Tag Archives: blue
Neural networks are powerful things, but they need a lot of juice. Engineers at MIT have now developed a new chip that cuts neural nets’ power consumption by up to 95 percent, potentially allowing them to run on battery-powered mobile devices.
Smartphones these days are getting truly smart, with ever more AI-powered services like digital assistants and real-time translation. But typically the neural nets crunching the data for these services are in the cloud, with data from smartphones ferried back and forth.
That’s not ideal, as it requires a lot of communication bandwidth and means potentially sensitive data is being transmitted and stored on servers outside the user’s control. But the huge amounts of energy needed to power the GPUs neural networks run on make it impractical to implement them in devices that run on limited battery power.
Engineers at MIT have now designed a chip that cuts that power consumption by up to 95 percent by dramatically reducing the need to shuttle data back and forth between a chip’s memory and processors.
Neural nets consist of thousands of interconnected artificial neurons arranged in layers. Each neuron receives input from multiple neurons in the layer below it, and if the combined input passes a certain threshold it then transmits an output to multiple neurons above it. The strength of the connection between neurons is governed by a weight, which is set during training.
This means that for every neuron, the chip has to retrieve the input data for a particular connection and the connection weight from memory, multiply them, store the result, and then repeat the process for every input. That requires a lot of data to be moved around, expending a lot of energy.
The new MIT chip does away with that, instead computing all the inputs in parallel within the memory using analog circuits. That significantly reduces the amount of data that needs to be shoved around and results in major energy savings.
The approach requires the weights of the connections to be binary rather than a range of values, but previous theoretical work had suggested this wouldn’t dramatically impact accuracy, and the researchers found the chip’s results were generally within two to three percent of the conventional non-binary neural net running on a standard computer.
This isn’t the first time researchers have created chips that carry out processing in memory to reduce the power consumption of neural nets, but it’s the first time the approach has been used to run powerful convolutional neural networks popular for image-based AI applications.
“The results show impressive specifications for the energy-efficient implementation of convolution operations with memory arrays,” Dario Gil, vice president of artificial intelligence at IBM, said in a statement.
“It certainly will open the possibility to employ more complex convolutional neural networks for image and video classifications in IoT [the internet of things] in the future.”
It’s not just research groups working on this, though. The desire to get AI smarts into devices like smartphones, household appliances, and all kinds of IoT devices is driving the who’s who of Silicon Valley to pile into low-power AI chips.
Apple has already integrated its Neural Engine into the iPhone X to power things like its facial recognition technology, and Amazon is rumored to be developing its own custom AI chips for the next generation of its Echo digital assistant.
The big chip companies are also increasingly pivoting towards supporting advanced capabilities like machine learning, which has forced them to make their devices ever more energy-efficient. Earlier this year ARM unveiled two new chips: the Arm Machine Learning processor, aimed at general AI tasks from translation to facial recognition, and the Arm Object Detection processor for detecting things like faces in images.
Qualcomm’s latest mobile chip, the Snapdragon 845, features a GPU and is heavily focused on AI. The company has also released the Snapdragon 820E, which is aimed at drones, robots, and industrial devices.
Going a step further, IBM and Intel are developing neuromorphic chips whose architectures are inspired by the human brain and its incredible energy efficiency. That could theoretically allow IBM’s TrueNorth and Intel’s Loihi to run powerful machine learning on a fraction of the power of conventional chips, though they are both still highly experimental at this stage.
Getting these chips to run neural nets as powerful as those found in cloud services without burning through batteries too quickly will be a big challenge. But at the current pace of innovation, it doesn’t look like it will be too long before you’ll be packing some serious AI power in your pocket.
Image Credit: Blue Planet Studio / Shutterstock.com Continue reading
I don’t have to open the doors of AImotive’s white 2015 Prius to see that it’s not your average car. This particular Prius has been christened El Capitan, the name written below the rear doors, and two small cameras are mounted on top of the car. Bundles of wire snake out from them, as well as from the two additional cameras on the car’s hood and trunk.
