algorithm
Latest
Google to switch on 'semantic search' within months, emphasize things as well as words
A search engine should be about more than just keywords. MC Hammer believes that passionately and Google must do too, because over the next few months and years it'll gradually adjust its own algorithms to put greater emphasis on "semantic search". Under this system, search queries are run through a vast knowledge database that discovers relationships with other words and facts. A Mountain View exec explained it thus: If you search for "Lake Tahoe", you won't just get ranked websites containing those two words but also key attributes about the lake, such as its location, altitude, average temperature and Bigfoot population. If a piece of knowledge isn't the in the ever-expanding database, the search engine will still use semantic search to help it recognize and evaluate information held on websites. In doing this, Big G hopes to compete with social networks that are amassing their own valuable (and sometimes intrusive) databases full of personal information, while also encouraging people to stay longer on its site and see more targeted ads. Google, who is Viviane Reding?
MIT-created algorithm predicts likelihood of running red lights, places blame where appropriate
The bad news: Math will always judge you. The good news: It'll still be there to judge everyone else. In a recent article published in the IEEE Transactions on Intelligent Transportation Systems, MIT's Professor Jonathan How and three colleagues announced they've created an algorithm that can predict whether an oncoming car is about to run a red light one or two seconds before a possible collision. The algorithm can compute the likelihood of a vehicle running a red light based on its rate of deceleration as it is approaching the intersection with a level of precision down to mere milliseconds. The team, which applied the algorithm to more than 15,000 vehicles during the study, used instruments that monitored vehicle speeds and locations as well as when the lights turned red. When the results were tallied, they found that they were able to correctly predict who would run a red light 85 percent of the time. In other news, MIT is working on a much simpler algorithm capable of predicting when your significant other will break up with you, the formula factoring in at least four behavioral elements from the last season of "Jersey Shore."
Magisto edits videos automagically, deluges the interwebs with idiot auteur savancy
Oh, the plague of social media and its irrational empowerment of at-home, amateur media moguls. Well, truth be told, not everyone has the tenacity to sit and slog through hours of footage to create a skillfully made, ready-for-prime time upload. Not to worry you talentless hacks, Magisto's got a web-based tool that'll automate your lack of video editing expertise, and churn out YouTube-worthy, ADD-style clips replete with background music and fancy multi-window effects. The service, which makes use of an algorithm to recognize "people, pets and landscapes and can even...[analyze] sounds and images," had formerly been available in a private beta, but is now open and free to anyone with a camera, a computer and a decent internet connection. We've seen the results of the company's handiwork and it's all pretty much the same thing -- an incoherent, tune-laden mashup. Which is to say, ideal for the Twitter and Facebook IV drips we've come to subsist upon. Go ahead and test the hyper-editing software out at the source below -- it's not like you actually have to do anything, anyway. Just click.
Researchers use wireless network to monitor breathing, could save lives
When Neal Patwari and his team of researchers developed a wireless network capable of seeing through walls, we assumed they were simply looking to cultivate their Alastor Moody-like superpowers. Turns out, they had far more important things on their minds. Patwari and his colleagues at the University of Utah have now penned a new study in which they demonstrate how their motion detecting technology could be used to monitor breathing patterns, as well, potentially enabling doctors to keep closer track of patients with sleep apnea or babies susceptible to sudden infant death syndrome (SIDS). To do this, Patwari reclined on a hospital bed and surrounded himself with 20 wireless transceivers operating at a frequency of 2.4GHz, as pictured above. He then timed his breathing at about 15 breaths per minute (the average rate for a resting adult), which he measured with his array of nodes and a carbon dioxide monitor. The engineer ultimately found that his system's algorithm could accurately measure respiration within 0.4 to 0.2 breaths per minute -- a relatively low error rate, since most monitors round off to the nearest full breath. Patwari says this development could offer a non-invasive and low-cost alternative to the devices used in most hospitals, and hopes to implement his technology into at-home baby monitors, as well. He acknowledges, however, that it will likely take at least five years before any of that happens -- so don't hold your breath. Full PR after the break.
