March 26, 2016

DARPA wants machine learning embedded into devices to collaboratively share wireless spectrum

The $2 million DARPA Spectrum Collaboration Challenge (SC2) is about getting the billions to trillions of wireless devices to efficiently share the wireless spectrum.

Teams will be rewarded for developing smart systems that collaboratively, rather than competitively, adapt in real time to today’s fast-changing, congested spectrum environment—redefining the conventional spectrum management roles of humans and machines in order to maximize the flow of radio frequency (RF) signals. DARPA officials unveiled the new Challenge before some 8000 engineers and communications professionals gathered in Las Vegas at the International Wireless Communications Expo (IWCE).

The primary goal of SC2 is to imbue radios with advanced machine-learning capabilities so they can collectively develop strategies that optimize use of the wireless spectrum in ways not possible with today’s intrinsically inefficient approach of pre-allocating exclusive access to designated frequencies. The challenge is expected to both take advantage of recent significant progress in the fields of artificial intelligence and machine learning and also spur new developments in those research domains, with potential applications in other fields where collaborative decision-making is critical.

“DARPA Challenges have traditionally rewarded teams that dominate their competitors, but when it comes to making the most of the electromagnetic spectrum, the team that shares most intelligently is going to win,” said SC2 program manager Paul Tilghman of DARPA’s Microsystems Technology Office (MTO). “We want to radically accelerate the development of machine-learning technologies and strategies that will allow on-the-fly sharing of spectrum at machine timescales.”

Quantum Fredkin Gate created which is usable for implementing Shor's algorithm for decryption

Researchers from Griffith University and the University of Queensland have overcome one of the key challenges to quantum computing by simplifying a complex quantum logic operation. They demonstrated this by experimentally realizing a challenging circuit -- the quantum Fredkin gate -- for the first time.

"The allure of quantum computers is the unparalleled processing power that they provide compared to current technology," said Dr Raj Patel from Griffith's Centre for Quantum Dynamics.

An artist's rendering of the quantum Fredkin (controlled-SWAP) gate, powered by entanglement, operating on photonic qubits. CREDIT Raj Patel and Geoff Pryde, Center for Quantum Dynamics, Griffith University.

First fully autonomous drone delivery to an urban residence in the USA

Drone delivery company, Flirtey, has made the first urban delivery in the United States by a fully autonomous drone.

Flirtey started in 2013 in Australia, delivering textbooks to universities, before it moved to Nevada. Its six-engine multicopter flew along a predetermined path. When it reached the target house, it lowered a package containing bottled water, emergency food, and a first aid kit. The house was uninhabited, as the flight was a demonstration of what a rescue drone might be able to carry to people in need. Flirtey already conducted a rural delivery test, so it makes sense that urban was next, even if that “urban” is defined as a fairly small town. According to Flirtey CEO Matthew Sweeney, 86% of packages are 5.5 pounds or less, and that the drone is designed to carry payloads that size up to 10 miles away.

UAVs can be operated commercially in a growing number of countries. Flirtey is in discussions with regulators all around the world, and in particular, Flirtey is pioneering drone delivery in New Zealand.

In July 2015, Australian commercial drone startup Flirtey recently made the first ever drone parcel delivery in Auckland, New Zealand by transporting auto parts 2km. Partnering with Fastway Couriers, Flirtey’s drone was able to make the 20-minute car trip in under five minutes.

Pathways to large amounts of solar energy

Solar and wind energy are still a tiny share of the overall energy production mix ARPA-E has the FOCUS (Full-Spectrum Optimized Conversion and Utilization of Sunlight) project to make solar energy a larger part of the US energy mix.

ARPA-E funding Xerox Machine for Super Solar Panels

Researchers at PARC, an R and D-focused subsidiary of Xerox, say they’re developing a new digital printing process that could make it much cheaper to mass-produce concentrated solar photovoltaic systems. Such systems can dramatically increase the efficiency of solar cells by using lenses to concentrate and focus the sunlight onto small cells.

Increasing efficiency could be an effective way to bring down the cost of solar modules, whose price has already fallen dramatically during the past several years. Much of the cost of conventional silicon systems is now due to things like wiring, installation, and permitting. More efficient panels would mean we need fewer of them to produce the same amount of power—which in turn cuts the costs of hardware and installation. But concentrated photovoltaic technology has so far failed to gain traction because it’s still too expensive and bulky to compete with conventional silicon solar panels.

PARC hopes to make the technology more competitive by shrinking the components and designing a new flat-panel form factor, and by developing a relatively inexpensive manufacturing process. The new process will build on a larger effort by PARC researchers to invent a new kind of printer that can precisely deposit “inks” made of tiny semiconductor chips, called “chiplets,” by using assembly principles similar to those behind Xerox photocopiers.