Inside is where things really get interesting, though. The trunk holds a computer the size of a microwave, and a large monitor covers the passenger glove compartment and dashboard. The center console has three switches labeled “Allowed,” “Error,” and “Active.”
Budapest-based AImotive is working to provide scalable self-driving technology alongside big players like Waymo and Uber in the autonomous vehicle world. On a highway test ride with CEO Laszlo Kishonti near the company’s office in Mountain View, California, I got a glimpse of just how complex that world is.
Camera-Based Feedback System
AImotive’s approach to autonomous driving is a little different from that of some of the best-known systems. For starters, they’re using cameras, not lidar, as primary sensors. “The traffic system is visual and the cost of cameras is low,” Kishonti said. “A lidar can recognize when there are people near the car, but a camera can differentiate between, say, an elderly person and a child. Lidar’s resolution isn’t high enough to recognize the subtle differences of urban driving.”
Image Credit: AImotive
The company’s aiDrive software uses data from the camera sensors to feed information to its algorithms for hierarchical decision-making, grouped under four concurrent activities: recognition, location, motion, and control.
Kishonti pointed out that lidar has already gotten more cost-efficient, and will only continue to do so.
“Ten years ago, lidar was best because there wasn’t enough processing power to do all the calculations by AI. But the cost of running AI is decreasing,” he said. “In our approach, computer vision and AI processing are key, and for safety, we’ll have fallback sensors like radar or lidar.”
aiDrive currently runs on Nvidia chips, which Kishonti noted were originally designed for graphics, and are not terribly efficient given how power-hungry they are. “We’re planning to substitute lower-cost, lower-energy chips in the next six months,” he said.
Testing in Virtual Reality
Waymo recently announced its fleet has now driven four million miles autonomously. That’s a lot of miles, and hard to compete with. But AImotive isn’t trying to compete, at least not by logging more real-life test miles. Instead, the company is doing 90 percent of its testing in virtual reality. “This is what truly differentiates us from competitors,” Kishonti said.
He outlined the three main benefits of VR testing: it can simulate scenarios too dangerous for the real world (such as hitting something), too costly (not every company has Waymo’s funds to run hundreds of cars on real roads), or too time-consuming (like waiting for rain, snow, or other weather conditions to occur naturally and repeatedly).
“Real-world traffic testing is very skewed towards the boring miles,” he said. “What we want to do is test all the cases that are hard to solve.”
On a screen that looked not unlike multiple games of Mario Kart, he showed me the simulator. Cartoon cars cruised down winding streets, outfitted with all the real-world surroundings: people, trees, signs, other cars. As I watched, a furry kangaroo suddenly hopped across one screen. “Volvo had an issue in Australia,” Kishonti explained. “A kangaroo’s movement is different than other animals since it hops instead of running.” Talk about cases that are hard to solve.
AImotive is currently testing around 1,000 simulated scenarios every night, with a steadily-rising curve of successful tests. These scenarios are broken down into features, and the car’s behavior around those features fed into a neural network. As the algorithms learn more features, the level of complexity the vehicles can handle goes up.
On the Road
After Kishonti and his colleagues filled me in on the details of their product, it was time to test it out. A safety driver sat in the driver’s seat, a computer operator in the passenger seat, and Kishonti and I in back. The driver maintained full control of the car until we merged onto the highway. Then he flicked the “Allowed” switch, his copilot pressed the “Active” switch, and he took his hands off the wheel.
What happened next, you ask?
A few things. El Capitan was going exactly the speed limit—65 miles per hour—which meant all the other cars were passing us. When a car merged in front of us or cut us off, El Cap braked accordingly (if a little abruptly). The monitor displayed the feed from each of the car’s cameras, plus multiple data fields and a simulation where a blue line marked the center of the lane, measured by the cameras tracking the lane markings on either side.
I noticed El Cap wobbling out of our lane a bit, but it wasn’t until two things happened in a row that I felt a little nervous: first we went under a bridge, then a truck pulled up next to us, both bridge and truck casting a complete shadow over our car. At that point El Cap lost it, and we swerved haphazardly to the right, narrowly missing the truck’s rear wheels. The safety driver grabbed the steering wheel and took back control of the car.
