Tag Archives: google
#433785 DeepMind’s Eerie Reimagination of the ...
If a recent project using Google’s DeepMind were a recipe, you would take a pair of AI systems, images of animals, and a whole lot of computing power. Mix it all together, and you’d get a series of imagined animals dreamed up by one of the AIs. A look through the research paper about the project—or this open Google Folder of images it produced—will likely lead you to agree that the results are a mix of impressive and downright eerie.
But the eerie factor doesn’t mean the project shouldn’t be considered a success and a step forward for future uses of AI.
From GAN To BigGAN
The team behind the project consists of Andrew Brock, a PhD student at Edinburgh Center for Robotics, and DeepMind intern and researcher Jeff Donahue and Karen Simonyan.
They used a so-called Generative Adversarial Network (GAN) to generate the images. In a GAN, two AI systems collaborate in a game-like manner. One AI produces images of an object or creature. The human equivalent would be drawing pictures of, for example, a dog—without necessarily knowing what a dog exactly looks like. Those images are then shown to the second AI, which has already been fed images of dogs. The second AI then tells the first one how far off its efforts were. The first one uses this information to improve its images. The two go back and forth in an iterative process, and the goal is for the first AI to become so good at creating images of dogs that the second can’t tell the difference between its creations and actual pictures of dogs.
The team was able to draw on Google’s vast vaults of computational power to create images of a quality and life-like nature that were beyond almost anything seen before. In part, this was achieved by feeding the GAN with more images than is usually the case. According to IFLScience, the standard is to feed about 64 images per subject into the GAN. In this case, the research team fed about 2,000 images per subject into the system, leading to it being nicknamed BigGAN.
Their results showed that feeding the system with more images and using masses of raw computer power markedly increased the GAN’s precision and ability to create life-like renditions of the subjects it was trained to reproduce.
“The main thing these models need is not algorithmic improvements, but computational ones. […] When you increase model capacity and you increase the number of images you show at every step, you get this twofold combined effect,” Andrew Brock told Fast Company.
The Power Drain
The team used 512 of Google’s AI-focused Tensor Processing Units (TPU) to generate 512-pixel images. Each experiment took between 24 and 48 hours to run.
That kind of computing power needs a lot of electricity. As artist and Innovator-In-Residence at the Library of Congress Jer Thorp tongue-in-cheek put it on Twitter: “The good news is that AI can now give you a more believable image of a plate of spaghetti. The bad news is that it used roughly enough energy to power Cleveland for the afternoon.”
Thorp added that a back-of-the-envelope calculation showed that the computations to produce the images would require about 27,000 square feet of solar panels to have adequate power.
BigGAN’s images have been hailed by researchers, with Oriol Vinyals, research scientist at DeepMind, rhetorically asking if these were the ‘Best GAN samples yet?’
However, they are still not perfect. The number of legs on a given creature is one example of where the BigGAN seemed to struggle. The system was good at recognizing that something like a spider has a lot of legs, but seemed unable to settle on how many ‘a lot’ was supposed to be. The same applied to dogs, especially if the images were supposed to show said dogs in motion.
Those eerie images are contrasted by other renditions that show such lifelike qualities that a human mind has a hard time identifying them as fake. Spaniels with lolling tongues, ocean scenery, and butterflies were all rendered with what looks like perfection. The same goes for an image of a hamburger that was good enough to make me stop writing because I suddenly needed lunch.
The Future Use Cases
GAN networks were first introduced in 2014, and given their relative youth, researchers and companies are still busy trying out possible use cases.
One possible use is image correction—making pixillated images clearer. Not only does this help your future holiday snaps, but it could be applied in industries such as space exploration. A team from the University of Michigan and the Max Planck Institute have developed a method for GAN networks to create images from text descriptions. At Berkeley, a research group has used GAN to create an interface that lets users change the shape, size, and design of objects, including a handbag.
For anyone who has seen a film like Wag the Dog or read 1984, the possibilities are also starkly alarming. GANs could, in other words, make fake news look more real than ever before.
For now, it seems that while not all GANs require the computational and electrical power of the BigGAN, there is still some way to reach these potential use cases. However, if there’s one lesson from Moore’s Law and exponential technology, it is that today’s technical roadblock quickly becomes tomorrow’s minor issue as technology progresses.
Image Credit: Ondrej Prosicky/Shutterstock Continue reading
Posted in Human Robots
#433689 The Rise of Dataism: A Threat to Freedom ...
What would happen if we made all of our data public—everything from wearables monitoring our biometrics, all the way to smartphones monitoring our location, our social media activity, and even our internet search history?
Would such insights into our lives simply provide companies and politicians with greater power to invade our privacy and manipulate us by using our psychological profiles against us?
A burgeoning new philosophy called dataism doesn’t think so.
