MachineLearning
Latest
You'll never believe what neural networks can do now
Clickbait headlines are the lowest form of journalism, but could they be written by a machine? After all, the Associated Press is using one to write complete financial articles, terrible as they are. Developer Lars Eidnes figured that "if this sort of writing truly is formulaic and unoriginal, we should be able to produce it automatically." Rather than building another Upworthy-style headline generator, however, Eidnes took it up a notch by enlisting a so-called recurrent neural network (RNN). That's the same type of machine learning used by SwiftKey, for one, on its beta SwiftKey Neural word-prediction app.
Robots can learn from their mistakes in real-time
Robots and other artificial intelligences can already learn from their mistakes, but they typically have to pause what they're doing to process what happened. They might not have to take a break in the future, though. Researchers have patented a technique, Integral Reinforcement Learning, that has devices continuously refining their actions based on each previous decision. If a machine doesn't already know the optimal way to handle a task, it can keep walking through the scenario (whether by predicting the outcome or actually trying) until it gets things right.
Qualcomm's Snapdragon 820 uses machine learning to fight malware
The trouble with security software is that it's always racing to catch up with new viruses and malware. They typically check against a database of known issues to protect you, which isn't very useful for brand spanking new attacks. Qualcomm is trying to fix that with its Smart Protect technology, which uses machine learning to keep an eye out for potential security issues in real time. Instead of relying on a static list of threats to protect you, it'll actually watch out for suspicious app behavior. Smart Protect will debut on Qualcomm's upcoming Snapdragon 820 mobile processor, details of which it's slowly leaking out. We already know the Snapdragon 820 will have faster graphics capabilities, for example, making it ideal for VR solutions. Naturally, Qualcomm is also offering an API for the new Smart Protect feature, allowing security software companies to take advantage of the new chip's heightened awareness.
Computers can categorize buildings into architectural styles
Even if you've never heard of "Byzantine," you can probably tell a Byzantine church from a Gothic one. Judging style differences is nearly impossible for a computer, however, and researchers from the University of Massachusetts want to fill in that gap. They used geometric matching, crowdsourcing and machine learning to teach an algorithm how to spot similar styles in buildings, furniture and other objects. That's something that could be incredibly useful for historians with mountains of photo archives, or game designers who need to auto-fill a level with historically accurate furniture.
Microsoft attempts to teach computers how to make a funny
Computers and artificial intelligence systems have long struggled with a human understanding of humor – as anyone who has ever asked Siri to tell a joke well knows. Bloomberg reports that recently, a researcher at Microsoft began working with The New Yorker on a project that aims to teach an AI system what is and what is not "funny."
Microsoft wants you to teach computers how to learn
As clever as learning computers may be, they only have as much potential as their software. What if you don't have the know-how to program one of these smart systems yourself? That's where Microsoft Research thinks it can help: it's developing a machine teaching tool that will let most anyone show computers how to learn. So long as you're knowledgeable about your field, you'd just have to plug in the right parameters. A chef could tell a computer how to create tasty recipes, for example, while a doctor could get software to sift through medical records and find data relevant to a new patient.
Military AI interface helps you make sense of thousands of photos
It's easy to find computer vision technology that detect objects in photos, but it's still tough to sift through photos... and that's a big challenge for the military, where finding the right picture could mean taking out a target or spotting a terrorist threat. Thankfully, the US' armed forces may soon have a way to not only spot items in large image libraries, but help human observers find them. DARPA's upcoming, artificial intelligence-backed Visual Media Reasoning system both detects what's in a shot and presents it in a simple interface that bunches photos and videos together based on patterns. If you want to know where a distinctive-looking car has been, for example, you might only need to look in a single group.
Facebook and Google get neural networks to create art
For Facebook and Google, it's not enough for computers to recognize images... they should create images, too. Both tech firms have just shown off neural networks that automatically generate pictures based on their understanding of what objects look like. Facebook's approach uses two of these networks to produce tiny thumbnail images. The technique is much like what you'd experience if you learned painting from a harsh (if not especially daring) critic. The first algorithm creates pictures based on a random vector, while the second checks them for realistic objects and rejects the fake-looking shots; over time, you're left with the most convincing results. The current output is good enough that 40 percent of pictures fooled human viewers, and there's a chance that they'll become more realistic with further refinements.
Amazon uses machine learning to show you more helpful reviews
Let's be blunt: Amazon's reviews sometimes suck. Many of them are hasty day-one reactions, others are horribly misinformed and a few are out-and-out fakes. The internet shopping giant thinks it knows how to sort the wheat from the chaff, however. It just launched a new machine learning system that understands which reviews are likely to be the most helpful, and floats them to the top. The artificial intelligence typically prefers reviews that are recent, receive a lot of up-votes or come from verified buyers. Amazon hopes that this will show you opinions that are not only more trustworthy, but reflect any fixes. In other words, you'll see reviews for the product you're actually likely to get.
Twitter buys a machine learning company to better study your tweets
Twitter thrives on its ability to understand both your tweets and the hot topic of the day, and it needs every bit of help it can get -- including from computers. Accordingly, the social network just snapped up Whetlab, a startup that makes it easier to implement machine learning (aka a form of artificial intelligence). The two companies are shy about what the acquisition means besides an improvement to Twitter's "internal machine learning efforts." However, the likely focus is on highlighting the content that's most relevant to you based on your activity and who you follow, as well as hiding abusive tweets before you have to reach for the "block" option. Whetlab's technology could get the ball rolling on these robotic discovery techniques much faster than before, and give you a custom-tailored Twitter experience that requires little effort on your part.
