Computed·Blg

Via Smithsonian

-----

Via Techcrunch

-----

Aiming to do for Machine Learning what MySQL did for database servers, U.S. and UK-based PredictionIO has raised $2.5 million in seed funding from a raft of investors including Azure Capital, QuestVP, CrunchFund (of which TechCrunch founder Mike Arrington is a Partner), Stanford University‘s StartX Fund, France-based Kima Ventures, IronFire, Sood Ventures and XG Ventures. The additional capital will be used to further develop its open source Machine Learning server, which significantly lowers the barriers for developers to build more intelligent products, such as recommendation or prediction engines, without having to reinvent the wheel.

Being an open source company — after pivoting from offering a “user behavior prediction-as-a-service” under its old TappingStone product name — PredictionIO plans to generate revenue in the same way MySQL and other open source products do. “We will offer an Enterprise support license and, probably, an enterprise edition with more advanced features,” co-founder and CEO Simon Chan tells me.

The problem PredictionIO is setting out to solve is that building Machine Learning into products is expensive and time-consuming — and in some instances is only really within the reach of major and heavily-funded tech companies, such as Google or Amazon, who can afford a large team of PhDs/data scientists. By utilising the startup’s open source Machine Learning server, startups or larger enterprises no longer need to start from scratch, while also retaining control over the source code and the way in which PredictionIO integrates with their existing wares.

In fact, the degree of flexibility and reassurance an open source product offers is the very reason why PredictionIO pivoted away from a SaaS model and chose to open up its codebase. It did so within a couple of months of launching its original TappingStone product. Fail fast, as they say.

“We changed from TappingStone (Machine Learning as a Service) to PredictionIO (open source server) in the first 2 months once we built the first prototype,” says Chan. “As developers ourselves, we realise that Machine Learning is useful only if it’s customizable to each unique application. Therefore, we decided to open source the whole product.”

The pivot appears to be working, too, and not just validated by today’s funding. To date, Chan says its open source Machine Learning server is powering thousands of applications with 4000+ developers engaged with the project. “Unlike other data science tools that focus on solving data researchers’ problems, PredictionIO is built for every programmer,” he adds.

Other competitors Chan cites include “closed ‘black box” MLaaS services or software’, such as Google Prediction API, Wise.io, BigML, and Skytree.

Examples of who is currently using PredictionIO include Le Tote, a clothing subscription/rental service that is using PredictionIO to predict customers’ fashion preferences, and PerkHub, which is using PredictionIO to personalize product recommendations in the weekly ‘group buying’ emails they send out.

 Via Computed·Blg

-----

Can computers learn to read? We think so. "Read the Web" is a research project that attempts to create a computer system that learns over time to read the web. Since January 2010, our computer system called NELL (Never-Ending Language Learner) has been running continuously, attempting to perform two tasks each day:

• First, it attempts to "read," or extract facts from text found in hundreds of millions of web pages (e.g., playsInstrument(George_Harrison, guitar)).
• Second, it attempts to improve its reading competence, so that tomorrow it can extract more facts from the web, more accurately.

So far, NELL has accumulated over 50 million candidate beliefs by reading the web, and it is considering these at different levels of confidence. NELL has high confidence in 2,132,551 of these beliefs — these are displayed on this website. It is not perfect, but NELL is learning. You can track NELL's progress below or @cmunell on Twitter, browse and download its knowledge base, read more about our technical approach, or join the discussion group.

Via MIT

-----

Small cubes with no exterior moving parts can propel themselves forward, jump on top of each other, and snap together to form arbitrary shapes.

A prototype of a new modular robot, with its innards exposed and its flywheel — which gives it the ability to move independently — pulled out.
Photo: M. Scott Brauer

In 2011, when an MIT senior named John Romanishin proposed a new design for modular robots to his robotics professor, Daniela Rus, she said, “That can’t be done.”

Two years later, Rus showed her colleague Hod Lipson, a robotics researcher at Cornell University, a video of prototype robots, based on Romanishin’s design, in action. “That can’t be done,” Lipson said.

In November, Romanishin — now a research scientist in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) — Rus, and postdoc Kyle Gilpin will establish once and for all that it can be done, when they present a paper describing their new robots at the IEEE/RSJ International Conference on Intelligent Robots and Systems.

Known as M-Blocks, the robots are cubes with no external moving parts. Nonetheless, they’re able to climb over and around one another, leap through the air, roll across the ground, and even move while suspended upside down from metallic surfaces. 

Inside each M-Block is a flywheel that can reach speeds of 20,000 revolutions per minute; when the flywheel is braked, it imparts its angular momentum to the cube. On each edge of an M-Block, and on every face, are cleverly arranged permanent magnets that allow any two cubes to attach to each other.