In fact, this trending ideology believes that liberating the flow of data is the supreme value of the universe, and that it could be the key to unleashing the greatest scientific revolution in the history of humanity.
What Is Dataism?
First mentioned by David Brooks in his 2013 New York Times article “The Philosophy of Data,” dataism is an ethical system that has been most heavily explored and popularized by renowned historian, Yuval Noah Harari.
In his 2016 book Homo Deus, Harari described dataism as a new form of religion that celebrates the growing importance of big data.
Its core belief centers around the idea that the universe gives greater value and support to systems, individuals, and societies that contribute most heavily and efficiently to data processing. In an interview with Wired, Harari stated, “Humans were special and important because up until now they were the most sophisticated data processing system in the universe, but this is no longer the case.”
Now, big data and machine learning are proving themselves more sophisticated, and dataists believe we should hand over as much information and power to these algorithms as possible, allowing the free flow of data to unlock innovation and progress unlike anything we’ve ever seen before.
Pros: Progress and Personal Growth
When you let data run freely, it’s bound to be mixed and matched in new ways that inevitably spark progress. And as we enter the exponential future where every person is constantly connected and sharing their data, the potential for such collaborative epiphanies becomes even greater.
We can already see important increases in quality of life thanks to companies like Google. With Google Maps on your phone, your position is constantly updating on their servers. This information, combined with everyone else on the planet using a phone with Google Maps, allows your phone to inform you of traffic conditions. Based on the speed and location of nearby phones, Google can reroute you to less congested areas or help you avoid accidents. And since you trust that these algorithms have more data than you, you gladly hand over your power to them, following your GPS’s directions rather than your own.
We can do the same sort of thing with our bodies.
Imagine, for instance, a world where each person has biosensors in their bloodstreams—a not unlikely or distant possibility when considering diabetic people already wear insulin pumps that constantly monitor their blood sugar levels. And let’s assume this data was freely shared to the world.
Now imagine a virus like Zika or the Bird Flu breaks out. Thanks to this technology, the odd change in biodata coming from a particular region flags an artificial intelligence that feeds data to the CDC (Center for Disease Control and Prevention). Recognizing that a pandemic could be possible, AIs begin 3D printing vaccines on-demand, predicting the number of people who may be afflicted. When our personal AIs tell us the locations of the spreading epidemic and to take the vaccine it just delivered by drone to our homes, are we likely to follow its instructions? Almost certainly—and if so, it’s likely millions, if not billions, of lives will have been saved.
But to quickly create such vaccines, we’ll also need to liberate research.
Currently, universities and companies seeking to benefit humankind with medical solutions have to pay extensively to organize clinical trials and to find people who match their needs. But if all our biodata was freely aggregated, perhaps they could simply say “monitor all people living with cancer” to an AI, and thanks to the constant stream of data coming in from the world’s population, a machine learning program may easily be able to detect a pattern and create a cure.
As always in research, the more sample data you have, the higher the chance that such patterns will emerge. If data is flowing freely, then anyone in the world can suddenly decide they have a hunch they want to explore, and without having to spend months and months of time and money hunting down the data, they can simply test their hypothesis.
Whether garage tinkerers, at-home scientists, or PhD students—an abundance of free data allows for science to progress unhindered, each person able to operate without being slowed by lack of data. And any progress they make is immediately liberated, becoming free data shared with anyone else that may find a use for it.
Any individual with a curious passion would have the entire world’s data at their fingertips, empowering every one of us to become an expert in any subject that inspires us. Expertise we can then share back into the data stream—a positive feedback loop spearheading progress for the entirety of humanity’s knowledge.
Such exponential gains represent a dataism utopia.
Unfortunately, our current incentives and economy also show us the tragic failures of this model.
As Harari has pointed out, the rise of datism means that “humanism is now facing an existential challenge and the idea of ‘free will’ is under threat.”
Cons: Manipulation and Extortion
In 2017, The Economist declared that data was the most valuable resource on the planet—even more valuable than oil.
Perhaps this is because data is ‘priceless’: it represents understanding, and understanding represents control. And so, in the world of advertising and politics, having data on your consumers and voters gives you an incredible advantage.
This was evidenced by the Cambridge Analytica scandal, in which it’s believed that Donald Trump and the architects of Brexit leveraged users’ Facebook data to create psychological profiles that enabled them to manipulate the masses.
How powerful are these psychological models?
A team who built a model similar to that used by Cambridge Analytica said their model could understand someone as well as a coworker with access to only 10 Facebook likes. With 70 likes they could know them as well as a friend might, 150 likes to match their parents’ understanding, and at 300 likes they could even come to know someone better than their lovers. With more likes, they could even come to know someone better than that person knows themselves.
Proceeding With Caution
In a capitalist democracy, do we want businesses and politicians to know us better than we know ourselves?
In spite of the remarkable benefits that may result for our species by freely giving away our information, do we run the risk of that data being used to exploit and manipulate the masses towards a future without free will, where our daily lives are puppeteered by those who own our data?
It’s extremely possible.
And it’s for this reason that one of the most important conversations we’ll have as a species centers around data ownership: do we just give ownership of the data back to the users, allowing them to choose who to sell or freely give their data to? Or will that simply deter the entrepreneurial drive and cause all of the free services we use today, like Google Search and Facebook, to begin charging inaccessible prices? How much are we willing to pay for our freedom? And how much do we actually care?
If recent history has taught us anything, it’s that humans are willing to give up more privacy than they like to think. Fifteen years ago, it would have been crazy to suggest we’d all allow ourselves to be tracked by our cars, phones, and daily check-ins to our favorite neighborhood locations; but now most of us see it as a worthwhile trade for optimized commutes and dating. As we continue navigating that fine line between exploitation and innovation into a more technological future, what other trade-offs might we be willing to make?
Image Credit: graphicINmotion / Shutterstock.com Continue reading
Posted in Human Robots
#433506 MIT’s New Robot Taught Itself to Pick ...
Back in 2016, somewhere in a Google-owned warehouse, more than a dozen robotic arms sat for hours quietly grasping objects of various shapes and sizes. For hours on end, they taught themselves how to pick up and hold the items appropriately—mimicking the way a baby gradually learns to use its hands.
Now, scientists from MIT have made a new breakthrough in machine learning: their new system can not only teach itself to see and identify objects, but also understand how best to manipulate them.
This means that, armed with the new machine learning routine referred to as “dense object nets (DON),” the robot would be capable of picking up an object that it’s never seen before, or in an unfamiliar orientation, without resorting to trial and error—exactly as a human would.
The deceptively simple ability to dexterously manipulate objects with our hands is a huge part of why humans are the dominant species on the planet. We take it for granted. Hardware innovations like the Shadow Dexterous Hand have enabled robots to softly grip and manipulate delicate objects for many years, but the software required to control these precision-engineered machines in a range of circumstances has proved harder to develop.
This was not for want of trying. The Amazon Robotics Challenge offers millions of dollars in prizes (and potentially far more in contracts, as their $775m acquisition of Kiva Systems shows) for the best dexterous robot able to pick and package items in their warehouses. The lucrative dream of a fully-automated delivery system is missing this crucial ability.
Meanwhile, the Robocup@home challenge—an offshoot of the popular Robocup tournament for soccer-playing robots—aims to make everyone’s dream of having a robot butler a reality. The competition involves teams drilling their robots through simple household tasks that require social interaction or object manipulation, like helping to carry the shopping, sorting items onto a shelf, or guiding tourists around a museum.
Yet all of these endeavors have proved difficult; the tasks often have to be simplified to enable the robot to complete them at all. New or unexpected elements, such as those encountered in real life, more often than not throw the system entirely. Programming the robot’s every move in explicit detail is not a scalable solution: this can work in the highly-controlled world of the assembly line, but not in everyday life.
Computer vision is improving all the time. Neural networks, including those you train every time you prove that you’re not a robot with CAPTCHA, are getting better at sorting objects into categories, and identifying them based on sparse or incomplete data, such as when they are occluded, or in different lighting.
But many of these systems require enormous amounts of input data, which is impractical, slow to generate, and often needs to be laboriously categorized by humans. There are entirely new jobs that require people to label, categorize, and sift large bodies of data ready for supervised machine learning. This can make machine learning undemocratic. If you’re Google, you can make thousands of unwitting volunteers label your images for you with CAPTCHA. If you’re IBM, you can hire people to manually label that data. If you’re an individual or startup trying something new, however, you will struggle to access the vast troves of labeled data available to the bigger players.
This is why new systems that can potentially train themselves over time or that allow robots to deal with situations they’ve never seen before without mountains of labelled data are a holy grail in artificial intelligence. The work done by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) is part of a new wave of “self-supervised” machine learning systems—little of the data used was labeled by humans.
The robot first inspects the new object from multiple angles, building up a 3D picture of the object with its own coordinate system. This then allows the robotic arm to identify a particular feature on the object—such as a handle, or the tongue of a shoe—from various different angles, based on its relative distance to other grid points.
This is the real innovation: the new means of representing objects to grasp as mapped-out 3D objects, with grid points and subsections of their own. Rather than using a computer vision algorithm to identify a door handle, and then activating a door handle grasping subroutine, the DON system treats all objects by making these spatial maps before classifying or manipulating them, enabling it to deal with a greater range of objects than in other approaches.
“Many approaches to manipulation can’t identify specific parts of an object across the many orientations that object may encounter,” said PhD student Lucas Manuelli, who wrote a new paper about the system with lead author and fellow student Pete Florence, alongside MIT professor Russ Tedrake. “For example, existing algorithms would be unable to grasp a mug by its handle, especially if the mug could be in multiple orientations, like upright, or on its side.”
Class-specific descriptors, which can be applied to the object features, can allow the robot arm to identify a mug, find the handle, and pick the mug up appropriately. Object-specific descriptors allow the robot arm to select a particular mug from a group of similar items. I’m already dreaming of a robot butler reliably picking my favourite mug when it serves me coffee in the morning.
Google’s robot arm-y was an attempt to develop a general grasping algorithm: one that could identify, categorize, and appropriately grip as many items as possible. This requires a great deal of training time and data, which is why Google parallelized their project by having 14 robot arms feed data into a single neural network brain: even then, the algorithm may fail with highly specific tasks. Specialist grasping algorithms might require less training if they’re limited to specific objects, but then your software is useless for general tasks.
As the roboticists noted, their system, with its ability to identify parts of an object rather than just a single object, is better suited to specific tasks, such as “grasp the racquet by the handle,” than Amazon Robotics Challenge robots, which identify whole objects by segmenting an image.
This work is small-scale at present. It has been tested with a few classes of objects, including shoes, hats, and mugs. Yet the use of these dense object nets as a way for robots to represent and manipulate new objects may well be another step towards the ultimate goal of generalized automation: a robot capable of performing every task a person can. If that point is reached, the question that will remain is how to cope with being obsolete.
Image Credit: Tom Buehler/CSAIL Continue reading
Posted in Human Robots
#433486 This AI Predicts Obesity ...
A research team at the University of Washington has trained an artificial intelligence system to spot obesity—all the way from space. The system used a convolutional neural network (CNN) to analyze 150,000 satellite images and look for correlations between the physical makeup of a neighborhood and the prevalence of obesity.
The team’s results, presented in JAMA Network Open, showed that features of a given neighborhood could explain close to two-thirds (64.8 percent) of the variance in obesity. Researchers found that analyzing satellite data could help increase understanding of the link between peoples’ environment and obesity prevalence. The next step would be to make corresponding structural changes in the way neighborhoods are built to encourage physical activity and better health.
Training AI to Spot Obesity
Convolutional neural networks (CNNs) are particularly adept at image analysis, object recognition, and identifying special hierarchies in large datasets.
Prior to analyzing 150,000 high-resolution satellite images of Bellevue, Seattle, Tacoma, Los Angeles, Memphis, and San Antonio, the researchers trained the CNN on 1.2 million images from the ImageNet database. The categorizations were correlated with obesity prevalence estimates for the six urban areas from census tracts gathered by the 500 Cities project.
The system was able to identify the presence of certain features that increased likelihood of obesity in a given area. Some of these features included tightly–packed houses, being close to roadways, and living in neighborhoods with a lack of greenery.
Visualization of features identified by the convolutional neural network (CNN) model. The images on the left column are satellite images taken from Google Static Maps API (application programming interface). Images in the middle and right columns are activation maps taken from the second convolutional layer of VGG-CNN-F network after forward pass of the respective satellite images through the network. From Google Static Maps API, DigitalGlobe, US Geological Survey (accessed July 2017). Credit: JAMA Network Open
Your Surroundings Are Key
In their discussion of the findings, the researchers stressed that there are limitations to the conclusions that can be drawn from the AI’s results. For example, socio-economic factors like income likely play a major role for obesity prevalence in a given geographic area.
However, the study concluded that the AI-powered analysis showed the prevalence of specific man-made features in neighborhoods consistently correlating with obesity prevalence and not necessarily correlating with socioeconomic status.
The system’s success rates varied between studied cities, with Memphis being the highest (73.3 percent) and Seattle being the lowest (55.8 percent).
AI Takes To the Sky
Around a third of the US population is categorized as obese. Obesity is linked to a number of health-related issues, and the AI-generated results could potentially help improve city planning and better target campaigns to limit obesity.
The study is one of the latest of a growing list that uses AI to analyze images and extrapolate insights.
A team at Stanford University has used a CNN to predict poverty via satellite imagery, assisting governments and NGOs to better target their efforts. A combination of the public Automatic Identification System for shipping, satellite imagery, and Google’s AI has proven able to identify illegal fishing activity. Researchers have even been able to use AI and Google Street View to predict what party a given city will vote for, based on what cars are parked on the streets.
In each case, the AI systems have been able to look at volumes of data about our world and surroundings that are beyond the capabilities of humans and extrapolate new insights. If one were to moralize about the good and bad sides of AI (new opportunities vs. potential job losses, for example) it could seem that it comes down to what we ask AI systems to look at—and what questions we ask of them.
Image Credit: Ocean Biology Processing Group at NASA’s Goddard Space Flight Center Continue reading
Posted in Human Robots