What happened, Kishonti explained, was that the shadows made it hard for the car’s cameras to see the lane markings. This was a new scenario the algorithm hadn’t previously encountered. If we’d only gone under a bridge or only been next to the truck for a second, El Cap may not have had so much trouble, but the two events happening in a row really threw the car for a loop—almost literally.
“This is a new scenario we’ll add to our testing,” Kishonti said. He added that another way for the algorithm to handle this type of scenario, rather than basing its speed and positioning on the lane markings, is to mimic nearby cars. “The human eye would see that other cars are still moving at the same speed, even if it can’t see details of the road,” he said.
After another brief—and thankfully uneventful—hands-off cruise down the highway, the safety driver took over, exited the highway, and drove us back to the office.
Driving into the Future
I climbed out of the car feeling amazed not only that self-driving cars are possible, but that driving is possible at all. I squint when driving into a tunnel, swerve to avoid hitting a stray squirrel, and brake gradually at stop signs—all without consciously thinking to do so. On top of learning to steer, brake, and accelerate, self-driving software has to incorporate our brains’ and bodies’ unconscious (but crucial) reactions, like our pupils dilating to let in more light so we can see in a tunnel.
Despite all the progress of machine learning, artificial intelligence, and computing power, I have a wholly renewed appreciation for the thing that’s been in charge of driving up till now: the human brain.
Kishonti seemed to feel similarly. “I don’t think autonomous vehicles in the near future will be better than the best drivers,” he said. “But they’ll be better than the average driver. What we want to achieve is safe, good-quality driving for everyone, with scalability.”
AImotive is currently working with American tech firms and with car and truck manufacturers in Europe, China, and Japan.
Image Credit: Alex Oakenman / Shutterstock.com Continue reading
Swarms of drones buzz overhead, while robotic vehicles crawl across the landscape. Orbiting satellites snap high-resolution images of the scene far below. Not one human being can be seen in the pre-dawn glow spreading across the land.
This isn’t some post-apocalyptic vision of the future à la The Terminator. This is a snapshot of the farm of the future. Every phase of the operation—from seed to harvest—may someday be automated, without the need to ever get one’s fingernails dirty.
In fact, it’s science fiction already being engineered into reality. Today, robots empowered with artificial intelligence can zap weeds with preternatural precision, while autonomous tractors move with tireless efficiency across the farmland. Satellites can assess crop health from outer space, providing gobs of data to help produce the sort of business intelligence once accessible only to Fortune 500 companies.
“Precision agriculture is on the brink of a new phase of development involving smart machines that can operate by themselves, which will allow production agriculture to become significantly more efficient. Precision agriculture is becoming robotic agriculture,” said professor Simon Blackmore last year during a conference in Asia on the latest developments in robotic agriculture. Blackmore is head of engineering at Harper Adams University and head of the National Centre for Precision Farming in the UK.
It’s Blackmore’s university that recently showcased what may someday be possible. The project, dubbed Hands Free Hectare and led by researchers from Harper Adams and private industry, farmed one hectare (about 2.5 acres) of spring barley without one person ever setting foot in the field.
The team re-purposed, re-wired and roboticized farm equipment ranging from a Japanese tractor to a 25-year-old combine. Drones served as scouts to survey the operation and collect samples to help the team monitor the progress of the barley. At the end of the season, the robo farmers harvested about 4.5 tons of barley at a price tag of £200,000.
“This project aimed to prove that there’s no technological reason why a field can’t be farmed without humans working the land directly now, and we’ve done that,” said Martin Abell, mechatronics researcher for Precision Decisions, which partnered with Harper Adams, in a press release.
I, Robot Farmer
The Harper Adams experiment is the latest example of how machines are disrupting the agricultural industry. Around the same time that the Hands Free Hectare combine was harvesting barley, Deere & Company announced it would acquire a startup called Blue River Technology for a reported $305 million.
