This paper presents a novel approach to address the challenging problem of autonomous on-ramp merging, where a self-driving vehicle needs to seamlessly integrate into a flow of vehicles on a multi-lane highway. We introduce the Lane-keeping, Lane-changing with Latent-state Inference and Safety Controller (L3IS) agent, designed to perform the on-ramp merging task safely without comprehensive knowledge about surrounding vehicles' intents or driving styles. We also present an augmentation of this agent called AL3IS that accounts for observation delays, allowing the agent to make more robust decisions in real-world environments with vehicle-to-vehicle (V2V) communication delays. By modeling the unobservable aspects of the environment through latent states, such as other drivers' intents, our approach enhances the agent's ability to adapt to dynamic traffic conditions, optimize merging maneuvers, and ensure safe interactions with other vehicles. We demonstrate the effectiveness of our method through extensive simulations generated from real traffic data and compare its performance with existing approaches. L3IS shows a 99.90% success rate in a challenging on-ramp merging case generated from the real US Highway 101 data. We further perform a sensitivity analysis on AL3IS to evaluate its robustness against varying observation delays, which demonstrates an acceptable performance of 93.84% success rate in 1-second V2V communication delay.

We study the question: How can we select the right data for fine-tuning to a specific task? We call this data selection problem active fine-tuning and show that it is an instance of transductive active learning, a novel generalization of classical active learning. We propose ITL, short for information-based transductive learning, an approach which samples adaptively to maximize information gained about the specified task. We are the first to show, under general regularity assumptions, that such decision rules converge uniformly to the smallest possible uncertainty obtainable from the accessible data. We apply ITL to the few-shot fine-tuning of large neural networks and show that fine-tuning with ITL learns the task with significantly fewer examples than the state-of-the-art.

Generative, multimodal artificial intelligence (GenAI) offers transformative potential across industries, but its misuse poses significant risks. Prior research has shed light on the potential of advanced AI systems to be exploited for malicious purposes. However, we still lack a concrete understanding of how GenAI models are specifically exploited or abused in practice, including the tactics employed to inflict harm. In this paper, we present a taxonomy of GenAI misuse tactics, informed by existing academic literature and a qualitative analysis of approximately 200 observed incidents of misuse reported between January 2023 and March 2024. Through this analysis, we illuminate key and novel patterns in misuse during this time period, including potential motivations, strategies, and how attackers leverage and abuse system capabilities across modalities (e.g. image, text, audio, video) in the wild.

As artificial intelligence (AI) systems surpass human capabilities in various tasks, ensuring alignment with human values and ethics is crucial. This is especially important for large language models (LLMs), which are trained on diverse datasets that may contain harmful content. Reinforcement Learning from Human Feedback (RLHF) is an effective method for AI alignment, with models like OpenAI's GPT-4, Google's Gemini, and Anthropic Claude showing safe and aligned behaviors. However, the vast majority of the current research in RLHF (Agarwal et al., 2020; Rafailov et al., 2023; Ouyang et al., 2022; Chakraborty et al., 2024; Swamy et al., 2024) focuses on the offline setting, which involves using a fixed dataset of responses generated by the supervised fine-tuned model (SFT), ranked by human experts. Consequently, these methods are inherently offline and heavily reliant on the quality of the offline data generated by the SFT model, which exhibits drawbacks such as insufficient coverage of response-query pairs leading to sub-optimal alignment. To deal with the above shortcomings, recent literature (Guo et al., 2024a; Sharma et al., 2024; Lee et al., 2023; Yuan et al., 2024b) has focused on designing online RLHF algorithms. The setting of online RLHF transcends the constraints of a static offline dataset and aims to address two critical questions: Q1: How should we generate new responses during fine-tuning?