“It’s one of these things that the [modular-robotics] community has been trying to do for a long time,” says Rus, a professor of electrical engineering and computer science and director of CSAIL. “We just needed a creative insight and somebody who was passionate enough to keep coming at it — despite being discouraged.”

Embodied abstraction

As Rus explains, researchers studying reconfigurable robots have long used an abstraction called the sliding-cube model. In this model, if two cubes are face to face, one of them can slide up the side of the other and, without changing orientation, slide across its top.

The sliding-cube model simplifies the development of self-assembly algorithms, but the robots that implement them tend to be much more complex devices. Rus’ group, for instance, previously developed a modular robot called the Molecule, which consisted of two cubes connected by an angled bar and had 18 separate motors. “We were quite proud of it at the time,” Rus says.

According to Gilpin, existing modular-robot systems are also “statically stable,” meaning that “you can pause the motion at any point, and they’ll stay where they are.” What enabled the MIT researchers to drastically simplify their robots’ design was giving up on the principle of static stability.

“There’s a point in time when the cube is essentially flying through the air,” Gilpin says. “And you are depending on the magnets to bring it into alignment when it lands. That’s something that’s totally unique to this system.”

That’s also what made Rus skeptical about Romanishin’s initial proposal. “I asked him build a prototype,” Rus says. “Then I said, ‘OK, maybe I was wrong.’”

Sticking the landing

To compensate for its static instability, the researchers’ robot relies on some ingenious engineering. On each edge of a cube are two cylindrical magnets, mounted like rolling pins. When two cubes approach each other, the magnets naturally rotate, so that north poles align with south, and vice versa. Any face of any cube can thus attach to any face of any other.

The cubes’ edges are also beveled, so when two cubes are face to face, there’s a slight gap between their magnets. When one cube begins to flip on top of another, the bevels, and thus the magnets, touch. The connection between the cubes becomes much stronger, anchoring the pivot. On each face of a cube are four more pairs of smaller magnets, arranged symmetrically, which help snap a moving cube into place when it lands on top of another.

As with any modular-robot system, the hope is that the modules can be miniaturized: the ultimate aim of most such research is hordes of swarming microbots that can self-assemble, like the “liquid steel” androids in the movie “Terminator II.” And the simplicity of the cubes’ design makes miniaturization promising.

But the researchers believe that a more refined version of their system could prove useful even at something like its current scale. Armies of mobile cubes could temporarily repair bridges or buildings during emergencies, or raise and reconfigure scaffolding for building projects. They could assemble into different types of furniture or heavy equipment as needed. And they could swarm into environments hostile or inaccessible to humans, diagnose problems, and reorganize themselves to provide solutions.

Strength in diversity

The researchers also imagine that among the mobile cubes could be special-purpose cubes, containing cameras, or lights, or battery packs, or other equipment, which the mobile cubes could transport. “In the vast majority of other modular systems, an individual module cannot move on its own,” Gilpin says. “If you drop one of these along the way, or something goes wrong, it can rejoin the group, no problem.”

“It’s one of those things that you kick yourself for not thinking of,” Cornell’s Lipson says. “It’s a low-tech solution to a problem that people have been trying to solve with extraordinarily high-tech approaches.”

“What they did that was very interesting is they showed several modes of locomotion,” Lipson adds. “Not just one cube flipping around, but multiple cubes working together, multiple cubes moving other cubes — a lot of other modes of motion that really open the door to many, many applications, much beyond what people usually consider when they talk about self-assembly. They rarely think about parts dragging other parts — this kind of cooperative group behavior.”

In ongoing work, the MIT researchers are building an army of 100 cubes, each of which can move in any direction, and designing algorithms to guide them. “We want hundreds of cubes, scattered randomly across the floor, to be able to identify each other, coalesce, and autonomously transform into a chair, or a ladder, or a desk, on demand,” Romanishin says.

Via MIT Technology Review

-----

Researchers have created software that predicts when and where disease outbreaks might occur based on two decades of New York Times articles and other online data. The research comes from Microsoft and the Technion-Israel Institute of Technology.

The system could someday help aid organizations and others be more proactive in tackling disease outbreaks or other problems, says Eric Horvitz, distinguished scientist and codirector at Microsoft Research. “I truly view this as a foreshadowing of what’s to come,” he says. “Eventually this kind of work will start to have an influence on how things go for people.” Horvitz did the research in collaboration with Kira Radinsky, a PhD researcher at the Technion-Israel Institute.

The system provides striking results when tested on historical data. For example, reports of droughts in Angola in 2006 triggered a warning about possible cholera outbreaks in the country, because previous events had taught the system that cholera outbreaks were more likely in years following droughts. A second warning about cholera in Angola was triggered by news reports of large storms in Africa in early 2007; less than a week later, reports appeared that cholera had become established. In similar tests involving forecasts of disease, violence, and a significant numbers of deaths, the system’s warnings were correct between 70 to 90 percent of the time.

