This Google Scientist Learned an Interesting Way to Teach Neural Networks

first_img Watch: Dolphin Leaps Feet Away From Unsuspecting SurferNASA Says 2 Asteroids Will Safely Fly By Earth This Weekend Stay on target Artificial intelligence is as trippy as it gets when it comes to thinking about how machines “think.” But what about when you consider how machines teach and learn from each other? That’s something one Ian Goodfellow dreamt up, as he discovered a way for neural networks to improve themselves and each other by actually teaming up and working together, as amazing and far-fetched as that sounds. As part of the Google Brain Team now, Goodfellow is all about improving the intelligence of computers and removing the need to hand-feed them data to get them started learning on a task.Before his discovery, Goodfellow was previously studying how neural networks can teach themselves real concepts without the assistance of humans, which is a super strange idea when you think about it, but certainly not out of the realm of possibility. While out at the pub one day, Goodfellow was thinking about how neural networks could conceivably work together on one thing. With one network learning about a particular set of data and generating examples while the other would try and determine whether the example was real or fake and would allow the first network to correct and improve itself going forward.After this flash of genius, Goodfellow went home and coded what’s known as a “generative adversarial network,” which would feature dueling neural networks working to perform impressive tasks, by improving the resolution of a pixelated image, create realistic fake photos, use certain filters, and more. All this, apparently from a stint at the pub. Sure, when Goodfellow goes to a pub or to a party he comes back home with million-dollar ideas, and when we come back home we get saddled with hangovers, right? That’s a bum deal. But hey, he’s certainly onto something, and it’ll be interesting to see where this goes.Let us know what you like about Geek by taking our survey.last_img read more

ULA Rocket Carrying US Spy Satellite Is Launched Into Orbit

first_img A United Launch Alliance (ULA) rocket carrying a U.S. spy satellite was successfully launched into orbit on Saturday.ULA announced the news in a press release, and said the Delta IV Heavy rocket, which contained an important payload for the National Reconnaissance Office (NRO) dubbed NROL-71, blasted off from ULA’s Space Launch Complex in California at 11:10 a.m. PST. With this rocket, the U.S. government can elevate national defense efforts in space.The Delta IV Heavy is a heavy lift launch vehicle that delivered high-priority missions for NASA, the U.S. Air Force, and the NRO. It can carry more than 14,000 pounds directly to geosynchronous orbit, making it a top rocket for launching government satellites into space.The mission initially launched on a Delta IV Heavy, which had three common booster cores that were powered by an Aerojet Rocketdyne RS-68A liquid hydrogen/liquid oxygen engine. The second stage of the mission involved an AR RL10B-2 liquid hydrogen/liquid oxygen engine.NROL-71 was ULA’s first launch of the year, and the company’s 132nd successful launch since 2006. ULA’s next scheduled launch is the U.S. Air Force’s WGS-10 mission. The launch, which will use another powerful Delta IV rocket, is set to take place on March 13 at Space Launch Complex-37 in Florida.More on Geek.com:Japan Sends Satellite Into Space to Deliver Artifical Meteor ShowerSaturn’s Rings Are Younger Than the Planet They OrbitHubble Space Telescope Should Last At Least 5 More Years Stay on target Watch: Dolphin Leaps Feet Away From Unsuspecting SurferNASA Says 2 Asteroids Will Safely Fly By Earth This Weekend last_img read more