Recent research from Stanford and Google has made the worst nightmare of some concerned with artificial intelligence (AI) all the more real. A machine learning agent was caught cheating by hiding information in “a nearly imperceptible, high-frequency signal.”
Clever, but also creepy.
The agent was instructed to turn aerial images into street maps and back again as part of research to improve Google’s process of turning satellite images into the widely used and relied upon Google Maps. The process involves CycleGAN, “a neural network that learns to transform images of type X and Y into one another, as efficiently yet accurately as possible.” Though the agent was performing this task quite well, it quickly became apparent that it was performing the task too well.
While the agent was supposed to be turning a satellite image into a street map and then back into a satellite image again using only the street map, details missing only on the street map version showed the research team that this wasn’t actually occurring. As seen in the example below, details such as skylights can be seen on both aerial versions of the map despite being absent on the street map (center).
What this ultimately means is that the agent wasn’t acting on its own accord or for malicious purposes, it was simply doing exactly what it was told to do – a “problem” with computers that has existed since their inception.
In order to figure out exactly what the machine learning agent was doing, the research team delved into the data generated by the AI. It turns out, because the agent was being graded on how close the resulting aerial map matched the original as well as the clarity of the generated street map, it didn’t learn how to make one using the other. What it did do was learn how to encode features present on the original satellite map into the noise of the street map via tiny color changes undetectable by the human eye. The agent would then read the secret messages it left for itself when creating the third image in the series.
The agent excelled to such an extent that it was actually able to encode any satellite map into any street map without paying attention to the street map. This means, the data to create the resulting aerial image was encoded in features already present on a street map, even on a completely different street map. An example of this can be seen below.
Encoding data into images, a process called steganography, certainly isn’t anything new. The first recorded use of steganography dates all the way back to 1499 when Johannes Trithemius wrote a book on cryptography and steganography, called Steganographia. Today the process is frequently used when adding watermarks images or metadata to image files. What is new is that a machine learning agent was able to create its own steganographic method to avoid actually learning to do a task.
So are computers getting smarter? Or are they just really good at cheating in a way that humans can’t easily detect? Chances are it’s the latter and we don’t have to worry about our worst AI nightmares coming true anytime soon. What we can deduce, however, is that for AI to do exactly what we want it to, we have to be really specific with our instructions. Especially when we’re instructing it to solve problems quickly and efficiently.
Dolphin Swims Through Louisiana Neighborhood in Aftermath of Hurricane Ida
A Louisiana family was shocked to find a dolphin swimming through their neighborhood in the aftermath of Hurricane Ida.
Amanda Huling and her family were assessing the damage to their neighborhood in Slidell, Louisiana, when they noticed the dolphin swimming through the inundated suburban landscape.
In video shot by Huling, the marine mammal’s dorsal fin can be seen emerging from the water.
“The dolphin was still there as of last night but I am in contact with an organization who is going to be rescuing it within the next few days if it is still there,” Huling told FOX 35.
Ida slammed into the coast of Louisiana this past weekend. The Category 4 hurricane ravaged the power grid of the region, plunging residents of New Orleans and upwards of 1 million homes and businesses in Louisiana and Mississippi into the dark for an indefinite period of time.
Officials have warned that the damage has been so extensive that it could take weeks to repair the power grid, reports Associated Press.
Also in Slidell, a 71-year-old man was attacked by an alligator over the weekend while he was in his flooded shed. The man went missing and is assumed dead, reports WDSU.
Internet users began growing weary last year about the steady stream of stories belonging to a “nature is healing” genre, as people stayed indoors and stories emerged about animals taking back their environs be it in the sea or in our suburbs.
However, these latest events are the surreal realities of a world in which extreme weather events are fast becoming the new normal – disrupting our lives in sometimes predictable, and occasionally shocking and surreal, ways.
Mom in LA Suburbs Fights Off Mountain Lion With Bare Hands, Rescues 5-Year-Old Son
A mother in Southern California is being hailed as a hero after rescuing her five-year-old son from an attacking mountain lion.
The little boy was playing outside his home in Calabasas, a city lying west of Los Angeles in the Santa Monica Mountains, when the large cat pounced on him.
The 65-pound (30 kg) mountain lion dragged the boy about 45 yards across the front lawn before the mother acted fast, running out and striking the creature with her bare hands and forcing it to free her son.
“The true hero of this story is his mom because she absolutely saved her son’s life,” California Department of Fish and Wildlife spokesman Captain Patrick Foy told Associated Press on Saturday.
“She ran out of the house and started punching and striking the mountain lion with her bare hands and got him off her son,” Foy added.
The boy sustained significant injuries to his head, neck and upper torso, but is now in stable condition at a hospital in Los Angeles, according to authorities.
The mountain lion was later located and killed by an officer with the California Department of Fish and Wildlife, who found the big cat crouching in the bushes with its “ears back and hissing” at the officer shortly after he arrived at the property.
“Due to its behavior and proximity to the attack, the warden believed it was likely the attacking lion and to protect public safety shot and killed it on sight,” the wildlife department noted in its statement.
The mountain lion attack is the first such attack on a human in Los Angeles County since 1995, according to Fish and Wildlife.
The Santa Monica Mountains is a biodiverse region teeming with wildlife such as large raptors, mountain lions, bears, coyote, deer, lizards, and snakes. However, their numbers have rapidly faded in recent years, causing local wildlife authorities to find new ways to manage the region’s endemic species.
Video Shows Taliban Taking Joyride in Captured US Blackhawk Helicopter
The rapid fall of Kabul to the Taliban has resulted in a number of surreal sights – from footage of the Islamist group’s fighters exercising at a presidential gym to clips of combatants having a great time on bumper cars at the local fun park.
However, a new video of Taliban members seemingly testing their skills in the cockpit of a commandeered UH-60 Black Hawk helicopter shows the chilling extent to which U.S. wares have fallen into the hands of a group it spent trillions of dollars, and exhaustive resources, to stamp out.
In the new video, shared on Twitter, the front-line utility helicopter can be seen taxiing on the ground at Kandahar Airport in southeastern Afghanistan, moving along the tarmac. It is unclear who exactly was sitting in the cockpit, and the Black Hawk cannot be seen taking off or flying.
It is unlikely that the Taliban have any combatants who are sufficiently trained to fly a UH-60 Black Hawk.
The helicopter, which carries a $6 million price tag, is just a small part of the massive haul that fell into the militant group’s hands after the country’s central government seemingly evaporated on Aug. 14 amid the withdrawal of U.S. and coalition troops.
Some 200,000 firearms, 20,000 Humvees and hundreds of aircraft financed by Washington for the now-defunct Afghan Army are believed to be in the possession of the Taliban.
The firearms include M24 sniper rifles, M18 assault weapons, anti-tank missiles, automatic grenade launchers, rocket-propelled grenades and mortars.
Taliban fighters in the elite Badri 313 Brigade have been seen in propaganda images showing off in uniforms and wielding weaponry meant for the special forces units of the Afghan Army.
The U.S. is known to have purchased 42,000 light tactical vehicles, 9,000 medium tactical vehicles and over 22,000 Humvees between 2003 and 2016.
The White House remains unclear on how much weaponry has fallen into Taliban hands, with National Security Adviser Jake Sullivan admitting last week that the U.S. lacks a “clear picture of just how much missing $83 billion of military inventory” the group has.