As LLMs become commonplace, machine-generated text has the potential to flood the internet with spam, social media bots, and valueless content. Watermarking is a simple and effective strategy for mitigating such harms by enabling the detection and documentation of LLM-generated text. Yet a crucial question remains: How reliable is watermarking in realistic settings in the wild? There, watermarked text may be modified to suit a user's needs, or entirely rewritten to avoid detection. We study the robustness of watermarked text after it is re-written by humans, paraphrased by a non-watermarked LLM, or mixed into a longer hand-written document. We find that watermarks remain detectable even after human and machine paraphrasing. While these attacks dilute the strength of the watermark, paraphrases are statistically likely to leak n-grams or even longer fragments of the original text, resulting in high-confidence detections when enough tokens are observed. For example, after strong human paraphrasing the watermark is detectable after observing 800 tokens on average, when setting a 1e 5 false positive rate. We also consider a range of new detection schemes that are sensitive to short spans of watermarked text embedded inside a large document, and we compare the robustness of watermarking to other kinds of detectors.

Motion capture through tracking retroreflectors obtains highly accurate pose estimation, which is frequently used in robotics. Unlike commercial motion capture systems, fiducial marker-based tracking methods, such as AprilTags, can perform relative localization without requiring a static camera setup. However, popular pose estimation methods based on fiducial markers have lower localization accuracy than commercial motion capture systems. We propose Mobile MoCap, a system that utilizes inexpensive near-infrared cameras for accurate relative localization even while in motion. We present a retroreflector feature detector that performs 6-DoF (six degrees-of-freedom) tracking and operates with minimal camera exposure times to reduce motion blur. To evaluate the proposed localization technique while in motion, we mount our Mobile MoCap system, as well as an RGB camera to benchmark against fiducial markers, onto a precision-controlled linear rail and servo. The fiducial marker approach employs AprilTags, which are pervasively used for localization in robotics. We evaluate the two systems at varying distances, marker viewing angles, and relative velocities. Across all experimental conditions, our stereo-based Mobile MoCap system obtains higher position and orientation accuracy than the fiducial approach. The code for Mobile MoCap is implemented in ROS 2 and made publicly available at https://github.com/RIVeR-Lab/mobile_mocap.

Federated clustering (FC) is an unsupervised learning problem that arises in a number of practical applications, including personalized recommender and healthcare systems. With the adoption of recent laws ensuring the "right to be forgotten", the problem of machine unlearning for FC methods has become of significant importance. We introduce, for the first time, the problem of machine unlearning for FC, and propose an efficient unlearning mechanism for a customized secure FC framework. Our FC framework utilizes special initialization procedures that we show are well-suited for unlearning. To protect client data privacy, we develop the secure compressed multiset aggregation (SCMA) framework that addresses sparse secure federated learning (FL) problems encountered during clustering as well as more general problems. To simultaneously facilitate low communication complexity and secret sharing protocols, we integrate Reed-Solomon encoding with special evaluation points into our SCMA pipeline, and prove that the client communication cost is logarithmic in the vector dimension. Additionally, to demonstrate the benefits of our unlearning mechanism over complete retraining, we provide a theoretical analysis for the unlearning performance of our approach. Simulation results show that the new FC framework exhibits superior clustering performance compared to previously reported FC baselines when the cluster sizes are highly imbalanced. Compared to completely retraining K-means++ locally and globally for each removal request, our unlearning procedure offers an average speed-up of roughly 84x across seven datasets. Our implementation for the proposed method is available at https://github.com/thupchnsky/mufc. The availability of large volumes of user training data has contributed to the success of modern machine learning models. For example, most state-of-the-art computer vision models are trained on large-scale image datasets including Flickr (Thomee et al., 2016) and ImageNet (Deng et al., 2009).

