About the Study

Title: Real-Time Neural Voice Camouflage.

Authors: Mia Chiquier, Chengzhi Mao, and Carl Vondrick; Department of Computer Science; Columbia Engineering

This research is based on work supported by the NSF CRII Award 1850069 and the NSF CAREER Award 2046910. MC is supported by a CAIT PhD fellowship. The views and conclusions should not be interpreted as necessarily representing the official policies, either expressed or implied, of the sponsors.

COI: The authors declare no financial or other conflicts of interest.

Learning the Earth with Artificial Intelligence and Physics (LEAP) center launched in 2021. What has been the core mission of the center?

LEAP’s mission is to increase the reliability, utility, and reach of climate projections by integrating climate and data science. Our primary strategy is to improve near-term climate projections by merging physical modeling with machine learning across a continuum, from expertise in climate science and climate modeling to cutting-edge machine learning algorithms. This will really help both the climate and data sciences communities--climate scientists and modelers struggle to fully integrate the plethora of existing datasets into their models, while machine learning algorithms have been good at emulating, such as ChatGPT, they are not so good at extrapolating or predicting extremes. By combining both approaches, we hope that LEAP will trigger a significant advancement in data science algorithms applied to physical problems. The center is incorporating physics and causal mechanisms into machine learning algorithms for better generalization and extrapolation, while optimally using the wealth of data available to climate science, in order to better predict the future. 

Can you touch on some of the key projects LEAP has been focusing on? Any breakthroughs?

LEAP is working on different aspects of climate science and data science, covering not only research but also knowledge transfer and education. Some of the recent research breakthroughs in our center have shown that by using AI, we can discover new previously unknown physics (of clouds or ocean turbulence). We hope to use this “new” physics in climate models to improve their accuracy, especially of extremes. Essentially we’re using vast amounts of data, such as those from satellites, so that we can refine climate models and their evaluation and improve our predictions. It’s also critical that we share this information with the public and private sectors in a user-friendly way.

We also hope to break some of historic silos with climate research as climate research doesn't translate easily to the public or to the private sectors, as it is very technical and difficult to use. We are creating a cloud platform where we can provide climate data more widely, and also engage with our colleagues in the field and beyond to see what is actually useful to them. For instance, a business will want to know how much the frequency of flooding or heat waves will change in the future so they could adapt their business. With LEAP, we expect to be able to refine our models so that we can help offer more precise predictions.

There has been a lot of attention given to AI, and AI is currently being applied in your climate research and others. How has AI revolutionized climate modeling, and what’s the state of that today?

Over the last five years, there has been an explosion in the use of AI to better understand climate models and to better represent physical processes (such as clouds, ocean, and terrestrial carbon cycle or ocean turbulence). The next big push is how to integrate those AI algorithms within climate models that have historically used empirical equations.

The information we are predicting in the future is really uncertain. And there are many reasons for that, including the sheer complexity of all of those processes that we are trained to do when we build climate models. So we are focused on reducing and narrowing those uncertainties to give everyone, from policy makers to business leaders to educators, accurate climate projections to inform their own decision-making. For instance in the agricultural sector, being able to provide precise information on future climate could heavily impact crop productivity and yield. 

The goal is to improve climate modeling so that we can say how many days a drought may last or what the likelihood of flooding is in New York City or any other specific low-lying area. These are questions that are really critical. Right now, the range of estimates is just so dramatic that it’s challenging to actually implement plans. We need to act now on this problem, and really then begin to fight climate change.

What excites you the most about where the field is headed and how LEAP's work will impact our future?

We are witnessing a true transformation and it’s centered around data – and the use of observation and simulation data to answer new hypotheses or questions that could not be addressed until now. Of course, as for any new field, we still need to be cautious and ensure that the results are sound and reproducible, but I am quite excited to see where the field is heading and witness the incredible pace of advances. I think that we are really witnessing a revolution in the climate sciences. 

