Tag Archives: better

#435676 Intel’s Neuromorphic System Hits 8 ...

At the DARPA Electronics Resurgence Initiative Summit today in Detroit, Intel plans to unveil an 8-million-neuron neuromorphic system comprising 64 Loihi research chips—codenamed Pohoiki Beach. Loihi chips are built with an architecture that more closely matches the way the brain works than do chips designed to do deep learning or other forms of AI. For the set of problems that such “spiking neural networks” are particularly good at, Loihi is about 1,000 times as fast as a CPU and 10,000 times as energy efficient. The new 64-Loihi system represents the equivalent of 8-million neurons, but that’s just a step to a 768-chip, 100-million-neuron system that the company plans for the end of 2019.

Intel and its research partners are just beginning to test what massive neural systems like Pohoiki Beach can do, but so far the evidence points to even greater performance and efficiency, says Mike Davies, director of neuromorphic research at Intel.

“We’re quickly accumulating results and data that there are definite benefits… mostly in the domain of efficiency. Virtually every one that we benchmark…we find significant gains in this architecture,” he says.

Going from a single-Loihi to 64 of them is more of a software issue than a hardware one. “We designed scalability into the Loihi chip from the beginning,” says Davies. “The chip has a hierarchical routing interface…which allows us to scale to up to 16,000 chips. So 64 is just the next step.”

Photo: Tim Herman/Intel Corporation

One of Intel’s Nahuku boards, each of which contains 8 to 32 Intel Loihi neuromorphic chips, shown here interfaced to an Intel Arria 10 FPGA development kit. Intel’s latest neuromorphic system, Pohoiki Beach, is made up of multiple Nahuku boards and contains 64 Loihi chips.

Finding algorithms that run well on an 8-million-neuron system and optimizing those algorithms in software is a considerable effort, he says. Still, the payoff could be huge. Neural networks that are more brain-like, such as Loihi, could be immune to some of the artificial intelligence’s—for lack of a better word—dumbness.

For example, today’s neural networks suffer from something called catastrophic forgetting. If you tried to teach a trained neural network to recognize something new—a new road sign, say—by simply exposing the network to the new input, it would disrupt the network so badly that it would become terrible at recognizing anything. To avoid this, you have to completely retrain the network from the ground up. (DARPA’s Lifelong Learning, or L2M, program is dedicated to solving this problem.)

(Here’s my favorite analogy: Say you coached a basketball team, and you raised the net by 30 centimeters while nobody was looking. The players would miss a bunch at first, but they’d figure things out quickly. If those players were like today’s neural networks, you’d have to pull them off the court and teach them the entire game over again—dribbling, passing, everything.)

Loihi can run networks that might be immune to catastrophic forgetting, meaning it learns a bit more like a human. In fact, there’s evidence through a research collaboration with Thomas Cleland’s group at Cornell University, that Loihi can achieve what’s called one-shot learning. That is, learning a new feature after being exposed to it only once. The Cornell group showed this by abstracting a model of the olfactory system so that it would run on Loihi. When exposed to a new virtual scent, the system not only didn't catastrophically forget everything else it had smelled, it learned to recognize the new scent just from the single exposure.

Loihi might also be able to run feature-extraction algorithms that are immune to the kinds of adversarial attacks that befuddle today’s image recognition systems. Traditional neural networks don’t really understand the features they’re extracting from an image in the way our brains do. “They can be fooled with simplistic attacks like changing individual pixels or adding a screen of noise that wouldn’t fool a human in any way,” Davies explains. But the sparse-coding algorithms Loihi can run work more like the human visual system and so wouldn’t fall for such shenanigans. (Disturbingly, humans are not completely immune to such attacks.)

Photo: Tim Herman/Intel Corporation

A close-up shot of Loihi, Intel’s neuromorphic research chip. Intel’s latest neuromorphic system, Pohoiki Beach, will be comprised of 64 of these Loihi chips.

