Current approaches to building general-purpose AI systems tend to produce systems with both beneficial and harmful capabilities. Further progress in AI development could lead to capabilities that pose extreme risks, such as offensive cyber capabilities or strong manipulation skills. We explain why model evaluation is critical for addressing extreme risks. Developers must be able to identify dangerous capabilities (through "dangerous capability evaluations") and the propensity of models to apply their capabilities for harm (through "alignment evaluations"). These evaluations will become critical for keeping policymakers and other stakeholders informed, and for making responsible decisions about model training, deployment, and security.

A challenging category of robotics problems arises when sensing incurs substantial costs. This paper examines settings in which a robot wishes to limit its observations of state, for instance, motivated by specific considerations of energy management, stealth, or implicit coordination. We formulate the problem of planning under uncertainty when the robot's observations are intermittent but their timing is known via a pre-declared schedule. After having established the appropriate notion of an optimal policy for such settings, we tackle the problem of joint optimization of the cumulative execution cost and the number of state observations, both in expectation under discounts. To approach this multi-objective optimization problem, we introduce an algorithm that can identify the Pareto front for a class of schedules that are advantageous in the discounted setting. The algorithm proceeds in an accumulative fashion, prepending additions to a working set of schedules and then computing incremental changes to the value functions. Because full exhaustive construction becomes computationally prohibitive for moderate-sized problems, we propose a filtering approach to prune the working set. Empirical results demonstrate that this filtering is effective at reducing computation while incurring only negligible reduction in quality. In summarizing our findings, we provide a characterization of the run-time vs quality trade-off involved.

Making the contents generated by Large Language Model (LLM) such as ChatGPT, accurate, credible and traceable is crucial, especially in complex knowledge-intensive tasks that require multi-step reasoning and each of which needs knowledge to solve. Introducing Information Retrieval (IR) to provide LLM with external knowledge is good potential to solve this problem. However, where and how to introduce IR into LLM is a big challenge. Previous work has the disadvantage that the wrong knowledge retrieved by IR misleads the LLM or breaks the reasoning chain of LLM. In this paper, we propose a novel framework called Search-in-the-Chain (SearChain) for the interaction between LLM and IR to solve the challenges. First, LLM generates the global reasoning chain called Chain-of-Query (CoQ) where each node consists of an IR-oriented query and the answer to the query. Second, IR verifies the answer of each node of CoQ, it corrects the answer that is not consistent with the retrieved information when IR gives high confidence, which improves the credibility. Third, LLM can mark its missing knowledge in CoQ and IR can provide this knowledge to LLM. These three operations improve the accuracy of LLM for complex knowledge-intensive tasks in terms of reasoning ability and knowledge. Finally, SearChain generates the reasoning process and marks references to supporting documents for each reasoning step, which improves traceability. SearChain transforms the topology of reasoning from chain to tree, which can modify the reasoning direction. Experiment shows that SearChain outperforms baselines on complex knowledge-intensive tasks including multi-hop question-answering, slot filling, fact checking, and long-form question-answering.

The field of mobile and wearable computing is undergoing a revolutionary integration of machine learning. Devices can now diagnose diseases, predict heart irregularities, and unlock the full potential of human cognition. However, the underlying algorithms powering these predictions are not immune to biases with respect to sensitive attributes (e.g., gender, race), leading to discriminatory outcomes. The goal of this work is to explore the extent to which the mobile and wearable computing community has adopted ways of reporting information about datasets and models to surface and, eventually, counter biases. Our systematic review of papers published in the Proceedings of the ACM Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT) journal from 2018-2022 indicates that, while there has been progress made on algorithmic fairness, there is still ample room for growth. Our findings show that only a small portion (5%) of published papers adheres to modern fairness reporting, while the overwhelming majority thereof focuses on accuracy or error metrics. To generalize these results across venues of similar scope, we analyzed recent proceedings of ACM MobiCom, MobiSys, and SenSys, IEEE Pervasive, and IEEE Transactions on Mobile Computing Computing, and found no deviation from our primary result. In light of these findings, our work provides practical guidelines for the design and development of mobile and wearable technologies that not only strive for accuracy but also fairness.