Chiplets in solution

So far, they have demonstrated only the ability to make small-scale devices by wiring together a few printed chiplets. But eventually the technique could allow people to design and print very large arrangements of small electrical and optical components similar to the way they would design and print a document using a Xerox printer.

They have a three year ARPA-E grant to apply the innovative printing process to build microscale arrays of photovoltaic cells on a flat panel.

PARC is working with Sandia National Laboratories (SNL) to develop a breakthrough digital manufacturing capability that will enable the fabrication of panel-sized micro-scale concentrated photovoltaic (micro-CPV) backplanes, overcoming the limitations of current pick-and-place processes.

PARC researchers say the new printing approach could lead to a high-speed manufacturing process that is orders of magnitude cheaper for building these microscale systems over large areas.

The project is called MOSAIC (Micro-scale Optimized Solar-cell Arrays with Integrated Concentration).

Recurrent Neural Nets

How can a net decipher and label patterns that change with time? For example, could a net be used to scan traffic footage and immediately flag a collision? Through the use of a recurrent net, these real-time interactions are now possible.

The Recurrent Neural Net (RNN) is the brainchild of Juergen Schmidhuber and Sepp Hochreiter. The three deep nets we’ve seen thus far – MLP, DBN, and CNN – are known as feedforward networks since a signal moves in only one direction across the layers. In contrast, RNNs have a feedback loop where the net’s output is fed back into the net along with the next input. Since RNNs have just one layer of neurons, they are structurally one of the simplest types of nets.

Like other nets, RNNs receive an input and produce an output. Unlike other nets, the inputs and outputs can come in a sequence. Here are some sample applications for different input-output scenarios:
- Single input, sequence of outputs: image captioning
- Sequence of inputs, single output: document classification
- Sequence of inputs, sequence of outputs: video processing by frame, statistical forecasting of demand in Supply Chain Planning

RNNs are trained using backpropagation through time, which reintroduces the vanishing gradient problem. In fact, the problem is worse with an RNN because each time step is the equivalent of a layer in a feedforward net. Thus if the net is trained for 1000 time steps, the gradient will vanish exponentially as it would in a 1000-layer MLP.

There are different approaches to address this problem, the most popular of which is gating. Gating takes the output of any time step and the next input, and performs a transformation before feeding the result back into the RNN. There are several types of gates, the LSTM being the most popular. Other approaches to address this problem include gradient clipping, steeper gates, and better optimizers.

GPUs are an essential tool for training an RNN. A team at Indico compared the speed boost from using a GPU over a CPU, and found a 250-fold increase. That’s the difference between 1 day and over 8 months!

A recurrent net has one additional capability – it can predict the next item in a sequence, essentially acting as a forecasting engine.

Why HPC supercomputing is speeding up machine learning research

Why is HPC (high performance computing) speeding up machine learning and deep learning research?

Experiments determined the right architecture for speech recognition.

A lot of AI progress is driven by experimentation.

How Scale is Enabling Deep Learning

AI is making rapid progress. To help you stay on the leading edge, Andrew Ng recording a few videos to share the latest trends.

Andrew Yan-Tak Ng is Chief Scientist at Baidu Research in Silicon Valley. In addition, he is an associate professor in the Department of Computer Science and the Department of Electrical Engineering by courtesy at Stanford University. He is chairman of the board of Coursera, an online education platform that he co-founded with Daphne Koller.

He researches primarily in machine learning and deep learning. His early work includes the Stanford Autonomous Helicopter project, which developed one of the most capable autonomous helicopters in the world, and the STAIR (STanford Artificial Intelligence Robot) project, which resulted in ROS, a widely used open-source robotics software platform.

Ng is also the author or co-author of over 100 published papers in machine learning, robotics and related fields, and some of his work in computer vision has been featured in a series of press releases and review

Eric Drexler talks about Cambrian Explosion in AI and Safe Access to Super Intelligence tools

Eric Drexler discusses the progress in Artificial Intelligence and Deep learning

  • a few years ago a Deep Learning network was 5 layers deep,
  • In 2014, the Google Image categorization network was 22 layers deep
  • In 2015, a new approach to information flow enables deep learning networks 200 layers deep

A specialized AI can be used to recursively improve the design of artificial intelligence

The AI software is not as software as we knew it. Here is a mathematical description describing the behavior and memory.
They are spontaneously organized patterns that are emperically engineered.

Eric argues we are being to anthropomorphic about our analysis of AI and AI risks.

Eric thinks we can develop superintelligence without using or creating AI agents.

Eric discusses constraint by construction.

Provide specialist AI only the information that they need for the tasks.

Use an architecture of specialized pieces.

