The emergence of increasingly sophisticated artificial intelligence (AI) systems have sparked intense debate among researchers, policymakers, and the public due to their potential to surpass human intelligence and capabilities in all domains. In this paper, I propose a game-theoretic framework that captures the strategic interactions between a human agent and a potential superhuman machine agent. I identify four key assumptions: Strategic Unpredictability, Access to Machine's Strategy, Rationality, and Superhuman Machine. The main result of this paper is an impossibility theorem: these four assumptions are inconsistent when taken together, but relaxing any one of them results in a consistent set of assumptions. Two straightforward policy recommendations follow: first, policymakers should control access to specific human data to maintain Strategic Unpredictability; and second, they should grant select AI researchers access to superhuman machine research to ensure Access to Machine's Strategy holds. My analysis contributes to a better understanding of the context that can shape the theoretical development of superhuman AI.

Deep neural networks have achieved significant success in the last decades, but they are not well-calibrated and often produce unreliable predictions. A large number of literature relies on uncertainty quantification to evaluate the reliability of a learning model, which is particularly important for applications of out-of-distribution (OOD) detection and misclassification detection. We are interested in uncertainty quantification for interdependent node-level classification. We start our analysis based on graph posterior networks (GPNs) that optimize the uncertainty cross-entropy (UCE)-based loss function. We describe the theoretical limitations of the widely-used UCE loss. To alleviate the identified drawbacks, we propose a distance-based regularization that encourages clustered OOD nodes to remain clustered in the latent space. We conduct extensive comparison experiments on eight standard datasets and demonstrate that the proposed regularization outperforms the state-of-the-art in both OOD detection and misclassification detection.

Abstaining classifiers have the option to abstain from making predictions on inputs that they are unsure about. These classifiers are becoming increasingly popular in high-stakes decision-making problems, as they can withhold uncertain predictions to improve their reliability and safety. When evaluating black-box abstaining classifier(s), however, we lack a principled approach that accounts for what the classifier would have predicted on its abstentions. These missing predictions matter when they can eventually be utilized, either directly or as a backup option in a failure mode. In this paper, we introduce a novel approach and perspective to the problem of evaluating and comparing abstaining classifiers by treating abstentions as missing data. Our evaluation approach is centered around defining the counterfactual score of an abstaining classifier, defined as the expected performance of the classifier had it not been allowed to abstain. We specify the conditions under which the counterfactual score is identifiable: if the abstentions are stochastic, and if the evaluation data is independent of the training data (ensuring that the predictions are missing at random), then the score is identifiable. Note that, if abstentions are deterministic, then the score is unidentifiable because the classifier can perform arbitrarily poorly on its abstentions. Leveraging tools from observational causal inference, we then develop nonparametric and doubly robust methods to efficiently estimate this quantity under identification. Our approach is examined in both simulated and real data experiments.

Modern deep learning requires large volumes of data, which could contain sensitive or private information that cannot be leaked. Recent work has shown for homogeneous neural networks a large portion of this training data could be reconstructed with only access to the trained network parameters. While the attack was shown to work empirically, there exists little formal understanding of its effective regime which datapoints are susceptible to reconstruction. In this work, we first build a stronger version of the dataset reconstruction attack and show how it can provably recover the \emph{entire training set} in the infinite width regime. We then empirically study the characteristics of this attack on two-layer networks and reveal that its success heavily depends on deviations from the frozen infinite-width Neural Tangent Kernel limit. Next, we study the nature of easily-reconstructed images. We show that both theoretically and empirically, reconstructed images tend to "outliers" in the dataset, and that these reconstruction attacks can be used for \textit{dataset distillation}, that is, we can retrain on reconstructed images and obtain high predictive accuracy.

From spotty leopards to stripy zebras, nature has no shortage of distinct patterns on animals and plants. Now, the age-old question of how these patterns developed may have finally been solved. Scientists have shown that the same physical process that helps remove dirt from laundry could play a role in how tropical fish get their colourful spots and stripes. For their study, the team at the University of Colorado Boulder drew on the groundbreaking work of British computer pioneer Alan Turing, dating back more than 70 years. They believe their findings could help develop new materials and even new drugs.