Graph Neural Networks (GNNs) are widely applied to graph learning problems such as node classification. When scaling up the underlying graphs of GNNs to a larger size, we are forced to either train on the complete graph and keep the full graph adjacency and node embeddings in memory (which is often infeasible) or mini-batch sample the graph (which results in exponentially growing computational complexities with respect to the number of GNN layers). Various sampling-based and historical-embedding-based methods are proposed to avoid this exponential growth of complexities. However, none of these solutions eliminates the linear dependence on graph size. This paper proposes a sketch-based algorithm whose training time and memory grow sublinearly with respect to graph size by training GNNs atop a few compact sketches of graph adjacency and node embeddings. Based on polynomial tensor-sketch (PTS) theory, our framework provides a novel protocol for sketching non-linear activations and graph convolution matrices in GNNs, as opposed to existing methods that sketch linear weights or gradients in neural networks. In addition, we develop a locality-sensitive hashing (LSH) technique that can be trained to improve the quality of sketches. Experiments on large-graph benchmarks demonstrate the scalability and competitive performance of our Sketch-GNNs versus their full-size GNN counterparts.

If you tilted your ears in a certain direction on Monday, you could make out a resounding cheer from the creative class across social media platforms and various Discord servers. That's because the Federal Trade Commission sued software company Adobe and two of its executives for "deceiving consumers" by all but forcing them "into year-long subscriptions through hidden early termination fees and numerous cancellation hurdles." "Adobe has had it coming for years," New York journalist Nolan Hicks stated. "I don't know of a single person who is rooting for Adobe on this," tweeted video essayist Scott Niswander. One viral meme urged the agency to "tear the bitch apart."

To boldly go where no man has gone before. That's the mission of the USS Enterprise -- and arguably the aim of a 1-million prize being offered through a foundation created to honor the father of the "Star Trek" franchise. The Roddenberry Foundation -- named for Gene Roddenberry -- announced Tuesday that this year's biennial award would focus on artificial intelligence that benefits humanity. Lior Ipp, chief executive of the foundation, told The Times there's a growing recognition that AI is becoming more ubiquitous and will affect all aspects of our lives. "We are trying to … catalyze folks to think about what AI looks like if it's used for good," Ipp said, "and what it means to use AI responsibly, ethically and toward solving some of the thorny global challenges that exist in the world."

The Biden campaign is facing its first major cheapfake scandal this week. Doctored clips of Biden at the G7 Summit and a Hollywood fundraiser have spread across platforms like X, claiming to show Biden wandering off, mumbling unintelligibly, or, uh, even pooping his pants. It's exactly the type of content the right-wing media apparatus drools over to play up Biden's age, despite the clips being edited in a manner reminiscent of the drunk Nancy Pelosi video from last cycle. And while we're all starting to get stressed over simple editing and cropping techniques again, Big Tech is training political campaigns on their generative AI tools. Could a little direction help mitigate the issue?

The AI Fringe returned for a second year on 5 June. The event was designed to complement the AI Seoul Summit which was co-hosted by the UK and South Korea governments. The goals of the AI Fringe are 1) to bring together the views of industry, civil society and academia on safe and beneficial AI, 2) to provide a platform for all communities to engage in the discussion, 3) to enhance understanding of AI and its impacts so organisations can harness its benefits. This year, the Fringe comprised a half-day event with the key elements being two panel discussions. The first addressed AI safety, and the panel reflected on progress over the last 12 months.

There has been an avalanche of coverage recently from liberal news outlets on so-called "cheap fakes," the term being used by both the media and the White House to describe viral clips of President Biden that critics say show signs of his cognitive decline. Biden's age has been the subject of intense scrutiny in recent days, with the president facing accusations of freezing and wandering off at various events showcased online by Republicans. One prominent example was footage showing Biden stepping away from other world leaders at the G-7 Summit to give a thumbs up to parachutists off-camera, prompting Italian Prime Minister Giorgia Meloni to corral him back to the group for a photo-op. "Selective editing of video and putting spin on interpretations of events has been going on in American politics for a long time," DePauw University journalism professor Jeffrey McCall said. "What has been surprising, however, is how eager the establishment media have been to parrot the White House spin, trying to dismiss concerns about Biden's capabilities as just cheap fake editing and razzle-dazzle."