Researchers have also been using Loihi to improve real-time control for robotic systems. For example, last week at the Telluride Neuromorphic Cognition Engineering Workshop—an event Davies called “summer camp for neuromorphics nerds”—researchers were hard at work using a Loihi-based system to control a foosball table. “It strikes people as crazy,” he says. “But it’s a nice illustration of neuromorphic technology. It’s fast, requires quick response, quick planning, and anticipation. These are what neuromorphic chips are good at.” Continue reading

Posted in Human Robots

#435674 MIT Future of Work Report: We ...

Robots aren’t going to take everyone’s jobs, but technology has already reshaped the world of work in ways that are creating clear winners and losers. And it will continue to do so without intervention, says the first report of MIT’s Task Force on the Work of the Future.

The supergroup of MIT academics was set up by MIT President Rafael Reif in early 2018 to investigate how emerging technologies will impact employment and devise strategies to steer developments in a positive direction. And the headline finding from their first publication is that it’s not the quantity of jobs we should be worried about, but the quality.

Widespread press reports of a looming “employment apocalypse” brought on by AI and automation are probably wide of the mark, according to the authors. Shrinking workforces as developed countries age and outstanding limitations in what machines can do mean we’re unlikely to have a shortage of jobs.

But while unemployment is historically low, recent decades have seen a polarization of the workforce as the number of both high- and low-skilled jobs have grown at the expense of the middle-skilled ones, driving growing income inequality and depriving the non-college-educated of viable careers.

This is at least partly attributable to the growth of digital technology and automation, the report notes, which are rendering obsolete many middle-skilled jobs based around routine work like assembly lines and administrative support.

That leaves workers to either pursue high-skilled jobs that require deep knowledge and creativity, or settle for low-paid jobs that rely on skills—like manual dexterity or interpersonal communication—that are still beyond machines, but generic to most humans and therefore not valued by employers. And the growth of emerging technology like AI and robotics is only likely to exacerbate the problem.

This isn’t the first report to note this trend. The World Bank’s 2016 World Development Report noted how technology is causing a “hollowing out” of labor markets. But the MIT report goes further in saying that the cause isn’t simply technology, but the institutions and policies we’ve built around it.

The motivation for introducing new technology is broadly assumed to be to increase productivity, but the authors note a rarely-acknowledged fact: “Not all innovations that raise productivity displace workers, and not all innovations that displace workers substantially raise productivity.”

Examples of the former include computer-aided design software that makes engineers and architects more productive, while examples of the latter include self-service checkouts and automated customer support that replace human workers, often at the expense of a worse customer experience.

While the report notes that companies have increasingly adopted the language of technology augmenting labor, in reality this has only really benefited high-skilled workers. For lower-skilled jobs the motivation is primarily labor cost savings, which highlights the other major force shaping technology’s impact on employment: shareholder capitalism.

The authors note that up until the 1980s, increasing productivity resulted in wage growth across the economic spectrum, but since then average wage growth has failed to keep pace and gains have dramatically skewed towards the top earners.

The report shies away from directly linking this trend to the birth of Reaganomics (something others have been happy to do), but it notes that American veneration of the shareholder as the primary stakeholder in a business and tax policies that incentivize investment in capital rather than labor have exacerbated the negative impacts technology can have on employment.

That means the current focus on re-skilling workers to thrive in the new economy is a necessary, but not sufficient, solution to the disruptive impact technology is having on work, the authors say.

Alongside significant investment in education, fiscal policies need to be re-balanced away from subsidizing investment in physical capital and towards boosting investment in human capital, the authors write, and workers need to have a greater say in corporate decision-making.

The authors point to other developed economies where productivity growth, income growth, and equality haven’t become so disconnected thanks to investments in worker skills, social safety nets, and incentives to invest in human capital. Whether such a radical reshaping of US economic policy is achievable in today’s political climate remains to be seen, but the authors conclude with a call to arms.