Horvitz says the performance is good enough to suggest that a more refined version could be used in real settings, to assist experts at, for example, government aid agencies involved in planning humanitarian response and readiness. “We’ve done some reaching out and plan to do some follow-up work with such people,” says Horvitz.

The system was built using 22 years of New York Times archives, from 1986 to 2007, but it also draws on data from the Web to learn about what leads up to major news events.

“One source we found useful was DBpedia, which is a structured form of the information inside Wikipedia constructed using crowdsourcing,” says Radinsky. “We can understand, or see, the location of the places in the news articles, how much money people earn there, and even information about politics.” Other sources included WordNet, which helps software understand the meaning of words, and OpenCyc, a database of common knowledge.

All this information provides valuable context that’s not available in news article, and which is necessary to figure out general rules for what events precede others. For example, the system could infer connections between events in Rwandan and Angolan cities based on the fact that they are both in Africa, have similar GDPs, and other factors. That approach led the software to conclude that, in predicting cholera outbreaks, it should consider a country or city’s location, proportion of land covered by water, population density, GDP, and whether there had been a drought the year before.

Horvitz and Radinsky are not the first to consider using online news and other data to forecast future events, but they say they make use of more data sources—over 90 in total—which allows their system to be more general-purpose.

There’s already a small market for predictive tools. For example, a startup called Recorded Future makes predictions about future events harvested from forward-looking statements online and other sources, and it includes government intelligence agencies among its customers (see “See the Future With a Search”). Christopher Ahlberg, the company’s CEO and cofounder, says that the new research is “good work” that shows how predictions can be made using hard data, but also notes that turning the prototype system into a product would require further development.

Microsoft doesn’t have plans to commercialize Horvitz and Radinsky’s research as yet, but the project will continue, says Horvitz, who wants to mine more newspaper archives as well as digitized books.

Many things about the world have changed in recent decades, but human nature and many aspects of the environment have stayed the same, Horvitz says, so software may be able to learn patterns from even very old data that can suggest what’s ahead. “I’m personally interested in getting data further back in time,” he says.

Via ExtremTech

-----

Of course, another possibility is that we simply keep the moths. After all, why should we spend time and money on an artificial system when mother nature, as always, has already done the hard work for us? In much the same way that miners used canaries and border police use sniffer dogs, why shouldn’t robots be controlled by insects? The silkmoth is graced with perhaps the most sensitive olfactory system in the world. For now it might only be sensitive to not-so-useful scents like the female sex pheromone, but who’s to say that genetic engineering won’t allow for silkmoths that can sniff out bombs or drugs or chemical spills?

Who nose: Maybe genetically modified insects with robotic exoskeletons are merely an intermediary step towards real nanobots that fly around, fixing, cleaning, and constructing our environment.

Via Human Rights Watch

-----

(Washington, DC) – Governments should pre-emptively ban fully autonomous weapons because of the danger they pose to civilians in armed conflict, Human Rights Watch said in a report released today. These future weapons, sometimes called “killer robots,” would be able to choose and fire on targets without human intervention. 

The United Kingdom’s Taranis combat aircraft, whose prototype was unveiled in 2010, is designed strike distant targets, “even in another continent.” While the Ministry of Defence has stated that humans will remain in the loop, the Taranis exemplifies the move toward increased autonomy.

© 2010 AP Photo

The 50-page report, “Losing Humanity: The Case Against Killer Robots,” outlines concerns about these fully autonomous weapons, which would inherently lack human qualities that provide legal and non-legal checks on the killing of civilians. In addition, the obstacles to holding anyone accountable for harm caused by the weapons would weaken the law’s power to deter future violations.

“Giving machines the power to decide who lives and dies on the battlefield would take technology too far,” said Steve Goose, Arms Division director at Human Rights Watch. “Human control of robotic warfare is essential to minimizing civilian deaths and injuries.”

The South Korean SGR-1 sentry robot, a precursor to a fully autonomous weapon, can detect people in the Demilitarized Zone and, if a human grants the command, fire its weapons. The robot is shown here during a test with a surrendering enemy soldier.

© 2007 Getty Images

“Losing Humanity” is the first major publication about fully autonomous weapons by a nongovernmental organization and is based on extensive research into the law, technology, and ethics of these proposed weapons. It is jointly published by Human Rights Watch and the Harvard Law School International Human Rights Clinic.