When politicians, tech executives, and researchers gathered in the UK last week to discuss the risks of artificial intelligence, one prominent worry was that algorithms might someday turn against their human masters. More quietly, the group made progress on controlling the use of AI for military ends. On November 1, at the US embassy in London, US vice president Kamala Harris announced a range of AI initiatives, and her warnings about the threat AI poses to human rights and democratic values got people's attention. But she also revealed a declaration signed by 31 nations to set guardrails around military use of AI. It pledges signatories to use legal reviews and training to ensure military AI stays within international laws, develop the technology cautiously and transparently, avoid unintended biases in systems that use AI, and continue to discuss how the technology can be developed and deployed responsibly.

'Botched experiments' by Elon Musk's Neuralink allegedly'kept suffering animals alive for no reason and malpractice caused monkey's brains to hemorrhage' during rushed brain chip testing, a former Neuralink employee and internal lab notes have previously revealed. The billionaire's startup is accused of violating the Animal Welfare Act with its experiments at the University of California, Davis, from 2017 through 2020, which'sacrificed all the animals involved,' a former Neuralink employee, who asked to remain anonymous, told DailyMail.com. One case stood out to them - a monkey sacrificed ahead of schedule due to errors allegedly made during surgery. The Physicians Committee for Responsible Medicine filed a lawsuit against the University of California, Davis, where the experiments were held, claiming it has to hand over video footage and photographs of the experiments under California's Public Records Act. Pictured is an image of a monkey shown on Neuralink's website'There was no reason to use it,' the former employee, who worked as a necropsy technician, told DailyMail.com.

Meta will require advertisers to disclose whether the ads they submit for its websites have been digitally altered, including through the use of AI tools, if they're political or social in nature. Ads that have been digitally altered will be marked as such on Meta's platforms, in the same way some advertisements come with a "Paid for" disclaimer. The company will start implementing the rule in the new year, just as the campaign period for what's expected to be a brutal and divisive 2024 US presidential elections heats up. In a blog post, Meta explained that advertisers have to disclose in the advertising flow if they submit a social issue, electoral or political ad with photorealistic images or videos -- or one with realistic sounding audio -- that was altered to make a real person say or do something they didn't actually say or do. They're also required to tell Meta whether they're submitting an ad with a realistic-looking person that doesn't exist, a realistic-looking event that didn't happen or an altered footage of a real event that truly occurred. If they submit a fake image, video or audio recording of an event that allegedly took place -- say, something they created with the help of AI image generators -- they have to notify Meta, as well.

If you look at the UN Sustainable Development Goals, it's clear that robots have a huge role to play in advancing the SDGs. However the field of Sustainable Robotics is more than just the application area. For every application that robotics can improve in sustainability, you have to also address the question – what are the additional costs or benefits all the way along the supply chain. What are the'externalities', or additional costs/benefits, of using robots to solve the problem. Solving our economic and environmental global challenges should not involve adding to the existing problems or creating new ones.

Building efficient, accurate and generalizable reduced order models of developed turbulence remains a major challenge. This manuscript approaches this problem by developing a hierarchy of parameterized reduced Lagrangian models for turbulent flows, and investigates the effects of enforcing physical structure through Smoothed Particle Hydrodynamics (SPH) versus relying on neural networks (NN)s as universal function approximators. Starting from Neural Network (NN) parameterizations of a Lagrangian acceleration operator, this hierarchy of models gradually incorporates a weakly compressible and parameterized SPH framework, which enforces physical symmetries, such as Galilean, rotational and translational invariances. Within this hierarchy, two new parameterized smoothing kernels are developed in order to increase the flexibility of the learn-able SPH simulators. For each model we experiment with different loss functions which are minimized using gradient based optimization, where efficient computations of gradients are obtained by using Automatic Differentiation (AD) and Sensitivity Analysis (SA). Each model within the hierarchy is trained on two data sets associated with weekly compressible Homogeneous Isotropic Turbulence (HIT): (1) a validation set using weakly compressible SPH; and (2) a high fidelity set from Direct Numerical Simulations (DNS). Numerical evidence shows that encoding more SPH structure improves generalizability to different turbulent Mach numbers and time shifts, and that including the novel parameterized smoothing kernels improves the accuracy of SPH at the resolved scales.