A group of nongovernmental organisations called on the Trump administration to clarify its policy on drone use, saying they are concerned about reported changes to US rules and a lack of transparency in the decision-making process. "We are deeply concerned that the reported new policy, combined with this administration's reported dramatic increase in lethal operations in Yemen and Somalia, will add to an increase in unlawful killings and in civilian casualties," a joint statement said. The organisations include Amnesty International, the US-based Center for Constitutional Rights, Human Rights Watch, the ACLU and others. President Donald Trump signed the 2018 National Defense Authorization Act in December. The act funds the US military but also requires Trump to make known to Congress any changes to previous drone policies by March 12.

In April, two semiautonomous drones, developed by Saildrone, a marine tech startup based in Alameda, California, in close collaboration with the National Oceanic and Atmospheric Administration in Washington, D.C., are set to return from an 8-month tour of the Pacific Ocean. This the first scientific test for the drones, which are powered only by the wind and sun, in the Pacific Ocean. The voyage is an important step in showing that such drones, carrying 15 different sensors, could help replace an aging and expensive array of buoys that are the main way scientists sniff out signs of climate-disrupting El Niño events. If successful, scientists envision fleets of similar drones spreading across the ocean, inviting thoughts of what it could be like to do oceanography without a ship.

Google is not only applying machine learning across its products, but also encouraging other developers to adopt it in third-party services and other use cases. It has now emerged that one of the latter examples is for drones from the U.S. government. Gizmodo this morning reported on a Google partnership last year with the U.S. Department of Defense to improve the latter organization's adoption of machine learning. The DoD's Project Maven is tasked with identifying objects in drone footage. Google supplied the military with TensorFlow APIs, which it notes in a statement are open source.

Afghanistan and Pakistan accuse each other of allowing militant groups to shelter in remote frontier regions to prepare cross-border attacks. U.S. officials have also accused Pakistan of not doing enough to crack down on such groups operating on its side of the border.

Islamabad, Pakistan - At least 20 members of the Pakistani Taliban, including the son of Tehreek-e-Taliban Pakistan (TTP) chief Mullah Fazlullah, have been killed in a US drone strike in the northeastern Afghan province of Kunar, according to a Taliban statement. The fighters, including Abdullah, were killed in a strike on a TTP camp, the group said in a statement texted to local journalists in Pakistan's Bajaur district, which borders Kunar, on Wednesday. According to the statement, 20 fidayeen, or suicide bombers, were killed, while another six fighters were wounded in the strike. Prominent TTP commander Gul Muhammad and fidayeen trainer Yaseen were also killed in the strike, Pakistani newspaper Dawn reported. Pakistan has often accused US and Afghan forces of not doing enough to target TTP forces resident on the Afghan side of the volatile Pakistan-Afghanistan border, where they fled following a series of military operations by Pakistan.

Only months after it was disclosed that the Pentagon was using artificial intelligence (AI) to hunt for terrorists, officials have now acknowledged that Google has been collaborating with the Department of Defense to use AI in analyzing drone footage. The disclosure comes amid an uproar among Google employees who aren't happy to be assisting in the development of military applications. While Google has had controversial contracts with the government before -- most notably with the NSA -- this is its "pilot" project with Project Maven, which is itself the Pentagon's own flagship weaponized AI program. The purpose of Project Maven is to implement Big Data and machine learning into the U.S. military, which officials say is currently in a new AI arms race with China and Russia. According to information from an internal mailing list, Google will now join this arms race and assist the DoD with storing and analyzing the massive troves of data from aerial drones. "We have long worked with government agencies to provide technology solutions.

Gizmodo reports that Google employees are enraged at the company after discovering that the AI they built is presently being used by the US Department of Defense (DoD) to analyze the massive troves of video footage captured by military drones. It's part of Project Maven, aka, Algorithmic Warfare Cross-Functional Team (AWCFT) (PDF), a Pentagon-driven project started last April to make use of big data and machine learning tech to advance the DoD's capabilities in the field. The idea has been to use Google's AI to identify vehicles and other objects in footage from drones; it was previously manually analyzed by humans, but these unmanned vehicles are now capturing more video than staff can keep up with, as the DoD is now racking up "millions of hours" of clips. It's understandable that employees at Google would be concerned and angry to learn that what they've built may be used to assist military activities they may not be aligned with. But the company has attempted to allay those fears with a statement. Google confirmed that it's providing the DoD with APIs for its TensorFlow machine learning system.

That would enable bystanders to use their smartphones to report misbehaving UAVs, Ford stated in a blog post, noting that they used this method to reliably identify drones up to 80 feet away during testing. While that's not far at all for surveillance drones, it might be close enough for drones endangering airports or driving paths -- though it's unclear how quickly the app could identify fast-moving aircraft. That range could be extended by up to 20 times using commonly-available DSLR lenses, Ford pointed out in the proposal's white paper (PDF). Light-based identification has its own problems, but at least it would be easier to implement than broadcasting by radio, an approach that could require industry standardization. It's a potentially cheap, efficient solution using equipment people already have (like smartphones), and given that consumers once again have to register their drones with the FAA, those 10-digit ID numbers will be more available.

"More than a dozen armed drones descended from an unknown location onto Russia's vast Hmeimim air base in northwestern Latakia province, the headquarters of Russia's military operations in Syria, and on the nearby Russian naval base at Tartus," The Washington Post reported. "Russia said that it shot down seven of the 13 drones and used electronic countermeasures to safely bring down the other six." And these drones appeared substantially less sophisticated and maneuverable than a DJI Phantom 4, the leading consumer drone. The National Academy notes that most of the counterstrategies that the Army has developed are "based on jamming radio frequency and GPS signals." The thinking was: Drones needed those information flows to navigate effectively.

A U.S. Air Force MQ-1B Predator unmanned aerial vehicle (UAV), carrying a Hellfire missile flies over an air base after flying a mission in the Persian Gulf region on January 7, 2016. A U.S. Air Force MQ-1B Predator unmanned aerial vehicle (UAV), carrying a Hellfire missile flies over an air base after flying a mission in the Persian Gulf region on January 7, 2016. A Google spokeswoman said the company provides its TensorFlow application programming interfaces, or APIs, to a pilot project with the Department of Defense to help automatically identify objects in unclassified data. APIs are software-based rules that let computer programs communicate. TensorFlow is a popular set of APIs and other tools for AI capabilities such as machine learning and computer vision.