Then think of using superintelligent tools

March 25, 2016

Hybrid airships nearing significant commercialization

Three hybrid airship projects are currently attracting the most attention:

  • the Airlander 10, which just launched this month, in England;
  • Lockheed Martin’s LMH-1; and
  • Pasternak’s Aeroscraft

Pasternak’s Aeroscraft ML866, ML868 and ML86X

Pasternak’s surveillance-airship business is worth millions, but wealth was never his goal. “It’s all about the cargo airship,” he said. His cargo prototype, the Dragon Dream, was in Hangar 2, on the other side of the base. Two hundred and sixty-six feet long—nearly the length of a football field—and ninety-six feet wide, the Dragon Dream was the largest rigid airship built in the U.S. since the nineteen-thirties. And yet it was just half the size of Pasternak’s proposed masterpiece, the Aeroscraft. The Aeroscraft will come in three sizes. The ML866 will be five hundred and fifty-five feet long and able to carry sixty-six tons of cargo. The ML868 will be about thirty per cent larger, with a capacity of two hundred and fifty tons. And the ML86X will be nine hundred and twenty feet (nearly three football fields) long, two hundred and fifteen feet (more than the Tower of Pisa) high, three hundred and fifty-five feet (two Boeing 747s) wide, and able to carry five hundred tons.

To get an idea of the scale of the ML86X, imagine a flying, elongated Houston Astrodome hauling a hundred and fifty elephants.

Worldwide Aeros Corp is an American manufacturer of airships based in Montebello, California. It was founded in 1992 by the current CEO and Chief Engineer, Igor Pasternak, who came to America from Ukraine. It currently employs more than 100 workers.

The company is beginning production of two examples, an ML866 and an ML868 model.

The Aeroscraft’s VTOL capability is like a submarine. For example, when a submarine needs to dive into the water, it takes on water to make it heavier. When the submarine needs to surface, it releases that water to become lighter. Similarly, the Aeroscraft can control its weight by releasing and taking on air, controlling the heaviness or lightness of the vehicle.

2013 - The Dragon Dream had its first float on January 3. The Pentagon declared the tests were a success. On July 4, the Dragon Dream rolled out of the hangar for the first time and on September 11, the first flight of the Dragon Dream occurred.
2014 - Aeros launches 40D 'Sky Dragon' S/N 22 into service following Congressional Christening ceremony; completes 'design freeze' for ML 866 (66-ton) Aeroscraft, initiates production of new 40E 'Sky Dragon' airship; develops and deploys a new tactical aerostat design; advances the company's IP position for the Aeroscraft.
2015 - Aeros receives Patent from USPTO for revolutionary COSH buoyancy management system; launches the North American Defense Advanced Technology Solutions (NADATS) division, readies launch of advanced 40E ‘Sky Dragon’ airship, and continues production on the world’s first cargo airship, the Aeroscraf

Dream Dragon half size prototype

Lockheed Martin LMH-1

Lockheed’s uses helium for only about eighty per cent of its lift; the rest comes from the aerodynamic form of the body. Its weight is intended to make it easier to control on the ground. The first version has a payload of 20 tons.

The LMH-1 should become certified by the Federal Aviation Administration by the end of 2017, paving the way for delivery in 2018.

Airlander 10 and Airlander 50

HAV (Hybrid Air Vehicle, UK) had won £6 million in grants from the U.K. and the European Union and raised £2.4 million from public crowdfunding to give the hybrid aircraft a second life. HAV officially began its “Return-to-Flight” program in May and gave the HAV304 a new name – the Airlander 10. (The “10” refers to the aircraft’s ability to carry ten tons.) Unlike its previous iteration, this one will be used for more than just military purposes.

The Airlander 10 incorporates lighter-than-air technology to create what the company calls “a new breed of hyper-efficient aircraft.” The aircraft will get 60 percent of its lift from internal helium gas and 40 percent from its aerodynamic
form. A critical component of the success of the aircraft is the boost it gets from advanced composite materials. “The Airlander hull is made of a flexible laminate utilizing Vectran as the structural fiber,” says Ashley Appleton head of rigid structures at HAV. “The laminate contains specific features to protect the material from the environment and to retain helium.”

The skin of Airlander 10’s hull is a combination of five tons of multilayered Vectran weave, Tedlar and Mylar surrounding a helium bubble.

Vectran, which consists of a high-performance multifilament yarn spun from liquid crystal polymer, is five times stronger
than steel and 10 times stronger than aluminum. Tedlar is a polyvinyl fluoride film that provides an outer coat and protects the hull from wearing away. Mylar, a form of polyester resin used to make heat-resistant plastic films, creates a gas barrier to minimize helium loss.

The hull is not the only structure on the revitalized aircraft to rely on composites. The engine frames also were made from an undisclosed carbon fiber prepreg. Underneath the Airlander 10 is a 149-foot-long structure made with CFRP and GFRP materials.