Computer algorithm picks the world's most creative art
Who would you trust to determine history's most creative art? A room full of seasoned critics? Rutgers University researchers think a machine can do the job. They've developed a computer vision algorithm that ranks the creativity of art based on how similar it is to earlier works in terms of everything from color and texture to the presence of familiar objects. The code treats art history as a network -- groundbreaking pieces are connected to later derivatives, and seemingly unique content may have a link to something produced in the distant past.
Google hopes to count the calories in your food photos
Be careful about snapping pictures of your obscenely tasty meals -- one day, your phone might judge you for them. Google recently took the wraps off Im2Calories, a research project that uses deep learning algorithms to count the calories in food photos. The software spots the individual items on your plate and creates a final tally based on the calorie info available for those dishes. If it doesn't properly guess what you're eating, you can correct it yourself and improve the system over time. Ideally, Google will also draw from the collective wisdom of foodies to create a truly smart dietary tool -- enough experience and it could give you a solid estimate of how much energy you'll have to burn off at the gym.
Microsoft thinks it can guess your age using facial recognition
Since we're right smack in the middle of Microsoft's BUILD dev conference, the company's showing off one of it's Azure APIs with a site you can put to the test. How-Old.net allows you to upload a picture before the site recognizes faces and analyzes them to determine their age. No, I'm not 41... I'm 31, and that picture is from over two years ago. Other folks here at Engadget received results closer to their real age, but it made us wonder: why not use a web cam to snap a picture under current conditions. You know, after I've had a chance to apply my daily dose of wrinkle remover. Perhaps that option on the way.
IBM's Watson cognitive computer has whipped up a cookbook
IBM's Watson learning computer system isn't just content with making the occasional meal -- it has a whole slew of recipes lined up. The tech company is launching Cognitive Cooking with Chef Watson, a cookbook based on Watson's knack for combining food in a way that produces unique (and typically tasty) flavors. There's only about 65 foodstuffs in the mix, but they're considered "greatest hits" that should work well in real life. Just be prepared to do more grocery shopping than usual when the book arrives on April 14th, since IBM's machine tends to choose ingredients that you probably don't have in the pantry.
Amazon's web services are smart enough to make predictions
You no longer have to run a tech giant (or work in a lab) to take advantage of learning computers. Amazon has launched a machine learning feature for Web Services that lets any developer use this computer intelligence to make predictions. Instead of having to sift through data yourself and spend ages fine-tuning algorithms, you let Amazon's servers comb through the info and create predictions largely on their own. This potentially saves you a ton of time, especially if you're running a small outfit that can't afford a lot of servers -- Amazon claims that it took 20 minutes to solve one problem that previously took 45 days.
This factory robot is small, precise and human-friendly
Picture a typical factory robot in your head and you'll probably see a cold, unsympathetic arm performing relatively simple tasks. You may want to shake that image soon, though. Rethink Robotics has taken the wraps off of Sawyer, a smaller sibling to its earlier Baxter model that's built for handling high-precision tasks that most machines can't tackle, such as testing circuit boards. The one-armed robot is designed to be as people-friendly as its predecessor, with a touchscreen for a face and software that lets you teach it by guiding it with your hands. The big improvements are in the arm itself. Sawyer is using new actuators and joints that make it smaller, faster and more precise, which should help with assembling or verifying lots of tiny parts.
Self-driving vehicles and robotic clerks could take your job in 20 years
It's no secret that computers and robots have been putting people out of work in recent years, but that trend is about to accelerate... at least, if you ask the computers themselves. A machine learning algorithm from Oxford University has sifted through US Bureau of Statistics data and believes that up to 47 percent of American jobs could be replaced by technology within the next 20 years. One of the biggest concerns is in logistics -- self-driving vehicles are advancing quickly enough that they could replace the likes of taxi drivers, truck drivers and forklift operators. Retail is also at risk, since companies can collect enough data about your shopping habits that they might predict what you want more effectively than human clerks.
NASA is using machine learning to predict the characteristics of stars
With so many stars in our galaxy to discover and catalog, NASA is adopting new machine learning techniques to speed up the process. Even now, telescopes around the world are capturing countless images of the night sky, and new projects such as the Large Synoptic Survey Telescope (LSST) will only increase the amount of data available at NASA's fingertips. To give its analysis a helping hand, the agency has been using some of its prior research and recordings to essentially "teach" computers how to spot patterns in new star data.
Google's short film examines the science of voice recognition
People used to think it's harder to make computers play chess (or Jeopardy) and do mathematics than it is to make them understand human language. Turns out the opposite is true -- yes, engineers have made great advancements in voice recognition (Siri and Google voice commands are perfect examples), but they've yet to create a system that can speak with us like another human can. Google's documentary (after the break) talks about the beginnings of voice recognition, the current state of language understanding, as well as the future of artificial neural network technology, which can be used to improve both. The main goal of scientists and engineers is to make computers reach human levels of language understanding, but whether that'll ever happen remains to be seen.
The SAFE project teaches computers to understand your musical vocab
The vocabulary we use to describe music can be tough enough for a human to grok (really, what does it mean when a guitar riff is "crunchy"?) but a team of tinkerers from Birmingham City University aren't interested in helping people understand that language. Nope -- instead, they've cooked up a way to teach your computer what you mean when you throw around words like "bright" or "fuzzy" or, yes, "crunchy" with a program they call the SAFE Project.