Structure-based drug design (SBDD) aims to design small-molecule ligands that bind with high affinity and specificity to pre-determined protein targets. In this paper, we formulate SBDD as a 3D-conditional generation problem and present DiffSBDD, an SE(3)-equivariant 3D-conditional diffusion model that generates novel ligands conditioned on protein pockets. Comprehensive in silico experiments demonstrate the efficiency and effectiveness of DiffSBDD in generating novel and diverse drug-like ligands with competitive docking scores. We further explore the flexibility of the diffusion framework for a broader range of tasks in drug design campaigns, such as off-the-shelf property optimization and partial molecular design with inpainting.

We introduce MABe22, a large-scale, multi-agent video and trajectory benchmark to assess the quality of learned behavior representations. This dataset is collected from a variety of biology experiments, and includes triplets of interacting mice (4.7 million frames video+pose tracking data, 10 million frames pose only), symbiotic beetle-ant interactions (10 million frames video data), and groups of interacting flies (4.4 million frames of pose tracking data). Accompanying these data, we introduce a panel of real-life downstream analysis tasks to assess the quality of learned representations by evaluating how well they preserve information about the experimental conditions (e.g. strain, time of day, optogenetic stimulation) and animal behavior. We test multiple state-of-the-art self-supervised video and trajectory representation learning methods to demonstrate the use of our benchmark, revealing that methods developed using human action datasets do not fully translate to animal datasets. We hope that our benchmark and dataset encourage a broader exploration of behavior representation learning methods across species and settings.

As California faces its first major heat wave of the summer this Fourth of July weekend, state officials are urging residents to not become complacent about the threat of wildfires this year. Standing beneath the blistering sun at the Grass Valley Air Attack Base in Nevada County, California Department of Forestry and Fire Protection chief Joe Tyler outlined the state's plans to battle blazes this year with new tools and technology, as well as increased vegetation management efforts. He cautioned that while the wet start to 2023 may have delayed the start of fire season, it has not deterred it. "The abundant rain has produced tall grass and other vegetation that's dried out already and is ready to burn," Tyler said. Additionally, portions of the state are expected to soar into the triple digits this weekend, including up to 110 degrees in the Sacramento Valley.

Tech Policy Center director Kara Fredrick explains how individuals and companies can mitigate the spread of misinformation by A.I. on'The Faulkner Focus.' British election regulators have urged politicians to pass new laws to limit spending on artificial intelligence (AI) as well as new requirements to identify AI-generated content. "The next U.K. general election is a ripe target for electronic disinformation given we are in the infancy of the AI age," Alan Mendoza, co-founder and executive director of the Henry Jackson Society, told Fox News Digital. "Many of the possible problems that may emerge have not even been considered." "As a result, we face a race against the clock to introduce appropriate protections, or run the nightmare risk of bad actors influencing campaigns and destroying public trust in our democratic process," he added.

The news: Disinformation generated by AI may be more convincing than disinformation written by humans, according to a new study. It found that people were 3% less likely to spot false tweets that had been generated by AI than real-life examples collected from Twitter. But the way in which GPT-3 orders information could have something to do with it, as AI-generated text tends to be more structured and condensed in comparison to how humans write. Why it matters: AI models can generate incorrect text that appears convincing, which could be used to generate false narratives quickly and cheaply for conspiracy theorists and disinformation campaigns. In theory, this could be spread further and faster than online disinformation networks manned by humans.

This month we heard about a fascinating AI experiment from a US Air Force colonel. An AI-controlled drone trained to autonomously carry out bombing missions had turned on its human operator when told not to attack targets; its programming prioritised successfully carrying out missions, so it saw human intervention as an obstacle in its way and decided to forcefully take it out. The only problem with the story was that it was nonsense. Firstly, as the colonel told it, the test was a simulation. Secondly, a US Air Force statement was hastily issued to clarify that the colonel, speaking at a UK conference, had "mis-spoke" and that no such tests had been carried out.

A first-of-its-kind law in New York City aims to make the use of AI in hiring and promotion both clearer and fairer. New York's Local Law 144, which goes into effect on 5 July, requires employers to get an independent audit of their automated employment decision tools to ensure that they do not demonstrate significant bias based on sex, race or ethnicity – though it does not cover discrimination based on factors such as …