Blue River has developed a “see-and-spray” system that combines computer vision and artificial intelligence to discriminate between crops and weeds. It hits the former with fertilizer and blasts the latter with herbicides with such precision that it can eliminate 90 percent of the chemicals used in conventional agriculture.
It’s not just farmland that’s getting a helping hand from robots. A California company called Abundant Robotics, spun out of the nonprofit research institute SRI International, is developing robots capable of picking apples with vacuum-like arms that suck the fruit straight off the trees in the orchards.
“Traditional robots were designed to perform very specific tasks over and over again. But the robots that will be used in food and agricultural applications will have to be much more flexible than what we’ve seen in automotive manufacturing plants in order to deal with natural variation in food products or the outdoor environment,” Dan Harburg, an associate at venture capital firm Anterra Capital who previously worked at a Massachusetts-based startup making a robotic arm capable of grabbing fruit, told AgFunder News.
“This means ag-focused robotics startups have to design systems from the ground up, which can take time and money, and their robots have to be able to complete multiple tasks to avoid sitting on the shelf for a significant portion of the year,” he noted.
Eyes in the Sky
It will take more than an army of robotic tractors to grow a successful crop. The farm of the future will rely on drones, satellites, and other airborne instruments to provide data about their crops on the ground.
Companies like Descartes Labs, for instance, employ machine learning to analyze satellite imagery to forecast soy and corn yields. The Los Alamos, New Mexico startup collects five terabytes of data every day from multiple satellite constellations, including NASA and the European Space Agency. Combined with weather readings and other real-time inputs, Descartes Labs can predict cornfield yields with 99 percent accuracy. Its AI platform can even assess crop health from infrared readings.
The US agency DARPA recently granted Descartes Labs $1.5 million to monitor and analyze wheat yields in the Middle East and Africa. The idea is that accurate forecasts may help identify regions at risk of crop failure, which could lead to famine and political unrest. Another company called TellusLabs out of Somerville, Massachusetts also employs machine learning algorithms to predict corn and soy yields with similar accuracy from satellite imagery.
Farmers don’t have to reach orbit to get insights on their cropland. A startup in Oakland, Ceres Imaging, produces high-resolution imagery from multispectral cameras flown across fields aboard small planes. The snapshots capture the landscape at different wavelengths, identifying insights into problems like water stress, as well as providing estimates of chlorophyll and nitrogen levels. The geo-tagged images mean farmers can easily locate areas that need to be addressed.
Growing From the Inside
Even the best intelligence—whether from drones, satellites, or machine learning algorithms—will be challenged to predict the unpredictable issues posed by climate change. That’s one reason more and more companies are betting the farm on what’s called controlled environment agriculture. Today, that doesn’t just mean fancy greenhouses, but everything from warehouse-sized, automated vertical farms to grow rooms run by robots, located not in the emptiness of Kansas or Nebraska but smack dab in the middle of the main streets of America.
Proponents of these new concepts argue these high-tech indoor farms can produce much higher yields while drastically reducing water usage and synthetic inputs like fertilizer and herbicides.
Iron Ox, out of San Francisco, is developing one-acre urban greenhouses that will be operated by robots and reportedly capable of producing the equivalent of 30 acres of farmland. Powered by artificial intelligence, a team of three robots will run the entire operation of planting, nurturing, and harvesting the crops.
Vertical farming startup Plenty, also based in San Francisco, uses AI to automate its operations, and got a $200 million vote of confidence from the SoftBank Vision Fund earlier this year. The company claims its system uses only 1 percent of the water consumed in conventional agriculture while producing 350 times as much produce. Plenty is part of a new crop of urban-oriented farms, including Bowery Farming and AeroFarms.
“What I can envision is locating a larger scale indoor farm in the economically disadvantaged food desert, in order to stimulate a broader economic impact that could create jobs and generate income for that area,” said Dr. Gary Stutte, an expert in space agriculture and controlled environment agriculture, in an interview with AgFunder News. “The indoor agriculture model is adaptable to becoming an engine for economic growth and food security in both rural and urban food deserts.”