“The failure of the US labor market to deliver broadly shared prosperity despite rising productivity is not an inevitable byproduct of current technologies or free markets,” they write. “We can and should do better.”

Image Credit: Simon Abrams / Unsplash/a> Continue reading

Posted in Human Robots

#435660 Toyota Research Developing New ...

With the Olympics taking place next year in Japan, Toyota is (among other things) stepping up its robotics game to help provide “mobility for all.” We know that Toyota’s HSR will be doing work there, along with a few other mobile systems, but the Toyota Research Institute (TRI) has just announced a new telepresence robot called the T-TR1, featuring an absolutely massive screen designed to give you a near-lifesize virtual presence.

T-TR1 is a virtual mobility/tele-presence robot developed by Toyota Research Institute in the United States. It is equipped with a camera atop a large, near-lifesize display.
By projecting an image of a user from a remote location, the robot will help that person feel more physically present at the robot’s location.
With T-TR1, Toyota will give people that are physically unable to attend the events such as the Games a chance to virtually attend, with an on-screen presence capable of conversation between the two locations.

TRI isn’t ready to share much more detail on this system yet (we asked, of course), but we can infer some things from the video and the rest of the info that’s out there. For example, that ball on top is a 360-degree camera (that looks a lot like an Insta360 Pro), giving the remote user just as good of an awareness of their surroundings as they would if they were there in person. There are multiple 3D-sensing systems, including at least two depth cameras plus a lidar at the base. It’s not at all clear whether the robot is autonomous or semi-autonomous (using the sensors for automated obstacle avoidance, say), and since the woman on the other end of the robot does not seem to be controlling it at all for the demo, it’s hard to make an educated guess about the level of autonomy, or even how it’s supposed to be controlled.

We really like that enormous screen—despite the fact that telepresence now requires pants. It adds to the embodiment that makes independent telepresence robots useful.

We really like that enormous screen—despite the fact that telepresence now requires pants. It adds to the embodiment that makes independent telepresence robots useful. It’s also nice that the robot can move fast enough to keep up a person walking briskly. Hopefully, it’s safe for it to move at that speed in an environment more realistic than a carpeted, half-empty conference room, although it’ll probably have to leverage all of those sensors to do so. The other challenge for the T-TR1 will be bandwidth—even assuming that all of the sensor data processing and stuff is done on-robot, 360 cameras are huge bandwidth hogs, plus there’s the primary (presumably high quality) feed from the main camera, and then the video of the user coming the other way. It’s a lot of data in a very latency-sensitive application, and it’ll presumably be operating in places where connectivity is going to be a challenge due to crowds. This has always been a problem for telepresence robots—no matter how amazing your robot is, the experience will often for better or worse be defined by Internet connections that you may have no control over.

We should emphasize that Toyota has only released the bare minimum of information about the T-TR1, although we’re told that we can expect more as the 2020 Olympics approach: opening ceremonies are one year from today.

[ TRI ] Continue reading

Posted in Human Robots

#435658 Video Friday: A Two-Armed Robot That ...

Video Friday is your weekly selection of awesome robotics videos, collected by your Automaton bloggers. We’ll also be posting a weekly calendar of upcoming robotics events for the next few months; here’s what we have so far (send us your events!):

ICRES 2019 – July 29-30, 2019 – London, U.K.
DARPA SubT Tunnel Circuit – August 15-22, 2019 – Pittsburgh, Pa., USA
IEEE Africon 2019 – September 25-27, 2019 – Accra, Ghana
ISRR 2019 – October 6-10, 2019 – Hanoi, Vietnam
Ro-Man 2019 – October 14-18, 2019 – New Delhi, India
Humanoids 2019 – October 15-17, 2019 – Toronto, Canada
Let us know if you have suggestions for next week, and enjoy today’s videos.