The Airlander 10 will eventually have a “big brother,” the Airlander 50, designed for remote access and logistics in markets such as mining, oil and gas and humanitarian relief. Allman says that composite knowledge and expertise gained on the Airlander 10 will be utilized for the Airlander 50

A variety of cargo planes can carry 70-250 tons of cargo.

122 tons for C-5 Galaxy (US Military)
134 tons for a converted Boeing 747

The main advantages of a hybrid airship would be more fuel efficiency and short takeoff or even servicing areas without any airfield. Hybrid airships are intended to fill the middle ground between the low operating cost and low speeds of traditional airships and the higher speed but higher fuel consumption of heavier-than-air craft. By combining dynamic and buoyant lift, hybrids are intended to provide improved airspeed, air-cargo payload capacity and (in some types) hovering capability compared to a pure airship, while having longer endurance and greater lifting capacity compared to a pure HTA type.

Hybrid aircraft technology is claimed to allow a wider range of flight-performance optimizations ranging from significantly heavier than air to near buoyant. This perception of uncommon dynamic flight range when coupled with an appropriate landing system is claimed to allow ultra heavy and affordable airlift transportation.

Hybrid airships offer the potential to deliver supplies directly to users, avoiding the complications inherent in multimodal port operations. From combat cargo lift to humanitarian assistance and disaster relief operations to civilian cargo delivery in austere environments, hybrid airship technology is now poised to transform the transportation landscape.

The cargo bay of the largest Aeroscraft is much larger than any bay container in any commercial freight aircraft carrier today (including the Boeing 747-8F and the Antonov 124 aircraft).

The current payload of the largest helicopters is 16-tons.

Russia upgrading current and future submarines with new missiles and underwater robotic drones

Russia is developing unmanned underwater vehicles for deployment by fifth-generation submarines.

New fifth generation nuclear submarines are expected to have advanced stealth, noise-reduction, automated reconnaissance and warning systems, Russian Navy's Commander-in-Chief Adm. Viktor Chirkov said last fall.

"[Unmanned submarine drones] will be released from the [main] submarine for environmental monitoring using different hardware or to attack the enemy. Torpedoes can be used as a weapon on these carriers. We are calling them underwater robots for now," the official added.

The Sevmash shipyard in northern Russia said in July that the construction of fifth-generation nuclear-powered submarines could begin by 2020.

Russia plans to upgrade its Project 971 [Akula] nuclear submarines with Kalibr cruise missiles, Rear Admiral Viktor Kochemazov said.

The Akula's truly remarkable feature is its low level of noise generation the Soviet and later Russian engineers were able to achieve. An upgraded version, known as the Akula II, was the quietest submarine at the time when it was commissioned, exceeding the upgraded version of the US Los Angeles-class subs.

The Russian Navy operates over ten Akulas as part of its Northern and Pacific Fleets. One Project 971 submarine, currently known as INS Chakra is on a ten-year lease in India.

Japan talking to Lockheed and Boeing about next generation air superiority fighter

Japan has opened talks with Western defense contractors—including Lockheed Martin and Boeing—to develop a next-generation air superiority fighter.

The new Japanese aircraft could be based in part on the technologies being matured on Mitsubishi’s X-2 ATD-X stealth fighter concept demonstrator. The X-2 prototype is set to take to the skies for the first time in days.

Japan needs to replace its F-15J Eagle air superiority fighters. In Tokyo’s estimation, the F-35—which was not designed primarily for the air-to-air role—won’t be enough to meet its requirements. While Tokyo is buying forty-two F-35s, those planes will replace Japan’s F-4J Kai Phantom II fleet. The plane Japan wanted as its F-15J replacement was the F-22, but U.S. law prohibited the Raptor’s export. “Japan really wanted the F-22 but it got the F-35,” a Japan-based source told Reuters. “This is a source of concern and frustration in Tokyo.”

That is one reason the Japanese are aiming to develop what is being called the F-3—potentially a operational variant of the X-2—as an air superiority fighter. Japan aims to field a new air superiority fighter in the 2030s—roughly at the same time the U.S. Air Force and U.S. Navy hope to field the F-X and F/A-XX respectively. That means that Japan might be able to leverage such an effort to join a U.S.-led sixth-generation fighter program. Tokyo probably can’t afford to pay for a $50 billion fighter development program on its own.

Japan's ATD-X protoytype

Boeing 6th generation fighter concept

US Army wants nano drone to support soldiers 500 meter awareness and targeting assistance in urban and other settings

Army leaders want to build a nano drone that weighs less than a pound with the entire unmanned system not weighing more than three pounds, according to a Request for Information released on on March 1.

The Army is calling its nano drone efforts, the Soldier Borne Sensors (SBS) program with Product Manager Soldier Maneuver Sensors (PM SMS) leading the work to outfit soldiers with these high tech unmanned systems.

