "Long-Context Language Modeling with Parallel Context Encoding" at 1pm on March 7th, 2024 in the 1st floor conference room of CSML (26 Prospect Ave). Advisor: Prof. Danqi Chen; Reader: Karthik Narasimhan Abstract: Extending large language models (LLMs) to process longer inputs is crucial for numerous applications. However, the considerable computational cost of transformers, coupled with limited generalization of positional encoding, restricts the size of their context window. We introduce Context Expansion with Parallel Encoding (CEPE), a framework that can be applied to any existing decoder-only LLMs to extend their context window. CEPE adopts a small encoder to process long inputs chunk by chunk and enables the frozen decoder to leverage additional contexts via cross-attention. CEPE is efficient, generalizable, and versatile: trained with 8K-token documents, CEPE extends the context window of LLAMA-2 to 128K tokens, offering 10x the throughput with only 1/6 of the memory. CEPE yields strong performance on language modeling and in-context learning. CEPE also excels in retrieval-augmented applications, while existing long-context models degenerate with retrieved contexts. We further introduce a CEPE variant that can extend the context window of instruction-tuned models with only unlabeled data, and showcase its effectiveness on LLAMA-2-CHAT, leading to a strong instruction-following model that can leverage very long context on downstream tasks. CS Grad Calendar: [ https://calendar.google.com/calendar/event?action=TEMPLATE&tmeid=MWhxOGEybz… | https://calendar.google.com/calendar/event?action=TEMPLATE&tmeid=MWhxOGEybz… ]

1 0

Princeton Robotics Seminar: Caitlin Mueller, Friday Feb 23
by Emily C. Lawrence 20 Feb '24

20 Feb '24

Speaker: [ http://digitalstructures.mit.edu/ | Caitlin Mueller ] , MIT Date: Friday, February 23, 2024 Time: 11am ET Location: Computer Science 105 Event page: [ https://robo.princeton.edu/seminar | https://robo.princeton.edu/seminar ] Title: Towards Robotic Construction of Sustainable Structures Abstract: Within the climate crisis, architecture and the built environment play an outsized role, both in contributions to current emissions and projections for the future. Standard design and construction practices are often wasteful, expensive, and exploitative, undermining a mission to create spaces of dignity, comfort, and delight. Within this context, this talk will present an alternative approach, focused on promoting design and construction methods that produce high-quality, low-carbon, and inexpensive outcomes, empowered by emerging advances in computational design and construction robotics. In addition to general methods, the talk will present a range of recent projects, including robotic planning and assembly of complex yet highly efficient truss structures, robotics-enabled fabrication of low-cost, low-carbon earthen and concrete structures, and algorithmic design and fabrication approaches for unconventional and circular material use. Bio: Caitlin Mueller is an Associate Professor at MIT's Department of Architecture and Department of Civil and Environmental Engineering, in the Building Technology Program, where she leads the Digital Structures research group. She works at the creative interface of architecture, structural engineering, and computation, and focuses on new computational design and digital fabrication methods for innovative, high-performance buildings and structures that empower a more sustainable and equitable future. Mueller holds three degrees from MIT in Architecture, Computation, and Building Technology, and one from Stanford in Structural Engineering. Her research is funded by federal agencies and industry partners, including the National Science Foundation, FEMA, the MIT Tata Center, the Dar Group, Holcim, Robert McNeel & Associates, and Altair Engineering. Mueller has won best paper awards from the International Association of Shell and Spatial Structures, the Symposium on Geometry Processing, and the Journal of Mechanical Design, and was awarded the ACADIA Innovative Research Award of Excellence by the Association for Computer Aided Design in Architecture in 2021 and the Diversity Achievement Award from the Association of Collegiate Schools of Architecture in 2022.

1 0