I’m sure you’ve seen this video already because you read this blog every day, but if you somehow missed it because you were skiing across Antarctica (the only valid excuse we’re accepting today), here’s our video introducing HMI’s Aquanaut transforming robot submarine.

And after you recover from all that frostbite, make sure and read our in-depth feature article here.

[ Aquanaut ]

Last week we complained about not having seen a ballbot with a manipulator, so Roberto from CMU shared a new video of their ballbot, featuring a pair of 7-DoF arms.

We should learn more at Humanoids 2019.

[ CMU ]

Thanks Roberto!

The FAA is making it easier for recreational drone pilots to get near-realtime approval to fly in lightly controlled airspace.

[ LAANC ]

Self-reconfigurable modular robots are usually composed of multiple modules with uniform docking interfaces that can be transformed into different configurations by themselves. The reconfiguration planning problem is finding what sequence of reconfiguration actions are required for one arrangement of modules to transform into another. We present a novel reconfiguration planning algorithm for modular robots. The algorithm compares the initial configuration with the goal configuration efficiently. The reconfiguration actions can be executed in a distributed manner so that each module can efficiently finish its reconfiguration task which results in a global reconfiguration for the system. In the end, the algorithm is demonstrated on real modular robots and some example reconfiguration tasks are provided.

[ CKbot ]

A nice design of a gripper that uses a passive thumb of sorts to pick up flat objects from flat surfaces.

[ Paper ] via [ Laval University ]

I like this video of a palletizing robot from Kawasaki because in the background you can see a human doing the exact same job and obviously not enjoying it.

[ Kawasaki ]

This robot cleans and “brings joy and laughter.” What else do we need?

I do appreciate that all the robots are named Leo, and that they’re also all female.

[ LionsBot ]

This is less of a dishwashing robot and more of a dishsorting robot, but we’ll forgive it because it doesn’t drop a single dish.

[ TechMagic ]

Thanks Ryosuke!

A slight warning here that the robot in the following video (which costs something like $180,000) appears “naked” in some scenes, none of which are strictly objectionable, we hope.

Beautifully slim and delicate motion life-size motion figures are ideal avatars for expressing emotions to customers in various arts, content and businesses. We can provide a system that integrates not only motion figures but all moving devices.

[ Speecys ]

The best way to operate a Husky with a pair of manipulators on it is to become the robot.

[ UT Austin ]

The FlyJacket drone control system from EPFL has been upgraded so that it can yank you around a little bit.

In several fields of human-machine interaction, haptic guidance has proven to be an effective training tool for enhancing user performance. This work presents the results of psychophysical and motor learning studies that were carried out with human participant to assess the effect of cable-driven haptic guidance for a task involving aerial robotic teleoperation. The guidance system was integrated into an exosuit, called the FlyJacket, that was developed to control drones with torso movements. Results for the Just Noticeable Difference (JND) and from the Stevens Power Law suggest that the perception of force on the users’ torso scales linearly with the amplitude of the force exerted through the cables and the perceived force is close to the magnitude of the stimulus. Motor learning studies reveal that this form of haptic guidance improves user performance in training, but this improvement is not retained when participants are evaluated without guidance.

[ EPFL ]

The SAND Challenge is an opportunity for small businesses to compete in an autonomous unmanned aerial vehicle (UAV) competition to help NASA address safety-critical risks associated with flying UAVs in the national airspace. Set in a post-natural disaster scenario, SAND will push the envelope of aviation.

[ NASA ]

Legged robots have the potential to traverse diverse and rugged terrain. To find a safe and efficient navigation path and to carefully select individual footholds, it is useful to predict properties of the terrain ahead of the robot. In this work, we propose a method to collect data from robot-terrain interaction and associate it to images, to then train a neural network to predict terrain properties from images.

[ RSL ]

Misty wants to be your new receptionist.