Still, the model is not without its own challenges and criticisms. Most of what these farms can produce falls into the “leafy greens” category and often comes with a premium price, which seems antithetical to the proposed mission of creating oases in the food deserts of cities. While water usage may be minimized, the electricity required to power the operation, especially the LEDs (which played a huge part in revolutionizing indoor agriculture), are not cheap.
Still, all of these advances, from robo farmers to automated greenhouses, may need to be part of a future where nearly 10 billion people will inhabit the planet by 2050. An oft-quoted statistic from the Food and Agriculture Organization of the United Nations says the world must boost food production by 70 percent to meet the needs of the population. Technology may not save the world, but it will help feed it.
Image Credit: Valentin Valkov / Shutterstock.com Continue reading
When it comes to biomolecules, RNA doesn’t get a lot of love.
Maybe you haven’t even heard of the silent workhorse. RNA is the cell’s de facto translator: like a game of telephone, RNA takes DNA’s genetic code to a cellular factory called ribosomes. There, the cell makes proteins based on RNA’s message.
But RNA isn’t just a middleman. It controls what proteins are formed. Because proteins wiz around the cell completing all sorts of important processes, you can say that RNA is the gatekeeper: no RNA message, no proteins, no life.
In a new study published in Nature, RNA finally took center stage. By adding bits of genetic material to the E. Coli bacteria, a team of biohackers at the Wyss Institute hijacked the organism’s RNA messengers so that they only spring into action following certain inputs.
The result? A bacterial biocomputer capable of performing 12-input logic operations—AND, OR, and NOT—following specific inputs. Rather than outputting 0s and 1s, these biocircuits produce results based on the presence or absence of proteins and other molecules.
“It’s the greatest number of inputs in a circuit that a cell has been able to process,” says study author Dr. Alexander Green at Arizona State University. “To be able to analyze those signals and make a decision is the big advance here.”
When given a specific set of inputs, the bacteria spit out a protein that made them glow neon green under fluorescent light.
But synthetic biology promises far more than just a party trick—by tinkering with a cell’s RNA repertoire, scientists may one day coax them to photosynthesize, produce expensive drugs on the fly, or diagnose and hunt down rogue tumor cells.
Illustration of an RNA-based ‘ribocomputing’ device that makes logic-based decisions in living cells. The long gate RNA (blue) detects the binding of an input RNA (red). The ribosome (purple/mauve) reads the gate RNA to produce an output protein. Image Credit: Alexander Green / Arizona State University
The software of life
This isn’t the first time that scientists hijacked life’s algorithms to reprogram cells into nanocomputing systems. Previous work has already introduced to the world yeast cells that can make anti-malaria drugs from sugar or mammalian cells that can perform Boolean logic.
Yet circuits with multiple inputs and outputs remain hard to program. The reason is this: synthetic biologists have traditionally focused on snipping, fusing, or otherwise arranging a cell’s DNA to produce the outcomes they want.
But DNA is two steps removed from proteins, and tinkering with life’s code often leads to unexpected consequences. For one, the cell may not even accept and produce the extra bits of DNA code. For another, the added code, when transformed into proteins, may not act accordingly in the crowded and ever-changing environment of the cell.
What’s more, tinkering with one gene is often not enough to program an entirely new circuit. Scientists often need to amp up or shut down the activity of multiple genes, or multiple biological “modules” each made up of tens or hundreds of genes.
It’s like trying to fit new Lego pieces in a specific order into a room full of Lego constructions. Each new piece has the potential to wander off track and click onto something it’s not supposed to touch.
Getting every moving component to work in sync—as you might have guessed—is a giant headache.
The RNA way
With “ribocomputing,” Green and colleagues set off to tackle a main problem in synthetic biology: predictability.
Named after the “R (ribo)” in “RNA,” the method grew out of an idea that first struck Green back in 2012.