New program makes it easier to turn your computer into a conversational chatterbox
We've already seen how awkward computers can be when they try to speak like humans, but researchers from North Carolina State and Georgia Tech have now developed a program that could make it easier to show them how it's done. Their approach, outlined in a recently published paper, would allow developers to create natural language generation (NLG) systems twice as fast as currently possible. NLG technology is used in a wide array of applications (including video games and customer service centers), but producing these systems has traditionally required developers to enter massive amounts of data, vocabulary and templates -- rules that computers use to develop coherent sentences. Lead author Karthik Narayan and his team, however, have created a program capable of learning how to use these templates on its own, thereby requiring developers to input only basic information about any given topic of conversation. As it learns how to speak, the software can also make automatic suggestions about which information should be added to its database, based on the conversation at hand. Narayan and his colleagues will present their study at this year's Artificial Intelligence and Interactive Digital Entertainment conference in October, but you can dig through it for yourself, at the link below.
Researchers developing software to finger phony reviews
Opinion spam isn't a new version of your favorite meat treat, repackaged for discerning canned ham consumers. According to a team of researchers at Cornell University, it's a growing problem affecting user generated review sites, and the gang is working to stop it dead in its tracks with a new program that's aimed at tracking down fake reviews. That software, which has been tested on reviews of Chicago hotels, uses keyword analysis and word combination patterns to bust opinion spammers -- fakers, for example, use more verbs than their truth-telling counterparts. The as of yet unnamed program apparently has the ability to post deceptive opinions with 90 percent accuracy, but is currently only trained on hospitality in the Windy City. Ultimately, the group sees the software as a filter for sites like Amazon, but, for now, you'll just have to trust the old noggin to do the detecting for you.
Social x-ray glasses can decode emotions, make your blind dates less awkward
You may consider yourself a world-class liar, but a new pair of "social x-ray" glasses could soon expose you for the fraud you really are. Originally designed for people suffering from autism, these specs use a rice grain-sized camera to pick up on a person's 24 "feature points" -- facial expressions that convey feelings of confusion, agreement and concentration, among others. Once recognized, these signals are analyzed by software, compared against a database of known expressions and then relayed to users via an attached headphone. If their date starts to feel uncomfortable, a blinking red light lets them know that it's time to shut up. Rosalina Picard, an electrical engineer who developed the prototype with Rana el Kaliouby, acknowledged that her algorithm still needs some fine tuning, but told New Scientist that the glasses have already proved popular with autistic users, who often have difficulty deciphering others' body language. No word yet on when these social specs could hit the market, but they'll probably make us even more anti-social once they do.
Scientists develop algorithm to solve Rubik's cubes of any size
A computer solving a Rubik's cube? P'shaw. Doing it in 10.69 seconds? Been there, record set. But to crack one of any size? Color us impressed. Erik Demaine of MIT claims to have done just that -- he and his team developed an algorithm that applies to cubes no matter how ambitious their dimensions. Pretty early on, he realized he needed to take a different angle than he would with a standard 3 x 3 x 3 puzzle, which other scientists have tackled by borrowing computers from Google to consider all 43 quintillion possible moves -- a strategy known simply as "brute force." As you can imagine, that's not exactly a viable solution when you're wrestling with an 11 x 11 x 11 cube. So Demaine and his fellow researchers settled on an approach that's actually a riff on one commonly used by Rubik's enthusiasts, who might attempt to move a square into its desired position while leaving the rest of the cube as unchanged as possible. That's a tedious way to go, of course, so instead the team grouped several cubies that all needed to go in the same direction, a tactic that reduced the number of moves by a factor of log n, with n representing the length of any of the cube's sides. Since moving individual cubies into an ideal spot requires a number of moves equal to n², the final algorithm is n²/log n. If we just lost you non-math majors with that formula, rest assured that the scientists expect folks won't be able to apply it directly, per se, though they do say it could help cube-solvers sharpen their strategy. Other that, all you overachievers out there, you're still on your own with that 20 x 20 x 20.
Bronze makes your music collection sound downright lazy
No matter how large your record collection is, there'll always be days when you're sick and tired of everything you own. There are some pretty easy fixes to this, of course -- streaming music services, internet radio, and even the old fashioned terrestrial variety, but none attack the problem quite as directly as Bronze. The app was co-created by Golden Silvers' front man, Gwilym Gold, in part as a clever way to promote his new track, "Flesh Freeze," a hushed, Robert Wyatt-esque number, which just so happens to be the only song that it's compatible with at present. Bronze combats musical stagnation by altering the music itself, assuring that no two listens of a song are ever quite the same, running an algorithm that jumbles the sequence and presence of elements including the instruments and vocals. The changes are generally fairly subtle from play to play, taking a few listens to really pick up on what's unfolding. At present, it's available as a download for the Mac, with Windows and mobile versions to follow. It's free for now, but that, like the song, should change soon enough.