On this Earth Day 2023, what would you like to tell your children about how we are investing in our planet, how important climate change research is, and what you hope to leave to them and future generations?

What I tell my three kids is to think about others. We need to use fewer resources, to limit our footprint, to recycle, and to emit less. In other words, we need to mitigate climate change. But we also need to adapt. Climate change is now part of our everyday lives, as evidenced by the explosion in the number–and intensity–of recent extreme events (droughts, floods, etc). It’s clear we need to change as a society. This change has to be embraced by everyone and thus we need widespread support (government, policies) so that the entire country can adapt to those changes, not just a small fraction of our society. Otherwise, we will have failed as a society and country. Social justice and climate change are tightly connected–we need both to achieve our goals

About the Study

Journal: Optica

Title: Picosecond-resolution single-photon time lens for temporal mode quantum processing

Authors: Chaitali Joshi 1,2, Ben M. Sparkes 3, Alessandro Farsi 1, Thomas Gerrits 4, Varun Verma 5, Sven Ramelow 6, Sae Woo Nam 5, and Alexander Gaeta 1

1. Applied Physics and Applied Mathematics, Columbia University, New York, New York 10027, USA
2. Applied and Engineering Physics, Cornell University, Ithaca, New York 14850, USA
3. Institute for Photonics and Advanced Sensing (IPAS) and School of Physical Sciences, University of Adelaide, Adelaide, SA, Australia
4. National Institute of Standards and Technology, Gaithersburg, Maryland 20899, USA
5. National Institute of Standards and Technology, Boulder, Colorado 80305, USA
6. Institut für Physik, Humboldt-Universität zu Berlin, Berlin, Germany

This work was supported by the National Science Foundation (PHY-1707918, OMA-1936345) and the Australian Research Council (DE170100752)

B.M.S. acknowledges support from the Fulbright Future Scholarship (funded by the Kinghorn Foundation). Certain commercial equipment and instruments are identified in this paper for completeness. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

About the Study

Conference: IEEE Symposium on Security and Privacy, presented May 22-25, 2022, San Francisco.

Title: Differential Privacy and Swapping: Examining De-Identication’s Impact on Minority Representation and Privacy Preservation in the U.S. Census.

Authors: Miranda Christ (Columbia University), Sarah Radway (Tufts University- formerly Columbia University), Steven M. Bellovin (Columbia University)

This research was supported in part by the U.S. Department of Energy (DOE), Office of Science, Office of Advanced Scientific Computing Research under award number DE-SC-0001234, a Google Faculty Grant, NSF grants CNS-1923528 and CCF-2107187, a grant from the Columbia-IBM center for Blockchain and Data Transparency, a grant from JPMorgan Chase & Co., and a grant from LexisNexis Risk Solutions.

COI: The authors declare no financial or other conflicts of interest.

Self-awareness in robots

The work is part of Lipson’s decades-long quest to find ways to grant robots some form of self-awareness. “Self-modeling is a primitive form of self-awareness,” he explained. “If a robot, animal, or human has an accurate self-model, it can function better in the world, it can make better decisions, and it has an evolutionary advantage.”

The researchers are aware of the limits, risks, and controversies surrounding granting machines greater autonomy through self-awareness. Lipson is quick to admit that the kind of self-awareness demonstrated in this study is, as he noted, “trivial compared to that of humans, but you have to start somewhere. We have to go slowly and carefully, so we can reap the benefits while minimizing the risks.”

About the Study

Title: Full-Body Visual Self-Modeling of Robot Morphologies

Authors: Boyuan Chen, Robert Kwiatkowski, Carl Vondrick, Hod Lipson

The study was supported by DARPA MTO Lifelong Learning Machines (L2M) Program W911NF-21-2-0071, NSF NRI Award 1925157, NSF AI Institute for Dynamical Systems 2112085, NSF CAREER Award 2046910, and gifts from Facebook Research and Northrop Grumman.

COI: The authors declare no financial or other conflicts of interest.