Developing nano drones to fly through buildings to protect soldiers before having to clear rooms is a high priority for Army leaders. The technology is catching up to Army expectations.

The British Army deploy the Black Hornet nano drone (pictured above) to Afghanistan in 2013. Built by Prox Dynamics, the nano drone measures 4 inches long and weighs only 16 grams to include batteries. It carries three cameras and can fly up to 11 mph.

The PD-100 Personal Reconnaisance System (PRS) made by the Norwegian company Prox Dynamics provides the modern day warrior with a game-changing piece of equipment for instant use on the battlefield. (Image courtesy Prox Dynamics)

US Army wants a bigger drone with more range and better cameras than the PD-100

  • The display will be visible in day or night and backlit with brightness and contrast controls.
  • The system will accept a standard interoperable charging configuration with Military Standard and widely commercially available chargers.
  • Air vehicle must operate in manual mode and operate in waypoint navigation mode.
  • Air vehicle’s operating radius: Line of Sight: 500 meters
  • Air vehicle’s mission endurance: 15 mins
  • Air vehicle shall have, at range of 100 meters day (50 meters night), sufficient resolution for a trained operator to detect a man-sized target with a 90% probability.
  • Air vehicle shall have, at range of 20 meters during the day, sufficient resolution for a trained operator to recognize a man-sized target and determine if armed with a rifle or not with a 90% probability.
  • Air vehicle will have the ability to be hand launched and recovered, without exposing the operator, while in the prone position and while from cover & concealment, a hide site, confined area and alleyway type locations.
  • Air vehicle’s launch time from stored to operational is less than 60 seconds.
  • Air vehicle must be operable in constant winds up to 10 knots and gusts to 15 knots.

Google makes cloud machine learning available for everyone to use

Google Cloud Machine Learning provides modern machine learning services, with pre-trained models and a platform to generate your own tailored models. The neural net-based ML platform has better training performance and increased accuracy compared to other large scale deep learning systems.

Major Google applications use Cloud Machine Learning, including Photos (image search), the Google app (voice search), Translate, and Inbox (Smart Reply). The platform is now available as a cloud service to bring unmatched scale and speed to your business applications.

Cloud Machine Learning will take machine learning mainstream, giving data scientists and developers a way to build a new class of intelligent applications. It provides access to the same technologies that power Google Now, Google Photos and voice recognition in Google Search as easy to use REST APIs.

It enables you to build powerful Machine Learning models on your data using the open-source TensorFlow machine learning library:

Cloud Machine Learning makes it easy for you to build sophisticated, large scale machine learning models in a short amount of time. It’s portable, fully managed and scalable. Cloud Machine Learning works with data in many formats and is well integrated with other Cloud Platform products such as Google Cloud Dataflow, Google BigQuery, Google Cloud Dataproc, Google Cloud Storage, and Google Cloud Datalab. You can easily build predictive analytics models using your own training data. For example, a financial services app that predicts values using regression models, or a classification service for images. Cloud Machine Learning will take care of everything from data ingestion through to prediction. The result: now any application can take advantage of the same deep learning techniques that power many of Google’s services.

Pre-trained Machine Learning models like Google Translate API and Cloud Vision API are being joined today by Google Cloud Speech API. We're very excited to now offer a full set of APIs that help your applications see, hear and translate. Cloud Speech API brings the same advanced neural network technology that powers voice search in the Google app and voice typing in Google Keyboard. It demonstrates speech-to-text conversion in 80+ languages with unparalleled accuracy, especially in noisy environments, and is blazingly fast. The technology that has already empowered developers to add speech to their Chrome and Android is now available for any application in real-time streaming or batch mode.

Google Cloud Speech API enables developers to convert audio to text by applying powerful neural network models in an easy to use API. The API recognizes over 80 languages and variants, to support your global user base. You can transcribe the text of users dictating to an application’s microphone, enable command-and-control through voice, or transcribe audio files, among many other use cases. Recognize audio uploaded in the request, and in upcoming releases, integrate with your audio storage on Google Cloud Storage.

March 24, 2016

Daimler demos truck platooning to save 7% on fuel and has network connected 365000 trucks already

Today, on the A52 autobahn near Düsseldorf, Daimler Trucks presented an impressive example of the possibilities opened up by the digital connection of trucks: Three WiFi-connected, autonomously driving trucks operated on the autobahn with authorisation for public traffic as a so-called platoon. Such a combination can reduce fuel consumption by up to seven percent and the road space requirement on motorways by almost half - while improving traffic safety at the same time. Based on the Daimler Trucks Highway Pilot system for autonomously driving heavy trucks, the three trucks link up to form an aerodynamically optimized, fully automated platoon. Daimler Trucks calls this advanced system development Highway Pilot Connect.

