Plumes of smoke drifted up from a fire steadily taking over a 30-acre prairie at Cedar Creek Ecosystem Science Reserve, north of the Twin Cities. Amid the haze, five black drones zipped around.

More than 150 feet below the flying robots, research student Nikil Krishnakumar raised the controller in the air. The work has been published on the arXiv preprint server.

“It’s all autonomous now,” he said. “I’m not doing anything.”

The aerial robotic team’s mission: examine the smoke from the prescribed burn and send the data to a computer on the ground. The computer then analyzes the smoke data to understand the fire’s flow patterns, Krishnakumar said.

The University of Minnesota project is the latest research into using artificial intelligence to detect and track wildfires. The work has become more urgent as climate change is expected to make wildfires, like those that devastated Manitoba this summer, larger and more frequent.

NOAA’s Next-Generation Fire System consists of two satellites 22,000 miles above the equator that detect new sources of heat and report them to local National Weather Service stations and its online dashboard. Earlier this year, the satellites were credited with spotting 19 fires in Oklahoma and preventing $850 million in structure and property damage, according to the agency.

In Minnesota, Xcel has installed tower-mounted, AI-equipped high-definition cameras near power lines in Mankato and Clear Lake. Thirty-six more are planned. When a fire is detected, local fire departments are notified.

Krishnakumar and other members of the U’s research team performed their 11th trial at the U’s field station in East Bethel on Friday, with notable improvements from their previous attempts.

The first-generation drones crashed several times during previous field tests, Krishnakumar said. The team upgraded sensors for better data collecting and autonomous steering, and improved the drones’ propulsion by making them bigger and fitting them with better propellers.

“The big picture is one day these drones can be used to understand where the wildfires go, how they behave and to perform large-scale surveillance of wildfires,” Krishnakumar said. “The major challenge we’re trying to understand is how far these smoke particles can be transported and the altitude at which they can go.”

Understanding the behavior of particles like embers can help firefighters prevent wildfires from spreading, said Yue Weng, another researcher on the team.

Though the project has a way to go before it can be used for large-scale wildfires, the research represents a significant step toward using fully autonomous drone systems for emergency response and scientific research missions, said Jiarong Hong, professor at the University of Minnesota’s Department of Mechanical Engineering.

This year, 1,200 wildfires have been recorded in Minnesota so far, according to the state Department of Natural Resources. On a smaller scale, the technology could also be used to better manage prescribed burns, Hong said. Between 2012 and 2021, prescribed burns that went out of control caused 43 wildfires nationwide, according to the Associated Press.

“To characterize and measure particle transport in the real field is very challenging. Traditionally, people do small-scale lab experiments and study this at a fundamental level,” Hong said. “Such an experiment doesn’t capture the complexity involved in the real field environment.”

Smoke changes direction with the wind. Deploying multiple drones—with one at the center managing the four around it—enables them to navigate in the air without human intervention, Hong said.

The 11-pound drones were custom-built by the students to autonomously collect particle data. Future improvements to the project include collecting more data and extending the battery life of the drones. The drones are currently able to operate in the air for about 25 minutes, less in colder temperatures, Hong said.

“We have drones flying out at different heights, so we can actually measure the particle composition at different elevations at the same time,” Hong said.

“Particles are in a very irregular shape and some of them are porous and have varying levels of density. But we have been able to characterize their morphology and shape for the very first time.”

2025 The Minnesota Star Tribune. Distributed by Tribune Content Agency, LLC

By the standards of the San Francisco Bay Area’s hard left, Casey Goonan’s crimes were unremarkable. A police SUV partially burned by an incendiary device on UC Berkeley’s campus. A planter of shrubs lit on fire after Goonan unsuccessfully tried to smash a glass office window and throw a firebomb into the federal building in downtown Oakland.

But thanks to a series of communiques where Goonan claimed to have carried out the summer 2024 attacks in solidarity with Hamas and the East Bay native’s anarchist beliefs, federal prosecutors claimed Goonan “intended to promote” terrorism on top of a felony count for using an incendiary device. Goonan’s original charges notably did not contain terrorism counts. In late September, US District Court Judge Jeffrey White sentenced Goonan, whom they called “a domestic terrorist” during the hearing, to 19 and a half years in prison plus 15 years probation. Prosecutors also asked that he be sent to the Bureau of Prisons facility that contains a Communications Management Units, a highly restrictive assignment reserved for what the government claims are “extremist” inmates with terrorism-related offenses or affiliations.

Although Goonan’s case began under the Biden Administration, it offers a glimpse of the approach the Department of Justice may take in President Donald Trump’s forthcoming offensive against the “left,” formalized in late September in National Security Presidential Memorandum 7 (NSPM-7), an executive order targeting anti-fascist beliefs, opposition towards Immigrations and Customs Enforcement raids, and criticism of capitalism and Christianity as potential “indicators of terrorism.”

In addition to Goonan’s purported admiration for Hamas—a designated terrorist organization since 1997—and cofounding of True Leap, a tiny Anarchist publisher, the 35-year-old doctorate in African-American Studies’ biography includes another trait being targeted by the Trump administration and its allies: Goonan identifies as a transgender person. While NPSM-7 cites “extremism migration, race, and gender” as an indicator of “this pattern of violent and terroristic tendencies,” the Heritage Foundation has attempted to link gender-fluid identity to mass shootings and is urging the FBI to create a new, specious domestic terrorism classification of “Transgender Ideology-Inspired Violent Extremism,” or TIVE.

The executive order, meanwhile, directs the American security state’s sprawling post-9/11 counterterrorism apparatus to be reoriented away from neo-Nazis, Proud Boys, white nationalists, Christian nationalists, and other extreme right-wing actors that have been overwhelmingly responsible for the majority of political violence in the past few decades, and towards opponents of ICE, anti-fascists, and the administration writ large. Along with potentially violent actors, NSPM-7 instructs federal law enforcement to scrutinize nonprofit groups and philanthropic foundations involved in funding organizations that espouse amorphous ideologies, from “support for the overthrow of the United States Government” to expressing “hostility towards those who hold traditional American views on family, religion, and morality.”

“NSPM-7 is the natural culmination of ‘radicalization theory’ as the basis for the American approach to counterterrorism,” says Mike German, a retired FBI agent who spent years infiltrating violent white supremacist groups and quit the Bureau in response to its post-9/11 shift in terrorism strategy. German explored radicalization theory’s trajectory in his 2019 book, Disrupt, Discredit and Divide: How the New FBI Damages Democracy.

Despite ongoing speculation around an investment bubble that may be set to burst, artificial intelligence (AI) technology is here to stay. And while an over-inflated market may exist at the level of the suppliers, AI is well-developed and has a firm foothold among organisations of all sizes.

But AI workloads place specific demands on IT infrastructure and on storage in particular. Data volumes can start big and then balloon, in particular during training phases as data is vectorised and checkpoints are created. Meanwhile, data must be curated, gathered and managed throughout its lifecycle.

In this article, we look at the key characteristics of AI workloads, the particular demands of training and inference on storage I/O, throughput and capacity, whether to choose object or file storage, and the storage requirements of agentic AI.