Robots Learn How to Make Friends and Influence People

Robots Learn How to Make Friends and Influence People

• 
  by Will Knight
  
• 
  May 17, 2016
  

Thankfully, researchers have developed an algorithm that could give robots the ability to deftly maneuver through spaces packed with unpredictable humans.

“The challenge is how to program these devices to respect human social conventions,” says Silvio Savarese at Stanford University.

Savarese and colleagues developed a computer-vision algorithm that predicts the movement of people in a busy space. They trained a deep-learning neural network using several publicly available data sets containing video of people moving around crowded areas. And they found their software to be better at predicting peoples’ movements than existing approaches for several of those data sets.

Savarese's team is testing its algorithm on a mobile robot called JackRabbot developed at Stanford. The two-wheeled robot, which is equipped with cameras, range sensors, and GPS, will explore busy indoor and outdoor spaces to test the approach in real situations.

At the moment, the most notable example of robots interacting directly with members of the public is Google’s self-driving vehicles. The company has acknowledged that its cars, while predominantly safe, have indirectly contributed to accidents due to a failure to understand the social norms of the road (“Google’s Self-Driving Car Chief Defends Safety Record”). As robots begin to proliferate into settings like shops and offices, awkward run-ins could become more common.

“The first problem is to understand the mostly unstated rules that people follow,” Savarese says. “How do people behave in crowds? How do they share resources, like sidewalks, parking spots? When should a person (or a robot) take its turn?”

A startup called Starship Technologies, which makes robots that deliver packages, is also working on this problem. The company has been testing its robots at several locations in the U.S. and the U.K., and besides dealing with uneven sidewalks and navigating around random obstacles, encounters with pedestrians pose the biggest challenge.

“Our robots have now come into contact with over 230,000 people around the world,” says Henry Harris-Burland, a spokesperson for Starship. Engineers at the company monitor the robots remotely as they go about mock deliveries. “Social acceptance is a core focus at the moment,” he says.

Jodi Forlizzi, at Carnegie Mellon University’s Human Computer Interaction Institute, says the Stanford algorithm adds to other research aimed at making robot behavior more humanlike. “Much research in human-robot interaction has looked at whether we can replicate the norms of human social interaction,” she says.

That goes way beyond just predicting a person’s movement. Forlizzi’s own research has involved trying to get robots to move around spaces in such a way that they form natural-seeming clusters with people. She says there is a definite need to teach robots how to blend in.

“There’s a whole class of robots that will be working with people and close to people also, so we need to understand how they should behave,” Forlizzi says.

SA

When Will Computers Have Common Sense? Ask Facebook

• 
  By Larry Greenemeier on June 20, 2016
  

Facebook is well known for its early and increasing use of artificial intelligence. The social media site uses AI to pinpoint its billion-plus users’ individual interests and tailor content accordingly by automatically scanning their newsfeeds, identifying people in photos and targeting them with precision ads. And now behind the scenes the social network’s AI researchers are trying to take this technology to the next level—from pure data-crunching logic to a nuanced form of “common sense” rivaling that of humans.

AI already lets machines do things like recognize faces and act as virtual assistants that can track down info on the Web for smartphone users. But to perform even these basic tasks the underlying learning algorithms rely on computer programs written by humans to feed them massive amounts of training data, a process known as machine learning. For machines to truly have common sense—to be able to figure out how the world works and make reasonable decisions based on that knowledge—they must be able to teach themselves without human supervision. Though this will not happen on a significant scale anytime soon, researchers are taking steps in that direction. In a blog posted Monday, for example, Facebook director of AI Yann LeCun and research engineer Soumith Chintala describe efforts at unsupervised machine learning through a technique called adversarial training.

This approach consists of two artificial neural networks, so called because they use algorithms designed to help them function a little like a human brain. A “generator” network creates images based on random data that it is fed. Researchers train the second “discriminator” network through machine learning to be able to tell the difference between a real image and a data file containing nonsensical patterns of shapes and colors. The discriminator then analyzes a series of files, some from a database of real images and others created by the generator network. Initially, the generator is not very good at creating realistic images and the discriminator easily flags them as fakes. Eventually, however, the generator is supposed to learn from the discriminator’s responses and begin to produce increasingly more realistic images. In this way the generator and discriminator are adversaries, with the former trying to fool the latter and the latter trying to avoid being fooled, according to Chintala.