[ Misty Robotics ]

For years, we’ve been pointing out that while new Roombas have lots of great features, older Roombas still do a totally decent job of cleaning your floors. This video is a performance comparison between the newest Roomba (the S9+) and the original 2002 Roomba (!), and the results will surprise you. Or maybe they won’t.

[ Vacuum Wars ]

Lex Fridman from MIT interviews Chris Urmson, who was involved in some of the earliest autonomous vehicle projects, Google’s original self-driving car among them, and is currently CEO of Aurora Innovation.

Chris Urmson was the CTO of the Google Self-Driving Car team, a key engineer and leader behind the Carnegie Mellon autonomous vehicle entries in the DARPA grand challenges and the winner of the DARPA urban challenge. Today he is the CEO of Aurora Innovation, an autonomous vehicle software company he started with Sterling Anderson, who was the former director of Tesla Autopilot, and Drew Bagnell, Uber’s former autonomy and perception lead.

[ AI Podcast ]

In this week’s episode of Robots in Depth, Per speaks with Lael Odhner from RightHand Robotics.

Lael Odhner is a co-founder of RightHand Robotics, that is developing a gripper based on the combination of control and soft, compliant parts to get better grasping of objects. Their work focuses on grasping and manipulating everyday human objects in everyday environments.This mimics how human hands combine control and flexibility to grasp objects with great dexterity.

The combination of control and compliance makes the RightHand robotics gripper very light-weight and affordable. The compliance makes it easier to grasp objects of unknown shape and differs from the way industrial robots usually grip. The compliance also helps in a more unstructured environment where contact with the object and its surroundings cannot be exactly predicted.

[ RightHand Robotics ] via [ Robots in Depth ] Continue reading

Posted in Human Robots

#435656 Will AI Be Fashion Forward—or a ...

The narrative that often accompanies most stories about artificial intelligence these days is how machines will disrupt any number of industries, from healthcare to transportation. It makes sense. After all, technology already drives many of the innovations in these sectors of the economy.

But sneakers and the red carpet? The definitively low-tech fashion industry would seem to be one of the last to turn over its creative direction to data scientists and machine learning algorithms.

However, big brands, e-commerce giants, and numerous startups are betting that AI can ingest data and spit out Chanel. Maybe it’s not surprising, given that fashion is partly about buzz and trends—and there’s nothing more buzzy and trendy in the world of tech today than AI.

In its annual survey of the $3 trillion fashion industry, consulting firm McKinsey predicted that while AI didn’t hit a “critical mass” in 2018, it would increasingly influence the business of everything from design to manufacturing.

“Fashion as an industry really has been so slow to understand its potential roles interwoven with technology. And, to be perfectly honest, the technology doesn’t take fashion seriously.” This comment comes from Zowie Broach, head of fashion at London’s Royal College of Arts, who as a self-described “old fashioned” designer has embraced the disruptive nature of technology—with some caveats.

Co-founder in the late 1990s of the avant-garde fashion label Boudicca, Broach has always seen tech as a tool for designers, even setting up a website for the company circa 1998, way before an online presence became, well, fashionable.

Broach told Singularity Hub that while she is generally optimistic about the future of technology in fashion—the designer has avidly been consuming old sci-fi novels over the last few years—there are still a lot of difficult questions to answer about the interface of algorithms, art, and apparel.

For instance, can AI do what the great designers of the past have done? Fashion was “about designing, it was about a narrative, it was about meaning, it was about expression,” according to Broach.

AI that designs products based on data gleaned from human behavior can potentially tap into the Pavlovian response in consumers in order to make money, Broach noted. But is that channeling creativity, or just digitally dabbling in basic human brain chemistry?

She is concerned about people retaining control of the process, whether we’re talking about their data or their designs. But being empowered with the insights machines could provide into, for example, the geographical nuances of fashion between Dubai, Moscow, and Toronto is thrilling.

“What is it that we want the future to be from a fashion, an identity, and design perspective?” she asked.