Bernhard: "We are creating a new, highly efficient and open logistical network."

Dr Wolfgang Bernhard, member of the Board of Management of Daimler AG with responsibility for Daimler Trucks & Buses, explains: “We are connecting the truck with the internet – making him the mobile data center of the logistics network. It connects all those involved in goods: drivers, schedulers, fleet operators, workshops, manufacturers and insurance companies or authorities. They receive information in real time which was previously unavailable: about the condition of the tractor unit and semitrailer, traffic and weather conditions, the parking availability at motorway service stations, rest areas and much more."

Magnetic memory chips at theoretical limit for low energy will use one million times less energy per computer operation

In a breakthrough for energy-efficient computing, UC Berkeley engineers have shown for the first time that magnetic chips can actually operate at the lowest fundamental energy dissipation theoretically possible under the laws of thermodynamics. This means that dramatic reductions in power consumption are possible — down to as little as one-millionth the amount of energy per operation used by transistors in modern computers.

This is critical for mobile devices, which demand powerful processors that can run for a day or more on small, lightweight batteries. On a larger, industrial scale, as computing increasingly moves into “the cloud,” the electricity demands of the giant cloud data centers are multiplying, collectively taking an increasing share of the country’s — and world’s — electrical grid.

“We wanted to know how small we could shrink the amount of energy needed for computing,” said senior author Jeffrey Bokor, a UC Berkeley professor of electrical engineering and computer sciences and a faculty scientist at the Lawrence Berkeley National Laboratory. “The biggest challenge in designing computers and, in fact, all our electronics today is reducing their energy consumption.”

Magnetic microscope image of three nanomagnetic computer bits. Each bit is a tiny bar magnet only 90 nanometers long. In the image on the left, the bright spots are at the “North” end of the magnet, and the dark spots are at the “South” end. The “H” arrow shows the direction of magnetic field applied to switch the direction of the magnets. (Image by Jeongmin Hong and Jeffrey Bokor, study in Science Advances.)

Berkeley labs new hybrid membrane has eight times more carbon dioxide permeability which will make carbon capture more efficient

A new, highly permeable carbon capture membrane developed by scientists from the U.S. Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) could lead to more efficient ways of separating carbon dioxide from power plant exhaust, preventing the greenhouse gas from entering the atmosphere and contributing to climate change.

The researchers focused on a hybrid membrane that is part polymer and part metal-organic framework, which is a porous three-dimensional crystal with a large internal surface area that can absorb enormous quantities of molecules.

In a first, the scientists engineered the membrane so that carbon dioxide molecules can travel through it via two distinct channels. Molecules can travel through the polymer component of the membrane, like they do in conventional gas-separation membranes. Or molecules can flow through “carbon dioxide highways” created by adjacent metal-organic frameworks.

Initial tests show this two-route approach makes the hybrid membrane eight times more carbon dioxide permeable than membranes composed only of the polymer.

The hybrid membrane has metal-organic frameworks which account for 50 percent of its weight, which is about 20 percent more than other hybrid membranes. Previously, the mechanical stability of a hybrid membrane limited the amount of metal-organic frameworks that could be packed in it.

Energy and Environmental Science - Enhanced permeation arising from dual transport pathways in hybrid polymer–MOF membranes

Berkeley Lab scientists make major advance in understanding a basic process of life with structure of protein organization at beginning of every gene

Your DNA governs more than just what color your eyes are and whether you can curl your tongue. Your genes contain instructions for making all your proteins, which your cells constantly need to keep you alive. But some key aspects of how that process works at the molecular level have been a bit of a mystery—until now.

Using cryo-electron microscopy (cryo-EM), Lawrence Berkeley National Laboratory (Berkeley Lab) scientist Eva Nogales and her team have made a significant breakthrough in our understanding of how our molecular machinery finds the right DNA to copy, showing with unprecedented detail the role of a powerhouse transcription factor known as TFIID.

Berkeley Lab scientists Eva Nogales and Robert Louder at the electron microscope. (Credit: Roy Kaltschmidt/Berkeley Lab)

This finding is important as it paves the way for scientists to understand and treat a host of malignancies. “Understanding this regulatory process in the cell is the only way to manipulate it or fix it when it goes bad,” said Nogales. “Gene expression is at the heart of many essential biological processes, from embryonic development to cancer. One day we’ll be able to manipulate these fundamental mechanisms, either to correct for expression of genes that should or should not be present or to take care of malignant states where the process has gone out of control.”

“We now have the structure of the whole protein organization that is formed at the beginning of every gene. This is something no one has come close to doing because it is really very difficult to study by traditional methodologies.”

