What Do a Robot's Dreams Look Like? Google Found Out

They may look odd, but it’s all part of Google’s plan to solve a huge issue in machine learning: recognizing objects in images.

Google's artificial neural networks produce some trippy images thanks to its Deep Dream program (photo credit: Michael Tyka/Google)
Google's artificial neural networks produce some trippy images thanks to its Deep Dream program (photo credit: Michael Tyka/Google)

When Google asked its neural network to dream, the machine begin to generating some pretty wild images. They may look odd, but it’s all part of Google’s plan to solve a huge issue in machine learning: recognizing objects in images.

To be clear, Google’s software engineers didn’t ask a computer to dream, but they did ask its neural network to alter the images based on an original photo they fed into it, by applying layers. This was all part of their Deep Dream program.
 
The purpose was to make it better at finding patterns, which computers are none too good at. So, engineers started by “teaching” the neural network to recognize certain objects by giving it 1.2 million images, complete with object classifications the computer could understand.

These classifications allowed Google’s AI to learn to detect the different qualities of certain objects in an image, like a dog and a fork. But Google’s engineers wanted to go one step further, which is where Deep Dream comes in, which allowed the neural network to add those hallucinogenic qualities to images




Google wanted to make its neural network better at detection to the point where it could pick out other objects in an image that may not contain that object (think of it as seeing the outline of a dog in the clouds). Deep Dream gave the computer the ability to change the rules and parameters of the images, which in turn allowed Google’s AI to recognize objects the images didn’t necessarily contain. So, an image might contain an image of a foot, but when it examined a few pixels of that image, it may have seen the outline of what looked like a dog’s nose.

So, when researchers began to ask its neural network to tell them what other objects they might be able to see in an image of a mountain, tree, or plant, it came up with these interpretations:


(Photo Credit: Michael Tyka/Google)

“The techniques presented here help us understand and visualize how neural networks are able to carry out difficult classification tasks, improve network architecture, and check what the network has learned during training,” software engineers Alexander Mordvintsev and Christopher Olah, and intern Mike Tyka wrote in a post about Deep Dream. “It also makes us wonder whether neural networks could become a tool for artists—a new way to remix visual concepts—or perhaps even shed a little light on the roots of the creative process in general.”

Just for fun, Google has opened up the tool to the public and you can generate your own Deep Dream art here: deepdreamgenerator.com

the-future-of-machine-learning


Massive 'Darth Vader' isopod found lurking in the Indian Ocean

The father of all giant sea bugs was recently discovered off the coast of Java.

A close up of Bathynomus raksasa

SJADE 2018
Surprising Science
  • A new species of isopod with a resemblance to a certain Sith lord was just discovered.
  • It is the first known giant isopod from the Indian Ocean.
  • The finding extends the list of giant isopods even further.
Keep reading Show less

Astronomers find more than 100,000 "stellar nurseries"

Every star we can see, including our sun, was born in one of these violent clouds.

Credit: NASA / ESA via Getty Images
Surprising Science

This article was originally published on our sister site, Freethink.

An international team of astronomers has conducted the biggest survey of stellar nurseries to date, charting more than 100,000 star-birthing regions across our corner of the universe.

Stellar nurseries: Outer space is filled with clouds of dust and gas called nebulae. In some of these nebulae, gravity will pull the dust and gas into clumps that eventually get so big, they collapse on themselves — and a star is born.

These star-birthing nebulae are known as stellar nurseries.

The challenge: Stars are a key part of the universe — they lead to the formation of planets and produce the elements needed to create life as we know it. A better understanding of stars, then, means a better understanding of the universe — but there's still a lot we don't know about star formation.

This is partly because it's hard to see what's going on in stellar nurseries — the clouds of dust obscure optical telescopes' view — and also because there are just so many of them that it's hard to know what the average nursery is like.

The survey: The astronomers conducted their survey of stellar nurseries using the massive ALMA telescope array in Chile. Because ALMA is a radio telescope, it captures the radio waves emanating from celestial objects, rather than the light.

"The new thing ... is that we can use ALMA to take pictures of many galaxies, and these pictures are as sharp and detailed as those taken by optical telescopes," Jiayi Sun, an Ohio State University (OSU) researcher, said in a press release.

"This just hasn't been possible before."

Over the course of the five-year survey, the group was able to chart more than 100,000 stellar nurseries across more than 90 nearby galaxies, expanding the amount of available data on the celestial objects tenfold, according to OSU researcher Adam Leroy.

New insights: The survey is already yielding new insights into stellar nurseries, including the fact that they appear to be more diverse than previously thought.

"For a long time, conventional wisdom among astronomers was that all stellar nurseries looked more or less the same," Sun said. "But with this survey we can see that this is really not the case."

"While there are some similarities, the nature and appearance of these nurseries change within and among galaxies," he continued, "just like cities or trees may vary in important ways as you go from place to place across the world."

Astronomers have also learned from the survey that stellar nurseries aren't particularly efficient at producing stars and tend to live for only 10 to 30 million years, which isn't very long on a universal scale.

Looking ahead: Data from the survey is now publicly available, so expect to see other researchers using it to make their own observations about stellar nurseries in the future.

"We have an incredible dataset here that will continue to be useful," Leroy said. "This is really a new view of galaxies and we expect to be learning from it for years to come."

Protecting space stations from deadly space debris

Tiny specks of space debris can move faster than bullets and cause way more damage. Cleaning it up is imperative.

Videos
  • NASA estimates that more than 500,000 pieces of space trash larger than a marble are currently in orbit. Estimates exceed 128 million pieces when factoring in smaller pieces from collisions. At 17,500 MPH, even a paint chip can cause serious damage.
  • To prevent this untrackable space debris from taking out satellites and putting astronauts in danger, scientists have been working on ways to retrieve large objects before they collide and create more problems.
  • The team at Clearspace, in collaboration with the European Space Agency, is on a mission to capture one such object using an autonomous spacecraft with claw-like arms. It's an expensive and very tricky mission, but one that could have a major impact on the future of space exploration.

This is the first episode of Just Might Work, an original series by Freethink, focused on surprising solutions to our biggest problems.

Catch more Just Might Work episodes on their channel:
https://www.freethink.com/shows/just-might-work

Personal Growth

Meet the worm with a jaw of metal

Metal-like materials have been discovered in a very strange place.

Quantcast