Robot Ruby solves Rubik's Cube in 10.69 seconds, still can't beat humans (video)
There are a few robots smart enough to solve a Rubik's Cube in seconds flat, but a group of students at Swinburne University of Technology think theirs may be the fastest on Earth. Their bot, named Ruby, recently mastered the puzzle in just 10.69 seconds, including the time spent analyzing the cube. To achieve this feat, the device scanned the toy with a webcam before its software processed the images to crank out a solution. According to the university, Ruby's 10-second mark smashes the current world robot record of 18.2 seconds, unofficially making it the fastest cube-solving machine on the planet. But Ruby still has a long way to go before it catches Feliks Zemdegs -- a 16-year-old Australian who solved a Rubik's Cube in 6.24 seconds and continues to carry the torch for all of humanity. Twist and turn your way past the break for the full PR and a video of Ruby in action, as well as a clip of Zemdegs showing us how the pros do it.
Tenacious robot ashamed of creator's performance, shows mankind how it's done (video)
Looks like researchers have made another step towards taking Skynet live: giving robots the groundwork for gloating. A Swiss team of misguided geniuses have developed learning algorithms that allow robot-kind to learn from human mistakes. Earthlings guide the robot through a flawed attempt at completing a task, such as catapulting a ball into a paper basket; the machine then extrapolates its goal, what went wrong in the human-guided example, and how to succeed, via trial and error. Rather than presuming human demonstrations represent a job well done, this new algorithm assumes all human examples are failures, ultimately using their bad examples to help the 'bot one-up its creators. Thankfully, the new algorithm is only being used with a single hyper-learning appendage; heaven forbid it should ever learn how to use the robot-internet.
Algorithm places September 11th victims next to friends at 9/11 Memorial
This year marks the 10th anniversary of the September 11th attacks. The brilliant memorial (partially shown above) to be unveiled that day is radical not only as a result of on-site electronics, but also thanks to a computer algorithm, responsible for resting fallen victims next to each other based on affiliation, not the alphabet. Victims will be grouped by employer, but also by their friendships -- before the towers fell, and now for generations to come.
Google buys Green Parrot Pictures, looking to make YouTube vids easier on the eye
If you can't fix it, buy someone who can. That must be Google's rationale behind this latest acquisition, as the proprietor of YouTube has just bought Green Parrot Pictures, a company concerned solely with enhancing and improving the quality of video content. Through the use of some fancy motion prediction algorithms, the Irish startup has been able to build a name for itself over the past few years, and now it's been snapped up by the biggest fish in the online video ocean. The removal of flicker, noise and blotches from poorly executed recordings sounds nice, but we're most excited by Green Parrot's video stabilization feature. With all the cameraphone video being uploaded nowadays, there's plenty of camera shake populating YouTube's archives, and the addition of such a potent post-production technique seems like a veritable boon to us. Check out video demos of the stabilization algorithm and Green Parrot's other technologies below.
Microsoft's OneVision Video Recognizer can detect, identify, and track your face on video... so smile!
Here's your classic case of "just because you can, doesn't mean you should." Microsoft's Innovation Labs have just demonstrated a OneVision Video Recognizer algorithm that's powerful enough to perform face detection duties on a running video feed. It can recognize and track humanoid visages even while they're moving, accept tags that allow auto-identification of people as they enter the frame, and can ultimately lead to some highly sophisticated video editing and indexing via its automated information gathering. Of course, it's that very ease with which it can keep a watchful eye on everyone that has us feeling uneasy right now, but what are you gonna do? Watch the video after the break, that's what.
MIT Media Lab gets a multiplicitous new logo (video)
Logos can be surprisingly divisive things, so the MIT Media Lab has decided to cheat a little bit with its new identity: it won't have just one logo, it'll have 40,000. You heard / read / imagined that right, the new Media Lab logo will simply be the concept of three intersecting "spotlights," composed of three colors, straight lines, three black squares, and a few blending gradients. There's an algorithm behind it all, which is used to generate a unique logo for every new member of staff, meaning that although trademark claims may be a headache to enforce, originality will continue thriving in the Lab for a long time to come. Hit the source link to learn more or leap past the break for a nice video rundown.