Nature - Structure of promoter-bound TFIID and model of human pre-initiation complex assembly

Areva joint development with Lightbridge on metallic fuel that will enable 17% uprates of existing reactors and China pebble bed and accelerator driven reactors

1. The first of two reactor pressure vessels has been installed at the demonstration HTR-PM high-temperature gas-cooled reactor unit under construction at Shidaowan in China's Shandong province. The twin-reactor unit is scheduled to start up next year.

The vessel - about 25 meters in height and weighing about 700 tonnes - was manufactured by Shanghai Electric Nuclear Power Equipment. It successfully completed factory acceptance on 29 February and was dispatched from the manufacturing plant on 2 March.

The demonstration plant's twin HTR-PM reactors will drive a single 210 MWe turbine. It is expected to start commercial operation in late 2017. An earlier proposal was for 18 further 210 MWe units - giving a total capacity of 3800 MWe - at the Shidaowan site, near Rongcheng in Weihai city, but this has been dropped.

A proposal to construct two 600 MWe HTR plants - each featuring three twin reactor and turbine units - at Ruijin city in China's Jiangxi province passed a preliminary feasibility review in early 2015. The design of the Ruijin HTRs is based on the smaller Shidaowan demonstration HTR-PM. Construction of the Ruijin reactors is expected to start next year, with grid connection in 2021.

The vessel is hoisted above the unit's reactor building (Image: China Huaneng)

2. Lightbridge Corporation has entered into a Joint Development Agreement with Areva NP to assess establishing a joint venture this year for the development, manufacture and commercialization of fuel assemblies based on Lightbridge's "next generation" metallic nuclear fuel technology. The parties will share the cost of the work scope to be performed under the JDA, with Areva contributing in-kind for its share of the costs.

Lightbridge president and CEO Seth Grae said Areva "has the resources and expertise to enable global deployment of our metallic fuel in commercial reactors". Grae added that the Nuclear Utility Fuel Advisory Board - comprising nuclear fuel and regulatory experts from Dominion Resources, Southern Nuclear Operating Company, Duke Energy, and Exelon Generation - have requested that the US Nuclear Regulatory Commission prepare to receive initial regulatory licensing documentation in 2017.

Lightbridge’s all metal fuel (AMF) assembly is comprised entirely of metallic fuel rods and is capable of providing up to 17% increase in power output in existing PWRs and up to a 30% power uprate in new build PWRs operating on 18-month fuel cycles. Due to certain constraints associated with the size of equipment that can fit in the containment structures of existing PWRs, there are limits as to the maximum power uprate level existing PWRs can accommodate without changing their existing containment structure. However, a new build unit can be constructed with a larger containment to allow for higher capacity equipment with relatively small capital cost increase.

Lightbridge is developing three primary nuclear fuel product offerings for power uprates and longer fuel cycles:

  • LTB17-1024™ all-metal fuel for up to 10% power uprates and 24-month operating cycles in existing PWRs;
  • LTB17-1718 1718™ all-metal fuel for up to 17% power uprates and 1818-month operating cycles in existing PWRs; and
  • LTB17-3018™ all-metal fuel for up to 30% power uprates and 18-month operating cycles in new-build PWRs.
  • In addition, Lightbridge is developing LTB17-Th18™, our™ thorium-based seed and blanket fuel, which offers significant back-end advantages and enhanced proliferation resistance of used fuel.

China's 5 year plan for doubling nuclear power generation

Under the latest Five-Year Plan, China should have some 58 GWe of nuclear generating capacity in operation by 2020, up from the current capacity of almost 27 GWe. In addition, a further 30 GWe of nuclear capacity will be under construction by 2020.

The 13th Five Year Plan also calls for the construction of a demonstration as well as a large commercial scale reprocessing plant to be accelerated. China National Nuclear Corporation (CNNC) and France's Areva signed an agreement in November 2007 to assess the feasibility of setting up an 800 tonne per year reprocessing plant for used fuel in China.

The latest Five-Year Plan also calls for construction of Phase III of the Tianwan plant in Jiangsu province to be accelerated. The State Council gave its approval for Tianwan units 5 and 6 - featuring 1080 MWe ACPR1000 reactors - in December. Tianwan units 1 to 4 are Russian-supplied VVER reactors.

The plan also calls for the preparation for construction of inland nuclear power plant

First order 657 Joint Light Tactical Vehicles and plans for 17,000

Oshkosh Defense has received the first production order for the 657 Joint Light Tactical Vehicle (JLTV) vehicles. The order, issued by the U.S. Army is valued $243 million. It also includes 2,977 installed kits and related support. The order will serve both the U.S. Army and Marine Corps. The JLTV production contract calls for Oshkosh to deliver a total of nearly 17,000 vehicles, as well as kits and services over an eight-year period with first vehicle delivery in October 2016.