Adversarial network generators tested up to now in AI labs—at Facebook and elsewhere—have typically failed to show significant improvement even after interacting with discriminators. In an attempt to remedy this, Facebook researchers created generators that have specially crafted structures of interconnected layers, an arrangement known in the AI community as a “deep convolutional generative adversarial network,” or DCGAN. Each of these layers consists of a particular algorithm applied to the input that the network gets. The generator’s first layer runs an algorithm that extracts raw pixels and simple motifs from a dataset representing an image. The next layer combines these motifs into slightly more complex arrangements. The next layers detect parts of objects, assemble them into objects and create scenes, respectively, until the entire image is created. “There’s a hierarchy of layers, which is where the word ‘deep’ comes from,” LeCun says.

The researchers found that their DCGANs could, among other things, learn to draw specific objects as the training progressed. They also got a better understanding of what happens as data move from layer to layer within the neural network. In addition, LeCun, Chintala and their colleagues tested their generator’s predictive capabilities by having it use raw data to produce video frames. In one experiment they fed the generator four frames of video and had it produce the next two frames based on those data. The resulting AI-generated frames looked like a realistic continuation of the action, whether it was a person walking or simply making head movements.

More intelligent assistants

LeCun thinks such predictive abilities could enhance Facebook’s ability to engage users, using the common sense the site has developed to essentially make educated guesses about them. “If we know how to build dialogue systems that have an idea of what the person dialoguing wants or thinks, that means we can have chatbots that are actually useful and interact with you in a natural way,” LeCun says. Improved predictive capabilities could likewise help improve the Facebook M virtual assistant, which faces growing competition from Apple’s Siri, Google’s upcoming Google Assistant, Amazon's Alexa and Microsoft's Cortana.

“There is still a long way—very long way—to go before [machines have common sense], but I share with [LeCun and his colleagues] the belief that exploring better unsupervised learning algorithms is a crucial key towards human-level AI,” says Yoshua Bengio, a University of Montreal computer science professor and a co-author of the 2014 study that introduced much of the AI world to generative adversarial networks. Bengio, who was not involved in the Facebook AI research, addressed deep learning’s progress in the June 2016 Scientific American article titled “Machines Who Learn.”

Facebook’s interest in unsupervised machine learning is part of a larger trend that has some of the largest Internet companies—including Amazon, Apple, Google, Microsoft and Twitter—buying AI startup companies and investing in their own studies. Earlier this month Microsoft Research announced its effort to develop a system that could tell a story based on a series of related images. Google’s AlphaGo program made headlines in March when it convincingly beat one of the world’s best Go players at his own game. AI likewise plays a crucial role in efforts by Alphabet, Inc.—Google’s parent company—to develop a driverless car. Apple executives in the past week talked at the company’s Worldwide Developer Conference about wrestling with efforts to advance AI in its products without hurting widely publicized efforts to ensure customer privacy. “It's obvious [that AI] is likely to completely change their business as well as the whole world's economy in a major way,” Bengio says.

LeCun agrees that AI is clearly a very strategic technology for any company that operates on the Web or has any kind of digital presence. “Not just for user interfaces or content filtering but in general,” he says. “People will interact with machines in a very natural way, and we need to get machines to understand people.”

Yahoo Finance

How robots paved the way for Donald Trump

July 14, 2016

See any humans? A Ford assembly plant in Missouri.

Think we don’t make anything in America any more? Think again.

US manufacturing output is close to a record high, even when adjusted for inflation. The reason that sounds surprising is manufacturing jobs have been disappearing since the late 1980s, and now that number is just 12.3 million. Since 1989, manufacturing output has surged 69% while employment has fallen by 32%.

Manufacturers are doing more with less because of technology: computerized machines, streamlined processes, and on just about any factory floor that’s been built or revamped during the last 20 years, robots. “Automation is eating jobs from the inside out,” says Moshe Vardi, a professor of computational engineering at Rice University in Houston. “It’s the major cause of job losses in manufacturing.”