Anthropogenic pollution of hydrological systems affects diverse communities and ecosystems around the world. Data analytics and modeling tools play a key role in fighting this challenge, as they can help identify key sources as well as trace transport and quantify impact within complex hydrological systems. Several tools exist for simulating and tracing pollutant transport throughout surface waters using detailed physical models; these tools are powerful, but can be computationally intensive, require significant amounts of data to be developed, and require expert knowledge for their use (ultimately limiting application scope). In this work, we present a graph modeling framework -- which we call ${\tt HydroGraphs}$ -- for understanding pollutant transport and fate across waterbodies, rivers, and watersheds. This framework uses a simplified representation of hydrological systems that can be constructed based purely on open-source data (National Hydrography Dataset and Watershed Boundary Dataset). The graph representation provides an flexible intuitive approach for capturing connectivity and for identifying upstream pollutant sources and for tracing downstream impacts within small and large hydrological systems. Moreover, the graph representation can facilitate the use of advanced algorithms and tools of graph theory, topology, optimization, and machine learning to aid data analytics and decision-making. We demonstrate the capabilities of our framework by using case studies in the State of Wisconsin; here, we aim to identify upstream nutrient pollutant sources that arise from agricultural practices and trace downstream impacts to waterbodies, rivers, and streams. Our tool ultimately seeks to help stakeholders design effective pollution prevention/mitigation practices and evaluate how surface waters respond to such practices.

Pretrained language models have demonstrated extraordinary capabilities in language generation. However, real-world tasks often require controlling the distribution of generated text in order to mitigate bias, promote fairness, and achieve personalization. Existing techniques for controlling the distribution of generated text only work with quantified distributions, which require pre-defined categories, proportions of the distribution, or an existing corpus following the desired distributions. However, many important distributions, such as personal preferences, are unquantified. In this work, we tackle the problem of generating text following arbitrary distributions (quantified and unquantified) by proposing Nano, a few-shot human-in-the-loop training algorithm that continuously learns from human feedback. Nano achieves state-of-the-art results on single topic/attribute as well as quantified distribution control compared to previous works. We also show that Nano is able to learn unquantified distributions, achieves personalization, and captures differences between different individuals' personal preferences with high sample efficiency.

Fighting fires was always dangerous. But with climate change, there have been more wildfires, which means even more risky rescue missions for local firefighting squads. That's why multiple different companies and teams of scientists are working to develop robots that can scope out burning buildings before human firefighters have to enter. The latest entrant is FireBot, a remote-controlled robot that can withstand temperatures as high as 650 degrees Celsius. At that scalding temperature, a firefighter wearing a protective suit can only withstand about 15 minutes of exposure.

Artificial intelligence is developing too fast for regulators to keep up, the UK's deputy prime minister is to announce as he aims to galvanise other countries to take the threat seriously in advance of the UK's AI safety summit in November. Oliver Dowden will use a speech at the UN general assembly on Friday to sound the alarm over the lack of regulation of AI, which he says is developing faster than many policymakers thought possible. Dowden will urge other countries to come together to create an international regulatory system, something the UK is keen to promote when it hosts the summit at Bletchley Park. According to comments released before the speech, Dowden will say: "The starting gun has been fired on a globally competitive race in which individual companies as well as countries will strive to push the boundaries as far and fast as possible. "In the past, leaders have responded to scientific and technological developments with retrospective regulation.

As the United Nations General Assembly gathered this week in New York, the U.N. Secretary-General's envoy on technology, Amandeep Gill, hosted an event titled Governing AI for Humanity, where participants discussed the risks that AI might pose and the challenges of achieving international cooperation on artificial intelligence. Secretary-General António Guterres and Gill have said they believe that a new U.N. agency will be required to help the world cooperate in managing this powerful technology. But the issues that the new entity would seek to address and its structure are yet to be determined, and some observers say that ambitious plans for global cooperation like this rarely get the required support of powerful nations. Gill has led efforts to make advanced forms of technology safer before. He was chair of the Group of Governmental Experts of the Convention on Certain Conventional Weapons when the Campaign to Stop Killer Robots, which sought to compel governments to outlaw the development of lethal autonomous weapons systems, failed to gain traction with global superpowers including the U.S. and Russia.

Google's mission statement is to make the'world's information universally accessible' - but that hasn't stopped it from self-censoring to avoid offending Russia. A new study has shown the search giant's artificial intelligence chatbot, Bard, mostly refuses to answer critical questions about Russian President Vladimir Putin. In fact, it won't answer 90 percent of queries regardless of how offensive or inoffensive they are. One of the two researchers in Switzerland who did the test believe Google is being'pushed' by the Kremlin to censor anything critical about the Russian regime. Google's artificial intelligence chatbot, Bard, mostly refuses to answer critical questions about Russian President Vladimir Putin Mykola Makhortykh, a post-doctoral lecturer at the University of Bern and one of the researchers, told DailyMail.com: 'My personal opinion is that Google might have been pushed by the Russian government to censor some of the results which were critical to the Kremlin similar to how it was done by Yandex.'