Bayesian methods hold significant promise for improving the uncertainty quantification ability and robustness of deep neural network models. Recent research has seen the investigation of a number of approximate Bayesian inference methods for deep neural networks, building on both the variational Bayesian and Markov chain Monte Carlo (MCMC) frameworks. A fundamental issue with MCMC methods is that the improvements they enable are obtained at the expense of increased computation time and model storage costs. In this paper, we investigate the potential of sparse network structures to flexibly trade-off model storage costs and inference run time against predictive performance and uncertainty quantification ability. We use stochastic gradient MCMC methods as the core Bayesian inference method and consider a variety of approaches for selecting sparse network structures. Surprisingly, our results show that certain classes of randomly selected substructures can perform as well as substructures derived from state-of-the-art iterative pruning methods while drastically reducing model training times.

I present arguments against the hypothesis put forward by Silver, Singh, Precup, and Sutton ( https://www.sciencedirect.com/science/article/pii/S0004370221000862 ) : reward maximization is not enough to explain many activities associated with natural and artificial intelligence including knowledge, learning, perception, social intelligence, evolution, language, generalisation and imitation. I show such reductio ad lucrum has its intellectual origins in the political economy of Homo economicus and substantially overlaps with the radical version of behaviourism. I show why the reinforcement learning paradigm, despite its demonstrable usefulness in some practical application, is an incomplete framework for intelligence -- natural and artificial. Complexities of intelligent behaviour are not simply second-order complications on top of reward maximisation. This fact has profound implications for the development of practically usable, smart, safe and robust artificially intelligent agents.

Multi-model ensemble analysis integrates information from multiple climate models into a unified projection. However, existing integration approaches based on model averaging can dilute fine-scale spatial information and incur bias from rescaling low-resolution climate models. We propose a statistical approach, called NN-GPR, using Gaussian process regression (GPR) with an infinitely wide deep neural network based covariance function. NN-GPR requires no assumptions about the relationships between models, no interpolation to a common grid, no stationarity assumptions, and automatically downscales as part of its prediction algorithm. Model experiments show that NN-GPR can be highly skillful at surface temperature and precipitation forecasting by preserving geospatial signals at multiple scales and capturing inter-annual variability. Our projections particularly show improved accuracy and uncertainty quantification skill in regions of high variability, which allows us to cheaply assess tail behavior at a 0.44$^\circ$/50 km spatial resolution without a regional climate model (RCM). Evaluations on reanalysis data and SSP245 forced climate models show that NN-GPR produces similar, overall climatologies to the model ensemble while better capturing fine scale spatial patterns. Finally, we compare NN-GPR's regional predictions against two RCMs and show that NN-GPR can rival the performance of RCMs using only global model data as input.

Much of the past work on fairness in machine learning has focused on forcing the predictions of classifiers to have similar statistical properties for individuals of different demographics. Yet, such methods often simply perform a rescaling of the classifier scores and ignore whether individuals of different groups have similar features. Our proposed method, Optimal Transport to Fairness (OTF), applies Optimal Transport (OT) to take this similarity into account by quantifying unfairness as the smallest cost of OT between a classifier and any score function that satisfies fairness constraints. For a flexible class of linear fairness constraints, we show a practical way to compute OTF as an unfairness cost term that can be added to any standard classification setting. Experiments show that OTF can be used to achieve an effective trade-off between predictive power and fairness.

WASHINGTON--The Internal Revenue Service is scrapping its use of a private facial-recognition system to authenticate taxpayers' identities for online accounts, the agency said Monday after criticism from lawmakers in both parties over privacy concerns. "Everyone should feel comfortable with how their personal information is secured, and we are quickly pursuing short-term options that do not involve facial recognition," IRS Commissioner Charles Rettig said in a statement on Monday.

It didn't take long for the Internal Revenue Service to respond to pressure to drop facial recognition. The agency has told Senator Ron Wyden it plans to back away from using facial recognition for verification purposes. Wyden cautioned the transition would "take time," but he saw this as evidence the Biden administration knew privacy and security weren't "mutually exclusive" concepts. The New York Times understood the shift would take place over weeks to minimize disruptions to tax filing season. We've asked ID.me, the company slated to provide facial recognition to the IRS, for comment.

The IRS said Monday it would “transition away” from using the company ID.me for new accounts “over the coming weeks” and would develop an additional authentication process that does not involve facial recognition.

It's not just Republican senators upset over the Internal Revenue Service's plans to adopt ID.me facial recognition. Democratic House Representatives Ted Lieu, Anna Eshoo, Pramila Jayapal and Yvette Clarke have sent a letter to IRS Commissioner Charles Rettig demanding his agency drop plans to use facial recognition starting this summer. They're concerned the plan will compromise privacy and security by forcing uploads of sensitive data to a database that could be a "prime target" for cyberattacks like the one that exposed license plates at Customs and Border Protection in 2019. The members of Congress were also worried about lingering accuracy and bias problems with facial recognition systems. While ID.me maintains its technology is equitable and inclusive, the Democrats pointed to a National Institute of Standards and Technology study that showed many more false positives for Asian and Black faces, even in one-to-one matching systems like the one ID.me

On Thursday, a group of Senate Republicans wrote a letter to Rettig that cited the government's "unfortunate history of data breaches." The IRS, the letter said, had "unilaterally decided to allow an outside contractor to stand as the gatekeeper between citizens and necessary government services. The decision millions of Americans are forced to make is to pay the toll of giving up their most personal information, biometric data, to an outside contractor or return to the era of a paper-driven bureaucracy."

Experts have been warning of something called the "quantum apocalypse" – the point when quantum computers become a reality and render most methods of internet encryption useless. Boris Johnson promised in November that the UK would "go big" on quantum computing – a new and more powerful way of processing information, based on quantum physics. If you imagine a standard computer to be like a horse and cart, then a quantum computer is "more like a sports car – a huge leap forward", explained the BBC. The UK is aiming to secure 50% of the global quantum computing market by 2040, said The Guardian, by investing in the National Quantum Computing Centre in Harwell, Oxfordshire. But the US and China have already taken huge steps to revolutionise research in the field, with the Americans achieving a "dramatic lead in quantum computing patents", said Scientific American. A leaked Google research paper published in 2019 suggested that a computer designed by the tech giant had achieved "quantum supremacy" – defined by The Independent as the ability to perform "a calculation that was far beyond the reach of today's most powerful supercomputers".