There’s a different storyline in the presidential campaign, with Donald Trump blaming bad trade deals and unfair labor practices in China and Mexico for the loss of decent-paying blue-collar jobs in the United States. “They’re eating our lunch,” Trump often says of trading partners that pay their workers below US standards and sell billions of cheap imports to Americans.

It’s easier to blame other countries for the loss of American jobs than it is to blame technology entrepreneurs, many of them American, who have revolutionized manufacturing and will continue to do so. But the numbers do suggest that technology has made many manufacturers far more productive and cut the need for human workers. These two charts show manufacturing output and employment during the last 30 years – and they’re clearly going in opposite directions.

Screen Shot 2016-07-13 at 2.04.46 PM

Screen Shot 2016-07-13 at 2.05.02 PM

The loss of manufacturing jobs is probably causing more damage to the middle-class than anything else, as families try to cope with stagnant or declining pay, shrinking opportunity and the alarming prospect that today’s digital economy is simply leaving them behind. Trump, more than any political figure of modern times, has tapped into that anxiety with his call to “make America great again” and remake trade deals he blames for the plight of the formerly middle class.

But Trump, the likely Republican nominee for president, may be targeting the wrong problem by focusing on trade rather than technology. If elected, it’s possible he could rework trade deals only to find that US manufacturers still aren’t hiring and the trend toward replacing workers with machines only intensifies.

Economist Larry Summers has argued that a large-scale substitution of technology for workers might be underway, and only just getting started. In restaurants, kiosks and tablets are replacing humans who used to take your order. Robots process packages in distribution warehouses. Some white-collar work, such as legal research, can be computerized. Researchers at Oxford University and consulting firm McKinsey estimate that nearly half of all jobs are susceptible to automation. The next sector likely to be robotified is transportation, as self-driving trucks, ships and construction machinery begin to come online.

In theory, people whose jobs are displaced by new technology are supposed to get retrained, move if necessary, and find new work in a growing field. But instead of “upscaling,” most displaced workers end up “downscaling” – taking lower-paying jobs and in many cases falling out of the middle class. “The economy for working-class people has been miserable,” says Vardi. “They are not sharing in economic growth from the Internet.”

While workers connected to the global, digital economy — the top 20% of earners, more or less – are generally doing fine, others are reeling from the twin crush of globalization and digital technology. The percentage of adult men who have a job or are looking for one is a scant 69.2%, nearly the lowest level on record. In the years following World War II, more than 85% of men had a job or wanted one. The decline has been most pronounced since 2008, and that growing segment of economic dropouts tend to be older white men without a college degree – Trump’s strongest supporters.

Trump wants importers such as China and Mexico to pay larger tariffs on goods they ship to the United States, to make home-grown goods more competitive and boost hiring at home. But he hasn’t said a word about disarming the robots. The march of technology may be the tougher challenge.

NYT

Directory: tlairson -> ibtech
tlairson -> The South China Sea Is the Future of Conflict
tlairson -> Nyt amid Tension, China Blocks Crucial Exports to Japan By keith bradsher published: September 22, 2010
tlairson -> China Alters Its Strategy in Diplomatic Crisis With Japan By jane perlez
tlairson -> The Asia-Pacific Journal, Vol 11, Issue 21, No. 3, May 27, 2013. Much Ado over Small Islands: The Sino-Japanese Confrontation over Senkaku/Diaoyu
tlairson -> Chapter 5 The Political Economy of Global Production and Exchange
tlairson -> Chapter IX power, Wealth and Interdependence in an Era of Advanced Globalization
tlairson -> Nyt india's Future Rests With the Markets By manu joseph published: March 27, 2013
tlairson -> Developmental State
tlairson -> The Economist Singapore The Singapore exception To continue to flourish in its second half-century, South-East Asia’s miracle city-state will need to change its ways, argues Simon Long
ibtech -> History of the Microprocessor and the Personal Computer, Part 2

Download 5.27 Mb.

Share with your friends: