Gamers get a chance to battle an AI on the QT. Plus: Robo-marines, and fisticuffs over facial recognition in Detroit
Rapid-fire summary of machine-learning news
Roundup Hello, here’s a few announcements from the world of machine learning beyond what we’ve already covered this week.
AlphaStar is coming out to play: AlphaStar, the StarCraft II-playing bot built by DeepMind researchers, will be facing human players in a series of 1v1 games online.
StarCraft II players can enter the open competition league set up by Blizzard Entertainment, the creators of the popular battle strategy game, and opt-in to play against AlphaStar. But nobody will know if they’re facing the bot, however, because it’ll be entering the matches anonymously.
Characters in the StarCraft II are from three species: Terran, Zerg or Protoss. AlphaStar will be able to play as and against players from these species, and will only dip in for a small number of games. DeepMind want to see how its agent fares against other players for research purposes, and is expected to post the results in a research paper in the future.
AlphaStar played and won against two top human players earlier this year. But it’s not invincible, just uses weird strategies that have probably never been seen before. The games played will not be used to train AlphaStar further, it’s more for testing.
Only players based in Europe will be allowed to against AlphaStar for this experiment.
Yikes, this facial recog meeting got out of hand: A police commissioner was thrown out and arrested during a facial recognition hearing held by the Board of Police Commissioners in Detroit this week.
The footage captured by The Detroit News shows an argument unfolding between police commissioner Willie Brown and Lisa Carter, who had just been sworn in as the new chairwoman of the board at the meeting.
Brown, apparently, asked Carter what she would do differently compared to the previous leader of the Board of Police Commissioners, and was told he was “out of order”. Brown then proceeded to talk and Carter ordered officers to remove him from the meeting. The pair continued yelling at one another, and Brown was reprimanded and pushed onto the floor by officers as he sat on his chair, whilst onlookers filmed the whole debacle.
Facial recognition is a tricky subject, especially in Detroit since the city has one of the largest percentage of black people in its population. The technology has been proven to perform worse for women and people with darker skin and there are concerns that the biases in the technology will have dire impacts when used by law enforcement in Detroit.
Isaac Robinson, a member of the Michigan House of Representatives, introduced a house bill calling for a moratorium that would ban Michigan state government from using the technology for five years. It directly challenged the Detroit Board of Police Commissioners, who recently voted to use the technology on traffic cameras.
“It is completely unconscionable to ask our families to sacrifice our freedoms for a policy that has the potential to put so many residents at risk. My bill will combat this by preventing law enforcement from using inherently flawed facial recognition software to analyze and identify people. We must reverse the dangerous trend of using technology that intrudes on our privacy and violates our Fourth Amendment rights,” Robinson said.
The current skirmish between the government and police is over whether facial recognition will be used over real time video footage or still photos, apparently. Detroit have been using the technology for over a year. The meeting was held after the Board of Police Commissioners asked to hold off on a vote that would affect what kind of applications of facial recognition technology were “acceptable” in June.
DARPA testing out how soldiers can use AI during battles: US Marines have been testing two systems that allow fighters to cooperate with unmanned vehicles on ground and in the air.
The systems developed by Lockheed Martin allow soldiers to surveil distant locations. The information from a four-wheeled robot and a drone were beamed to a team of Marines, who were wearing vests kitted out with sensors, radars, and cameras. They could then control what parts of the battleground they wanted to view from the eyes of the robot or drone with a few taps of a screen.
“A human would be involved in any lethal action,” Liutenant Colonel Phil Root, the deputy director of the Defense Sciences Office at DARPA, the US military research arm said. “But we’re establishing superior situational awareness through sufficient input and AI, and then the ability to do something about it at fast time scales.”
The experiment was carried out at the Air Ground Combat Center in Twentynine Palms, California. “The feedback process, in conjunction with the actual experimentation, gives the Marines the ability to use the technology and start seeing what it can do and, more specifically, what it can’t do,” said Andrew Hall, Staff Sergeant with the Marine Corps Tactics and Operations Group working with DARPA.
DARPA will conduct a second phase of the experiment to test whether upgraded version of both AI systems will be able to operate for five hours and longer later this year. ®