The base L-ATV does not have a standard armament, however it can be fitted with a selection of weapons including light, medium, and heavy machine guns, automatic grenade launchers, or anti-tank guided missiles (ATGMs) depending on user requirements. The weapons can be operated from ring mounts or a remote weapon station. Smoke grenade launchers for self-defense can also be fitted if required.

During the L-ATV design process, every component was optimized for survivability, resulting in the same level of protection in a vehicle 30 percent smaller. This resulted in a curb weight for the JLTV requirement of 14,000 lb (6,400 kg), almost one-third the weight of the heavier MRAP (4x4) models, and almost half the weight of the original MRAP models. Payload allowance for the JLTV in Combat Tactical Vehicle (CTV) configuration is four passengers and 3,500 lb (1,600 kg) of cargo, and in Combat Support Vehicle (CSV) configuration is two passengers and 5,100 lb (2,300 kg) of cargo

Up Armored Hummers weigh about 2.6 tons
the JLTV weighs 6.4 tons
MRAPS weigh 14-18 tons

Oshkosh's JLTV will deliver a level of protection similar to that of current, but far heavier and less maneuverable, Mine Resistant Ambush Protected (MRAP) class designs these having far more protection from blast than even the latest up-armored HMMWVs. The cost is similar to up armored hummers.

March 23, 2016

China plans to launch Tiangong 2 space laboratory this year and complete the full space station by 2020

China plans to initiate 20 space missions this year including the testing of its most sophisticated rocket and the launching of a habitual space module. But security analysts in the US are expressing suspicions they have military implications.

China is already testing surface-to-air missiles that could strike targets in orbit and it's also working with experimental lasers that can scramble or "blind" satellites.

This year, China intends to launch the Tiangong 2 space laboratory. China plans to later send astronauts to this habitual module, which will serve as a stepping stone toward a major space station. Chinese officials plan to have this station fully operational by 2020. Some security analysts fear the station will have military implications.

China still trails far behind Russia and the US in terms of space technology. But in the last 12 years, China has sought to rapidly reduce the gaps it holds with both nations, and it's showing no intentions of slowing down.

March 22, 2016

African researcher wants to bring tablet-based AI teaching software to Ethiopian children

African researchers have recently launched the YaNetu teaching tablet crowdfunding project. This effort aims to bring an AI based educational tablet to African children. The researchers hope to create:

- An Android-based teaching tablet for primary school age children in the developing world, with both offline and online applications

- A built-in curriculum, customized with local languages, designed to grow and develop over the years along with the child

- Artificial Intelligence systems, represented by human-like avatars, designed in collaboration with leading American AI researcher Dr. Ben Goertzel. Our AI avatars offer the student not only information and coaching, but also emotional and motivational feedback.

In an interview for Next Big Future with Sander Olson, iCOG researcher Hruy Tsegave describes why he believes that teaching tablets could be an effective and efficient method for providing large numbers of African children with a versatile and compelling teaching tool.

Question: Tell us about yourself. What is your background, and where are you located?

My Name is Hruy Tsegaye. I live in Addis Ababa the capital city of Ethiopia. I got my BA degree in Journalism and Literature. As a fresh graduate, I was looking for adventure and I went to a very remote region in Ethiopia called Afar. I began my professional career in that wretched place as a Program Officer for a governmental agency called HIV AIDS Prevention and Control Office (HAPCO). I had worked there for two years and by the end of the second year, I decided I had had enough! The extent of extreme poverty was unmatched and the depressing tone of the environment which you can feel floating in the burning desert wind was too much for any 22 years old novice even if there is this wild urge for adventure. I come back to the capital and Joined Fortune Newspaper as a junior reporter. Working in the largest English weekly newspaper has its merits when it comes to understanding whats going on in your country. After two years, an irreconcilable conflict of interest with my writing and the newspaper forced me to leave that paper. After that, I met a guy in the tech business and he convinced me that the future for the tech business in Ethiopia is bright. I started working for a company called 4Afri Mobile Technologies as Head of Idea Development. The Start Up Company was doing well but after 16 months the CEO (who is also the owner) decided to change directions and liquidated 4Afri in a move to invest in the country's Energy Sector. I was not interested in that; the tech world had already become my new love. A friend of mine told me about iCog and I joined the company as its 4th permanent staff and I can say that I am one of the founders.

Alphago explained by Demi sHassabis

Demi explains the policy and value neural networks.

Policy is neural network trained to make reasonable moves based upon supervised learning of 100,000 games.

Value network is built from tens of millions of games to be able to determine what winning positions are.
The value scoring of positions was previously believed to be impossible.

Alphago used policy to reduce the breadth of search
Alphago used value network to reduce the depth of the search for good moves.

Alphago also uses fast rollouts to play a few thousand games to determine statistics for which moves are good.

The Alphago program is improving by several levels every few months.

Форма для связи


Email *

Message *