Human Rights Watch and the International Human Rights Clinic called for an international treaty that would absolutely prohibit the development, production, and use of fully autonomous weapons. They also called on individual nations to pass laws and adopt policies as important measures to prevent development, production, and use of such weapons at the domestic level.

Fully autonomous weapons do not yet exist, and major powers, including the United States, have not made a decision to deploy them. But high-tech militaries are developing or have already deployed precursors that illustrate the push toward greater autonomy for machines on the battlefield. The United States is a leader in this technological development. Several other countries – including China, Germany, Israel, South Korea, Russia, and the United Kingdom – have also been involved. Many experts predict that full autonomy for weapons could be achieved in 20 to 30 years, and some think even sooner.

“It is essential to stop the development of killer robots before they show up in national arsenals,” Goose said. “As countries become more invested in this technology, it will become harder to persuade them to give it up.”

Fully autonomous weapons could not meet the requirements of international humanitarian law, Human Rights Watch and the Harvard clinic said. They would be unable to distinguish adequately between soldiers and civilians on the battlefield or apply the human judgment necessary to evaluate the proportionality of an attack – whether civilian harm outweighs military advantage.

These robots would also undermine non-legal checks on the killing of civilians. Fully autonomous weapons could not show human compassion for their victims, and autocrats could abuse them by directing them against their own people. While replacing human troops with machines could save military lives, it could also make going to war easier, which would shift the burden of armed conflict onto civilians.

Finally, the use of fully autonomous weapons would create an accountability gap. Trying to hold the commander, programmer, or manufacturer legally responsible for a robot’s actions presents significant challenges. The lack of accountability would undercut the ability to deter violations of international law and to provide victims meaningful retributive justice.

While most militaries maintain that for the immediate future humans will retain some oversight over the actions of weaponized robots, the effectiveness of that oversight is questionable, Human Rights Watch and the Harvard clinic said. Moreover, military statements have left the door open to full autonomy in the future.

“Action is needed now, before killer robots cross the line from science fiction to feasibility,” Goose said.

Via DVICE

-----

The techno-wizards over at Google X, the company's R&D laboratory working on its self-driving cars and Project Glass, linked 16,000 processors together to form a neural network and then had it go forth and try to learn on its own. Turns out, massive digital networks are a lot like bored humans poking at iPads.

The pretty amazing takeaway here is that this 16,000-processor neural network, spread out over 1,000 linked computers, was not told to look for any one thing, but instead discovered that a pattern revolved around cat pictures on its own.

This happened after Google presented the network with image stills from 10 million random YouTube videos. The images were small thumbnails, and Google's network was sorting through them to try and learn something about them. What it found — and we have ourselves to blame for this — was that there were a hell of a lot of cat faces.

"We never told it during the training, 'This is a cat,'" Jeff Dean, a Google fellow working on the project, told the New York Times. "It basically invented the concept of a cat. We probably have other ones that are side views of cats."

The network itself does not know what a cat is like you and I do. (It wouldn't, for instance, feel embarrassed being caught watching something like this in the presence of other neural networks.) What it does realize, however, is that there is something that it can recognize as being the same thing, and if we gave it the word, it would very well refer to it as "cat."

So, what's the big deal? Your computer at home is more than powerful enough to sort images. Where Google's neural network differs is that it looked at these 10 million images, recognized a pattern of cat faces, and then grafted together the idea that it was looking at something specific and distinct. It had a digital thought.

Andrew Ng, a computer scientist at Stanford University who is co-leading the study with Dean, spoke to the benefit of something like a self-teaching neural network: "The idea is that instead of having teams of researchers trying to find out how to find edges, you instead throw a ton of data at the algorithm and you let the data speak and have the software automatically learn from the data." The size of the network is important, too, and the human brain is "a million times larger in terms of the number of neurons and synapses" than Google X's simulated mind, according to the researchers.

"It'd be fantastic if it turns out that all we need to do is take current algorithms and run them bigger," Ng added, "but my gut feeling is that we still don't quite have the right algorithm yet."

The full article&videos @The New York Times

Via Flexibility Envelope

-----

Mobot by Barobo.com

Soon you can get your hands on the Mobot modular robot for a very reasonable $270 a module (pre-orders available now). A number of connection plates and attachments will also be available, and I guess you can 3D print your own stuff.

Mobot by Barobo.com

I like the gripper that is powered and controlled by the rotating faceplate. I am sure the same concept can be used to 3D print some cool things in the future. A connector would be an awesome thing and definitely worth a price of some sort.

In general, it seems to be a very competent modular robotics system. It uses a snap together connector, making it simple and fast to use, but maybe not as strong as a system that screws together.

There is a Graphical User Interface RobotController, and you can program it with the C/C++ interpreter Ch so everyone from beginner to hard core hacker should be able to do some really cool stuff.

More info on Barobo.com