In adversarial data collection (ADC), a human workforce interacts with a model in real time, attempting to produce examples that elicit incorrect predictions. Researchers hope that models trained on these more challenging datasets will rely less on superficial patterns, and thus be less brittle. However, despite ADC's intuitive appeal, it remains unclear when training on adversarial datasets produces more robust models. In this paper, we conduct a large-scale controlled study focused on question answering, assigning workers at random to compose questions either (i) adversarially (with a model in the loop); or (ii) in the standard fashion (without a model). Across a variety of models and datasets, we find that models trained on adversarial data usually perform better on other adversarial datasets but worse on a diverse collection of out-of-domain evaluation sets. Finally, we provide a qualitative analysis of adversarial (vs standard) data, identifying key differences and offering guidance for future research.

Inpainting is a learned interpolation technique that is based on generative modeling and used to populate masked or missing pieces in an image; it has wide applications in picture editing and retouching. Recently, inpainting started being used for watermark removal, raising concerns. In this paper we study how to manipulate it using our markpainting technique. First, we show how an image owner with access to an inpainting model can augment their image in such a way that any attempt to edit it using that model will add arbitrary visible information. We find that we can target multiple different models simultaneously with our technique. This can be designed to reconstitute a watermark if the editor had been trying to remove it. Second, we show that our markpainting technique is transferable to models that have different architectures or were trained on different datasets, so watermarks created using it are difficult for adversaries to remove. Markpainting is novel and can be used as a manipulation alarm that becomes visible in the event of inpainting.

With the capability of accurately representing a functional relationship between the inputs of a physical system's model and output quantities of interest, neural networks have become popular for surrogate modeling in scientific applications. However, as these networks are over-parameterized, their training often requires a large amount of data. To prevent overfitting and improve generalization error, regularization based on, e.g., $\ell_1$- and $\ell_2$-norms of the parameters is applied. Similarly, multiple connections of the network may be pruned to increase sparsity in the network parameters. In this paper, we explore the effects of sparsity promoting $\ell_1$-regularization on training neural networks when only a small training dataset from a high-fidelity model is available. As opposed to standard $\ell_1$-regularization that is known to be inadequate, we consider two variants of $\ell_1$-regularization informed by the parameters of an identical network trained using data from lower-fidelity models of the problem at hand. These bi-fidelity strategies are generalizations of transfer learning of neural networks that uses the parameters learned from a large low-fidelity dataset to efficiently train networks for a small high-fidelity dataset. We also compare the bi-fidelity strategies with two $\ell_1$-regularization methods that only use the high-fidelity dataset. Three numerical examples for propagating uncertainty through physical systems are used to show that the proposed bi-fidelity $\ell_1$-regularization strategies produce errors that are one order of magnitude smaller than those of networks trained only using datasets from the high-fidelity models.

As the world begins to travel again, Europe is sending migrants a loud message: Stay away! Greek border police are firing bursts of deafening noise from an armored truck over the frontier into Turkey. Mounted on the vehicle, the long-range acoustic device, or "sound cannon," is the size of a small TV set but can match the volume of a jet engine. It's part of a vast array of physical and experimental new digital barriers being installed and tested during the quiet months of the coronavirus pandemic at the 200-kilometer (125-mile) Greek border with Turkey to stop people entering the European Union illegally. Nearby observation towers are being fitted with long-range cameras, night vision, and multiple sensors.

Hackers are having a field day, and weak authentication is a major cause. The vast majority of cyberattacks -- some 80%, statistics show -- have their roots in compromised passwords that hackers get hold of. All it takes is one stolen password for hackers to wreak havoc; and according to experts, that single password breach can cost enterprise firms over $7 million. Many schemes have been tried to build up password security, including increased education and 2FA. But despite that, password compromise statistics remain stubbornly high, cybersecurity education programs, although widespread, don't seem to work, and 2FA has its own security issues.

A robotic arm attached to the outside of the International Space Station has been hit with space junk and visibly damaged, according to the Canadian Space Agency. In a blog post, the CSA notes that'a small section of the arm boom and thermal blanket' of Canadarm2 was hit. The space agency first noticed the incident'during a routine inspection' on May 12. 'Despite the impact, results of the ongoing analysis indicate that the arm's performance remains unaffected,' CSA wrote in the post, adding that the robotic arm is'continuing to conduct its planned operations.' A robotic arm attached to the outside of the International Space Station has been hit with space junk and visibly damaged, according to the Canadian Space Agency. According to the US space agency, more than 27,000 pieces of space junk are tracked.

Eric Schmidt, who leads a U.S. government initiative to integrate AI into national security, warned Monday that the EU's AI transparency requirements would be "very harmful to Europe." Speaking at POLITICO's AI Summit, Schmidt criticized the provisions of the EU's AI bill that require algorithms to be transparent. "It's just a proposal, but if you would adopt it without modification, it would be a very big setback for Europe," said Schmidt, who chairs the National Security Commission on Artificial Intelligence (NSCAI) and is a former CEO of Google. The EU's proposal "requires that the system would be able to explain itself. But machine learning systems cannot fully explain how they make their decisions," Schmidt said.

Artificial Intelligence (AI) today is rapidly changing the face of technology. But with the ability to create devices and systems capable of autonomous decisions, arises the need for legislation to monitor AI. Amazon's now scrapped AI recruiting tool is a prime example where it was discovered that the AI tool had a bias towards men since it had been trained on 10 years of data when men held most tech positions. As we constantly move towards a more technology integrated world, the need for the right balance in legislation grows more important. It needs to protect the rights of the citizens alongside ensuring that it is not a hindrance to technology and business growth.

'Tucker Carlson Tonight' host examines how the left uses'science' to push their agenda This is a rush transcript from "Tucker Carlson Tonight," May 28, 2021. This copy may not be in its final form and may be updated. At some point, though it's hard to imagine now, the current revolution will end. Ultimately, all revolutions do end, they can't be sustained. And when ours does, we'll wake up one morning in a country where we don't have to lie about everything all the time, where Math is allowed, where we can acknowledge the profound and inherent differences between men and women without being fired for it. The question is, when it does come, what will be left of our society? We can't know the answer in detail. We hope that reason remains. Reason, logic, the ability to think clearly and rationally. That's the one thing we can't lose. We're going to need it to rebuild. That's why of all the moral atrocities being committed at the moment in the name of equity and inclusion, it is the relentless attacks on science that should command our special attention. So for the next hour, we're going to consider those attacks in some detail and we are going to start with America's response to the COVID pandemic. When the coronavirus first arrived in our country last winter, most Americans uncritically accepted what the authorities said about it. They thought they could trust the people in charge. Few imagined that our leaders would leverage a public health emergency for their own political gain. That seemed like the one line that even politicians wouldn't cross, and yet almost immediately, they crossed it.

An advanced artificial intelligence flight computer will help the UAE's lunar rover explore the surface of the Moon. The navigation computer is being developed by Canadian space firm Mission Control Space Services. It will recognise geological features as the Emirati rover, Rashid, drives around the unstable terrain of the lunar surface. The computer will be installed on a Japanese lander that will take Rashid to the Moon next year, from where it will receive data from the rover. It will also send information back to Earth to be studied by scientists at the Mohammed bin Rashid Space Centre.