In a new series of experiments, artificial intelligence (A.I.) algorithms were able to influence people's preferences for fictitious political candidates or potential romantic partners, depending on whether recommendations were explicit or covert. Ujué Agudo and Helena Matute of Universidad de Deusto in Bilbao, Spain, present these findings in the open-access journal PLOS ONE on April 21, 2021. From Facebook to Google search results, many people encounter A.I. algorithms every day. Private companies are conducting extensive research on the data of their users, generating insights into human behavior that are not publicly available. Academic social science research lags behind private research, and public knowledge on how A.I. algorithms might shape people's decisions is lacking. To shed new light, Agudo and Matute conducted a series of experiments that tested the influence of A.I. algorithms in different contexts.

The weights of a deep neural network model are optimized in conjunction with the governing flow equations to provide a model for sub-grid-scale stresses in a temporally developing plane turbulent jet at Reynolds number $Re_0=6\,000$. The objective function for training is first based on the instantaneous filtered velocity fields from a corresponding direct numerical simulation, and the training is by a stochastic gradient descent method, which uses the adjoint Navier--Stokes equations to provide the end-to-end sensitivities of the model weights to the velocity fields. In-sample and out-of-sample testing on multiple dual-jet configurations show that its required mesh density in each coordinate direction for prediction of mean flow, Reynolds stresses, and spectra is half that needed by the dynamic Smagorinsky model for comparable accuracy. The same neural-network model trained directly to match filtered sub-grid-scale stresses -- without the constraint of being embedded within the flow equations during the training -- fails to provide a qualitatively correct prediction. The coupled formulation is generalized to train based only on mean-flow and Reynolds stresses, which are more readily available in experiments. The mean-flow training provides a robust model, which is important, though a somewhat less accurate prediction for the same coarse meshes, as might be anticipated due to the reduced information available for training in this case. The anticipated advantage of the formulation is that the inclusion of resolved physics in the training increases its capacity to extrapolate. This is assessed for the case of passive scalar transport, for which it outperforms established models due to improved mixing predictions.

Abstract--In this paper we tackle two important challenges related to the accurate partial singular value decomposition (SVD) and numerical rank estimation of a huge matrix to use in low-rank learning problems in a fast way. We use the concepts of Krylov subspaces such as the Golub-Kahan bidiagonalization process as well as Ritz vectors to achieve these goals. Our experiments identify various advantages of the proposed methods compared to traditional and randomized SVD (R-SVD) methods with respect to the accuracy of the singular values and corresponding singular vectors computed in a similar execution time. The proposed methods are appropriate for applications involving huge matrices where accuracy in all spectrum of the desired singular values, and also all of corresponding singular vectors is essential. We evaluate our method in the real application of Riemannian similarity learning (RSL) between two various image datasets of MNIST and USPS.

One of the priorities announced in the 2021 Examination Priorities Report of the U.S. Securities and Exchange Commission's Division of Examinations ("EXAMS") is a review of robo-advisory firms that build client portfolios with exchange-traded funds ("ETF's") and mutual funds. EXAMS notes that these clients are almost entirely retail investors without investments large enough to support the costs of regular human investment advisers. EXAMS sees that the risks involved in these robo-advisor accounts pose particular issues, that retail clients may well not recognize. Accordingly, it may help to reflect on the Laws of Robotics invented by that science fiction author Isaac Asimov (for "I Robot," a short story in his 1950 collection), particularly the First Law: A robot may not injure a human being or, through inaction, allow a human being to come to harm. Investors may not understand the risks associated with specific investments; the risk profiles of mutual funds and of ETF's vary widely, from diversified to concentrated, from simple to complex strategies.

Iran's foreign minister apologized Sunday for recorded comments that were leaked to the public last week that offered a blunt appraisal of the country's power struggles, sparking a political firestorm in Iran less than two months before presidential elections and apparently drawing the ire of Iran's supreme leader. The recordings of Mohammad Javad Zarif included frank comments about the powerful late Iranian Gen. Qassem Soleimani, who was killed by a U.S. drone strike in Iraq last year, as well as criticism of his polices in Syria and his relations with Russia. "I hope that the great people of Iran and all the lovers of General (Soleimani) and especially the great family of Soleimani, will forgive me," Zarif said in an Instagram post. In a speech broadcast later Sunday, Iran's supreme leader Ayatollah Ali Khamenei appeared to lambast Zarif for departing from the official line, although he didn't call him out by name. "It's a big mistake that must not be made by an official of the Islamic Republic," Khamenei said in veiled reference to the leaked comments.

Wendy Gonzalez is the CEO of Sama, the provider of accurate data for ambitious AI. As we reflect on the year that's passed since the start of the Covid-19 pandemic's lockdowns and stay-at-home orders, we can evaluate the rapid acceleration of digital transformation across industries. Where many verticals have made the transition quickly, there's one in particular that cannot afford to make any mistakes with its strategy: healthcare. With increased global accessibility, artificial intelligence (AI) is rapidly becoming a part of long-term transformation plans within healthcare. Through its adaptability and customization, organizations can harness AI to address a range of scenarios.

LOS ALAMOS, N.M., April 22, 2021--A new machine-learning model that generates realistic seismic waveforms will reduce manual labor and improve earthquake detection, according to a study published recently in JGR Solid Earth. "To verify the efficacy of our generative model, we applied it to seismic field data collected in Oklahoma," said Youzuo Lin, a computational scientist in Los Alamos National Laboratory's Geophysics group and principal investigator of the project. "Through a sequence of qualitative and quantitative tests and benchmarks, we saw that our model can generate high-quality synthetic waveforms and improve machine learning-based earthquake detection algorithms." Quickly and accurately detecting earthquakes can be a challenging task. Visual detection done by people has long been considered the gold standard, but requires intensive manual labor that scales poorly to large data sets.

Today, the European Commission proposed regulations for the European Union (EU). The proposed regulations are discussed on the EU site. They are of interest for more than only facial recognition, but as the start of what will be increasing regulation for many aspects of artificial intelligence (AI). There should be zero surprise that facial recognition is the first major aspect of AI to meet with government regulations. This technology is very intrusive and can directly impact the lives of all citizens in many ways.

There has been recently a growing interest in studying adversarial examples on natural language models in the black-box setting. These methods attack natural language classifiers by perturbing certain important words until the classifier label is changed. In order to find these important words, these methods rank all words by importance by querying the target model word by word for each input sentence, resulting in high query inefficiency. A new interesting approach was introduced that addresses this problem through interpretable learning to learn the word ranking instead of previous expensive search. The main advantage of using this approach is that it achieves comparable attack rates to the state-of-the-art methods, yet faster and with fewer queries, where fewer queries are desirable to avoid suspicion towards the attacking agent. Nonetheless, this approach sacrificed the useful information that could be leveraged from the target classifier for that sake of query efficiency. In this paper we study the effect of leveraging the target model outputs and data on both attack rates and average number of queries, and we show that both can be improved, with a limited overhead of additional queries.

If you don't have enough to worry about already, consider a world where AIs are hackers. Hacking is as old as humanity. We are creative problem solvers. We exploit loopholes, manipulate systems, and strive for more influence, power, and wealth. To date, hacking has exclusively been a human activity.