“The synthetic biological circuits to date have relied heavily on protein-based regulators that are difficult to scale up,” Green wrote at the time. We only have a limited handful of “designable parts” that work well, and these circuits require significant resources to encode and operate, he explains.
RNA, in comparison, is a lot more predictable. Like its more famous sibling DNA, RNA is composed of units that come in four different flavors: A, G, C, and U. Although RNA is only single-stranded, rather than the double helix for which DNA is known for, it can bind short DNA-like sequences in a very predictable manner: Gs always match up with Cs and As always with Us.
Because of this predictability, it’s possible to design RNA components that bind together perfectly. In other words, it reduces the chance that added RNA bits might go rogue in an unsuspecting cell.
Normally, once RNA is produced it immediately rushes to the ribosome—the cell’s protein-building factory. Think of it as a constantly “on” system.
However, Green and his team found a clever mechanism to slow them down. Dubbed the “toehold switch,” it works like this: the artificial RNA component is first incorporated into a chain of A, G, C, and U folded into a paperclip-like structure.
This blocks the RNA from accessing the ribosome. Because one RNA strand generally maps to one protein, the switch prevents that protein from ever getting made.
In this way, the switch is set to “off” by default—a “NOT” gate, in Boolean logic.
To activate the switch, the cell needs another component: a “trigger RNA,” which binds to the RNA toehold switch. This flips it on: the RNA grabs onto the ribosome, and bam—proteins.
String a few RNA switches together, with the activity of each one relying on the one before, and it forms an “AND” gate. Alternatively, if the activity of each switch is independent, that’s an “OR” gate.
“Basically, the toehold switches performed so well that we wanted to find a way to best exploit them for cellular applications,” says Green. They’re “kind of the equivalent of your first transistors,” he adds.
Once the team optimized the designs for different logic gates, they carefully condensed the switches into “gate RNA” molecules. These gate RNAs contain both codes for proteins and the logic operations needed to kickstart the process—a molecular logic circuit, so to speak.
If you’ve ever played around with an Arduino-controlled electrical circuit, you probably know the easiest way to test its function is with a light bulb.
That’s what the team did here, though with a biological bulb: green fluorescent protein, a light-sensing protein not normally present in bacteria that—when turned on—makes the microbugs glow neon green.
In a series of experiments, Green and his team genetically inserted gate RNAs into bacteria. Then, depending on the type of logical function, they added different combinations of trigger RNAs—the inputs.
When the input RNA matched up with its corresponding gate RNA, it flipped on the switch, causing the cell to light up.
Their most complex circuit contained five AND gates, five OR gates, and two NOTs—a 12-input ribocomputer that functioned exactly as designed.
That’s quite the achievement. “Everything is interacting with everything else and there are a million ways those interactions could flip the switch on accident,” says RNA researcher Dr. Julies Lucks at Northwestern University.
The specificity is thanks to RNA, the authors explain. Because RNAs bind to others so predictably, we can now design massive libraries of gate and trigger units to mix-and-match into all types of nano-biocomputers.
Although the technology doesn’t have any immediate applications, the team has high hopes.
For the first time, it’s now possible to massively scale up the process of programming new circuits into living cells. We’ve expanded the library of available biocomponents that can be used to reprogram life’s basic code, the authors say.
What’s more, when freeze-dried onto a piece of tissue paper, RNA keeps very well. We could potentially print RNA toehold switches onto paper that respond to viruses or to tumor cells, the authors say, essentially transforming the technology into highly accurate diagnostic platforms.
But Green’s hopes are even wilder for his RNA-based circuits.
“Because we’re using RNA, a universal molecule of life, we know these interactions can also work in other cells, so our method provides a general strategy that could be ported to other organisms,” he says.
Ultimately, the hope is to program neural network-like capabilities into the body’s other cells.
Imagine cells endowed with circuits capable of performing the kinds of computation the brain does, the authors say.
Perhaps one day, synthetic biology will transform our own cells into fully programmable entities, turning us all into biological cyborgs from the inside. How wild would that be?
Image Credit: Wyss Institute at Harvard University Continue reading