Three years after its involvement with the military's controversial Project Maven program led to employee strife within its walls, Google reportedly hopes to once again work with the Pentagon. According to The New York Times, the company is "aggressively" pursuing the Defense Department's Joint Warfighting Cloud Capability cloud contract. A Google spokesperson confirmed to Engadget it was pursuing a bid. Announced at the start of July, the program is a replacement for the military's cancelled $10 billion Joint Enterprise Defense Infrastructure initiative. With JEDI, the Pentagon had planned to modernize its IT infrastructure with help from Microsoft.

A United States drone attack that killed 10 civilians, including seven children, in the Afghan capital Kabul in August did not violate the laws of war, an internal Pentagon review has concluded. Speaking to reporters at the Pentagon on Wednesday, US Air Force Lieutenant General Sami Said said "execution errors", including "confirmation bias" and "communication breakdowns", contributed to the deadly attack. But Said, who acts as inspector general of the US Air Force, said it was not a violation of the law of war or a result of negligence. "It was an honest mistake," Said said. The drone strike on August 29 came amid the US's chaotic military withdrawal from Kabul, and Said also stressed that it took place as American forces were contending with threats from the Islamic State in Khorasan Province, ISKP (ISIS-K), an affiliate of ISIS (ISIL).

Notes from the President, based on a panel for CDO Magazine "Building AI/ML Organizations that are Ethical by Design" Some answers to some thoughtful questions, based on a panel completed by our president, Amina Al Sherif, moderated by Anno.Ai Chief AI Officer Ashley Antonides. AI and data ethics can most broadly be defined as guidelines by your organization that determines what right from wrong is. We loved this definition, because it acknowledges how ethically right and wrong algorithms largely depend on an organization's internal culture and ethical fabric, as well as what vertical or sector that organization serves. AI and data ethics is mostly about your data, and how it is handled in training a machine learning (ML) system. Models inherently cannot be unethical- because of their extreme dependencies on the data on which they are trained.

In as little as 10 to 15 years, software engineering may look more like a technical conversation between humans and computers than a process of manually refining specifications and code, and the software ecosystem must prepare for that future. That is just one of the conclusions of a study titled "Architecting the Future of Software Engineering: A National Agenda for Software Engineering Research & Development" that was released today by the Software Engineering Institute (SEI) at Carnegie Mellon University. For this study, the SEI engaged the software engineering community and assembled an advisory board of senior thought leaders across commercial industry, academia and government, with participation from Microsoft, Google, SpaceX, Lockheed Martin, Boeing, DARPA and others. The board was chaired by Deb Frincke, associate laboratory director for national security sciences, Oak Ridge National Laboratory. With their input, the study team worked to create the multiyear research and development vision and roadmap for engineering next-generation software-reliant systems.

THE ROAD TO WELLVILLE?: AI-powered health apps are proliferating. The mental stresses of the pandemic have fueled a boom in wellness devices that track speech, facial expressions and even eye blinks to assess emotional states. This kind of "affective computing" can replicate therapy or detect depression when in-person care isn't available -- and even remotely monitor workers and children. But the burst of interest is heightening concerns about whether there's enough government oversight of the technology, known as "emotion AI". Future Pulse spoke about the tension points with Alexandrine Royer, a doctoral candidate studying the field and the digital economy at the University of Cambridge and a student fellow at the Leverhulme Centre for the Future of Intelligence.

Last November, the Chinese government ordered Ant Group, a business spun out of Alibaba that operates the ubiquitous Alipay mobile payments platform and other financial services, to cancel its hotly anticipated IPO at the last moment. Shortly after ride-hailing giant DiDi went ahead with an IPO in New York this summer despite government concerns, officials removed the company's app from Chinese app stores and ordered it to comply with an extensive cybersecurity review. Soon after, ByteDance, operator of wildly popular news and entertainment apps, as well as the short-video sensation TikTok outside of China, shelved its own plans for an IPO to comply with tighter government rules around data protection and security. So it's a little odd that two titans of China's artificial intelligence industry, SenseTime and Megvii, are proceeding with plans for IPOs seemingly unbothered, with listings on the Hong Kong and Shanghai stock exchanges, respectively. After a decade of unchecked growth, many Chinese tech firms now face a stark new reality, with canceled IPOs, stricter regulations, and hefty fines.

To mark Israel's Memorial Day in 2021, the Israel Defense Forces musical ensembles collaborated with a company that specializes in synthetic videos, also known as "deepfake" technology, to bring photos from the 1948 Israeli-Arab war to life. They produced a video in which young singers clad in period uniforms and carrying period weapons sang "Hareut," an iconic song commemorating soldiers killed in combat. As they sing, the musicians stare at faded black-and-white photographs they hold. The past comes to life, Harry Potter style. For the past few years, my colleagues and I at UMass Boston's Applied Ethics Center have been studying how everyday engagement with AI challenges the way people think about themselves and politics. We've found that AI has the potential to weaken people's capacity to make ordinary judgments.

Seoul – A toy-like robot with scowling blue eyes and a North Korean flag across its chest roams around a classroom at a university in Pyongyang in a recent demonstration of tools aimed at helping children learn basic math, music and English. The footage, broadcast by North Korean state television KRT, also showed two other larger plastic robots, each with a vaguely humanoid appearance. North Korean leader Kim Jong Un has been pushing for education reform in recent years by spurring technological and scientific innovation. "I help teach educational technology that enhances children's intelligence," said the 80-centimeter (31.5 inches) tall robot in a female voice, waving its arms. A second robot featured a smiley face on a screen embedded inside a white round head, while another wore a blue plastic suit and white-rimmed glasses, the KRT footage showed.

Atlas and Spot won't have blank spaces on their resumes. The Boston Dynamics robots, famous for their YouTube parkour and dancing exploits, could land a position in a heartbeat, as can many non-human job-seekers, part of the wave of robot hires amid a human worker shortage. As the Associated Press reported, the pandemic ushered in these workplace changes. Companies are starting to automate service sector jobs, thanks to higher labor costs and the aforementioned worker shortages. Machines can do many tasks such as toss pizza dough, transport hospital linens, inspect gauges and sort goods.

Increasing levels of renewable generation motivate a growing interest in data-driven approaches for AC optimal power flow (AC OPF) to manage uncertainty; however, a lack of disciplined dataset creation and benchmarking prohibits useful comparison among approaches in the literature. To instill confidence, models must be able to reliably predict solutions across a wide range of operating conditions. This paper develops the OPF-Learn package for Julia and Python, which uses a computationally efficient approach to create representative datasets that span a wide spectrum of the AC OPF feasible region. Load profiles are uniformly sampled from a convex set that contains the AC OPF feasible set. For each infeasible point found, the convex set is reduced using infeasibility certificates, found by using properties of a relaxed formulation. The framework is shown to generate datasets that are more representative of the entire feasible space versus traditional techniques seen in the literature, improving machine learning model performance.