Off on the Right Foot
Silicon Valley and some of the biggest brands in the industry offer a few answers about where AI and fashion are headed (though not at the sort of depths that address Broach’s broader questions of aesthetics and ethics).

Take what is arguably the biggest brand in fashion, at least by market cap but probably not by the measure of appearances on Oscar night: Nike. The $100 billion shoe company just gobbled up an AI startup called Celect to bolster its data analytics and optimize its inventory. In other words, Nike hopes it will be able to figure out what’s hot and what’s not in a particular location to stock its stores more efficiently.

The company is going even further with Nike Fit, a foot-scanning platform using a smartphone camera that applies AI techniques from fields like computer vision and machine learning to find the best fit for each person’s foot. The algorithms then identify and recommend the appropriately sized and shaped shoe in different styles.

No doubt the next step will be to 3D print personalized and on-demand sneakers at any store.

San Francisco-based startup ThirdLove is trying to bring a similar approach to bra sizes. Its 20-member data team, Fortune reported, has developed the Fit Finder quiz that uses machine learning algorithms to help pick just the right garment for every body type.

Data scientists are also a big part of the team at Stitch Fix, a former San Francisco startup that went public in 2017 and today sports a market cap of more than $2 billion. The online “personal styling” company uses hundreds of algorithms to not only make recommendations to customers, but to help design new styles and even manage the subscription-based supply chain.

Future of Fashion
E-commerce giant Amazon has thrown its own considerable resources into developing AI applications for retail fashion—with mixed results.

One notable attempt involved a “styling assistant” that came with the company’s Echo Look camera that helped people catalog and manage their wardrobes, evening helping pick out each day’s attire. The company more recently revisited the direct consumer side of AI with an app called StyleSnap, which matches clothes and accessories uploaded to the site with the retailer’s vast inventory and recommends similar styles.

Behind the curtains, Amazon is going even further. A team of researchers in Israel have developed algorithms that can deduce whether a particular look is stylish based on a few labeled images. Another group at the company’s San Francisco research center was working on tech that could generate new designs of items based on images of a particular style the algorithms trained on.

“I will say that the accumulation of many new technologies across the industry could manifest in a highly specialized style assistant, far better than the examples we’ve seen today. However, the most likely thing is that the least sexy of the machine learning work will become the most impactful, and the public may never hear about it.”

That prediction is from an online interview with Leanne Luce, a fashion technology blogger and product manager at Google who recently wrote a book called, succinctly enough, Artificial Intelligence and Fashion.

Data Meets Design
Academics are also sticking their beakers into AI and fashion. Researchers at the University of California, San Diego, and Adobe Research have previously demonstrated that neural networks, a type of AI designed to mimic some aspects of the human brain, can be trained to generate (i.e., design) new product images to match a buyer’s preference, much like the team at Amazon.

Meanwhile, scientists at Hong Kong Polytechnic University are working with China’s answer to Amazon, Alibaba, on developing a FashionAI Dataset to help machines better understand fashion. The effort will focus on how algorithms approach certain building blocks of design, what are called “key points” such as neckline and waistline, and “fashion attributes” like collar types and skirt styles.

The man largely behind the university’s research team is Calvin Wong, a professor and associate head of Hong Kong Polytechnic University’s Institute of Textiles and Clothing. His group has also developed an “intelligent fabric defect detection system” called WiseEye for quality control, reducing the chance of producing substandard fabric by 90 percent.

Wong and company also recently inked an agreement with RCA to establish an AI-powered design laboratory, though the details of that venture have yet to be worked out, according to Broach.

One hope is that such collaborations will not just get at the technological challenges of using machines in creative endeavors like fashion, but will also address the more personal relationships humans have with their machines.

“I think who we are, and how we use AI in fashion, as our identity, is not a superficial skin. It’s very, very important for how we define our future,” Broach said.

Image Credit: Inspirationfeed / Unsplash Continue reading

Posted in Human Robots