Google makes rich richer, poor poorer in search results
If you've been paying attention to the state of search as of late, you'll know that Google's between a proverbial rock and hard place right now. Some individuals and companies claim Mountain View's beloved search engine is losing to the spammers, squatters, scrapers and content farms by failing to weed them from the system -- though you can now do that on your own -- while others say it's squashing the little guy by unfairly downranking competitors in search results. We're not sure if either is truly the case, the company's made a mildly controversial move this week: it's tweaked the search algorithms to "reduce rankings for low-quality sites," and "provide better rankings for high-quality" ones. As ever, Mountain View's not talking about what that change mathematically entails, though it says about 11.8 percent of queries will be affected as a result. In short: some will be happy, some sad, some angry, and many won't notice at all.
Gmail Priority Inbox now available on iPhone
The official Gmail blog has announced that the Priority Inbox feature, which uses an algorithm to mark what it thinks are important emails in your inbox, is now available on the web version of Gmail for the iPhone. If you've set up the service to work on your usual browser, it'll now also work on Gmail for Mobile, which is up and running in HTML5 on iPhones everywhere. I use Gmail almost exclusively for all of my email, yet aside from a short trial when it was first released, I've never really had much interest in Priority Inbox. I do a much better job of sorting my email than Google's algorithm, and I'm too worried that I'll miss something important that should have been prioritized if I depend on the formula. But a feature is a feature, and since Gmail is free anyway, those of you who use Priority Inbox will probably be happy to see it the next time you log into Gmail from a mobile device. [via MacStories]
DARPA's MSEE to develop new mathematical language, race of sentient machines
The hyper-ambitious folks at DARPA are totally over the current state of military data collection, and they're pretty sure they know how to fix it: teach sensors how to think. Well, they've got an idea how to fix it, but they've put out a call for mathematician to do the dirty work. The Mathematics of Sensing, Exploitation, and Execution (MSEE) program seeks a unified mathematical language that cane teach sensors not only to collect data, but to interpret, and act on it too. The aim is to eliminate the "data deluge" that comes from ever-increasing streams of information, like cellphone intercepts and video drone feeds, allowing analysts to focus on the important stuff. Currently the onus falls on humans to interpret the overwhelming amount of information collected by military sensors, but DARPA is confident that the right algorithm could have machines interpreting the world as early as 2014. Which gives you right around three years to fulfill every fantasy you've ever had.
Google under fire for promoting own content ahead of competing websites
This isn't exactly a new allegation, but the idea's spreading fast: Google is tuning search results to favor itself, and perhaps that's not entirely fair. Though the European Union is already investigating Google for potential antitrust violations, a recent article in the Wall Street Journal cites several US businesses that aren't too happy themselves, claiming that Google Places, Product Search and the like took a big chomp out of their traffic. Google's defense, as usual, rests on its secret algorithms, which it claims aren't rigged in any way, adding that the prominent placement of location- and product-based search results are just a way to get users quicker answers to their queries. If you type in "day spa nyc," you're looking for some catered suggestions, right? And what of those who argue differently? Well, obviously they're in league with Microsoft.
App Store algorithms down and back up, probably tweaked
Regular users may not have noticed this, but I'm sure developers have -- the App Store's charts have been in a freeze for the past few days, leaving apps in exactly the same positions as they were last Thursday. That might not seem like too big a deal, but for many developers, the App Store charts are where they make or break a profit -- not only do the charts provide a good inkling of how popular an app is compared to others, but the top 10 on the charts provides a huge amount of exposure -- just getting in the top 10 often boosts app sales just by virtue of the extra eyeballs. For the past few days, though, the App Store charts haven't changed, and that usually means that Apple is messing with something, either tweaking the way the charts are calculated, or just plain changing the way that they work. Last year the company fixed an issue where developers could constantly update their apps to keep them on "Recently Released" charts, but this year we haven't quite figured out what changes have been made. At any rate, it seems likely Apple was simply getting things ready for what's sure to be a busy holiday season for the App Store. Hopefully developers who released their apps on Halloween didn't fall too far through the cracks while things were down.
