artificialintelligence
Latest
iPhone Assistant called a "world-changing event" by Siri co-founder
One of the expected announcements in tomorrow's "Let's Talk iPhone" event revolves around the iPhone Assistant, a powerful voice recognition and artificial intelligence tool that will enable real-time searching and control of iPhones. The tool, based on technology acquired in Apple's 2010 purchase of Siri, is so powerful that Siri co-founder Norman Winarsky is referring to its release as a "world-changing event." Winarsky is no longer with Siri, so he's offering pure speculation based on his prior knowledge of the company's technology. He believes that inclusion of the Siri artificial intelligence capability might explain why an iPhone announcement wasn't made until October (rather than the usual summer timeframe), noting that "AI takes a lot of computing power." Rumors floating around the web note that Assistant may only run on a next-generation iPhone due to the need for more RAM and a faster processor. In an interview with iPodNN News, Winarsky waxed enthusiastic about the capabilities of Assistant, noting that "If the rumors are true, Apple will enable millions upon millions of people to interact with machines with natural language ... We're talking another technology revolution. A new computing paradigm shift."
IBM's Watson set to tackle health insurance, takes 'Diagnosis for $1,000'
After tackling your tech support woes, the famed Watson is moving on to mop up the health insurance industry. That's right, the IBM showstopper we all know and love for trouncing trivia kings on Jeopardy has been hired by one of the largest health insurance company's in the US. WellPoint Inc. will make use of the system's breakneck speed and healthcare database alongside patient records -- allowing the supercomputer to guide treatment options and prescribe medicines. Once implemented, data will be combined from three sources in a matter of seconds: a patient's chart / records from a doctor, the insurance company's patient history and the medical knowledge that Watson already possesses. A pilot program will roll out next year to a number of cancer facilities, academic medical centers and oncology practices. No word yet on when The Watson School of Medicine will start accepting applications.
When two chatbots have a conversation, everyone wins (video)
What did one chatbot say to the other chatbot? Quite a lot, actually -- but good luck making any sense out of it. That's what researchers from Cornell's Creative Machines Lab recently discovered, after pitting two bots against one another for a good ol' fashioned talk-off. It's all part of the lab's submission to this year's Loebner Prize Competition in Artificial Intelligence -- an event that awards $100,000 to the team whose computer programs can conduct the most human-like conversations. Unfortunately for Cornell's squad, their chatbots still have a long way to go before achieving conversational coherence, though they could easily get hired as anchors on most cable news networks. Throughout the course of their frenetic (and often snippy) discussion, one bot raised heady questions about God and existence, while the other boldly claimed to be a unicorn. Basically, they had the exact same conversation we used to have in our dorm rooms every night, at around 4 am. Watch it for yourself after the break. It's nothing short of sublime.
Robot skin captures super detailed 3D surface images
Remember those awesome pin art toys where you could press your hand (or face) into the pins to leaving a lasting impression? Researchers at MIT have taken the idea one (or two) steps further with "GelSight," a hunk of synthetic rubber that creates a detailed computer visualized image of whatever surface you press it against. It works as such: push the reflective side of the gummy against an object (they chose a chicken feather and a $20 bill) and the camera on the other end will capture a 3-D image of the microscopic surface structure. Originally designed as robot "skin," researchers realized the tool could be used in applications from criminal forensics (think bullets and fingerprints) to dermatology. The Coke can-sized machine is so sensitive, it can capture surface subtleties as small as one by two micrometer in surface -- finally solving the mystery of who stole the cookies from the cookie jar. (Hint: we know it was you Velvet Sledgehammer).
Stanford schooling unwashed masses with free online Intro to Artificial Intelligence (video)
If you fancy yourself a Stanford (wo)man, but lack the requisite dollars to actually attend, now's your chance to collect those collegiate bragging rights. Starting October 10th, you can join Professor Sebastian Thrun and Google's Director of Research, Peter Norvig, in a free, online version of the school's Introduction to Artificial Intelligence course. The class covers, "knowledge representation, inference, machine learning, planning and game playing, information retrieval, and computer vision and robotics," and ambitiously aims to be the largest online AI course ever taught. If you're feeling the ole red and white, you can register at the source link below, but if you're looking for the official Stanford stamp of approval, we're afraid you're barking up the wrong tree -- non-students will receive a certificate of completion from the instructors only. Still interested? Check out the video introduction after the break and hit the source for more details.
Tokyo Institute of Technology's SOINN robot teaches itself to serve humans (video)
Robots have been replacing more and more human workers for quite some time now, but in most instances they're still just being programmed to perform specific tasks. As evidenced by this bot developed by the Hasegawa Group at the Tokyo Institute of Technology, however, there's also a growing number capable of teaching themselves some new tricks, and they're getting smarter every day. This particular one employs what's called a self-organizing incremental neural network (or SOINN), which lets it build up a base of knowledge that it can apply to new tasks and make educated guesses about how to proceed with them -- in this case, pouring a glass of water and then dropping an ice cube in it (or what's supposed to be water and an ice cube, at least). Head on past the break for a video.
Khepera swarm robots learn to build a mobile quadrocopter landing platform (video)
Last we saw the Khepera swarm robots from the Georgia Robotics and Intelligent Systems lab, they were just beginning to learn how move in formation and spell out "GRITS" on the floor. Well, these bots are growing up fast. The lab's latest video shows a group of four of the robots following a leader, and arranging themselves to form a mobile landing platform for a hovering quadrocopter -- a feat they manage to make seem easy. What will they think of next? We're a little scared to ask.
Calling for tech support? IBM's Watson might be on the other end
Watson may have Jeopardy! and the medical realm under lock, but retail / service industries? Not yet, but soon. Very soon. According to a new piece in Hemispheres Magazine, IBM's now looking to shop the supercomputer's world-class vocal recognition technologies to outfits in retail and customer service, with those enterprises in particular drooling at the thought of having a sophisticated machine recognizing human speech. In theory, at least, basic questions could potentially be answered entirely by Watson, but that's honestly not a future we're too fond of. There's also the possibility of using analytical data that Watson collects in order to better position deals, service and other tech support centers based on what kinds of requests come in the most. So, eager to speak with a kindhearted, potentially confused robot? Or will that flustered, potentially sympathetic Earthling still suffice?
AI competition pits Ms. Pac-Man against ghosts in the Manichean struggle of our time (video)
While the world breathlessly awaits the Pac-Man reality TV show, the University of Essex held a programming competition starring that other yellow chomper. The Ms. Pac-Man vs. Ghost Team contest pitted 13 competitors from nine different countries against one another, to see who could create the most elusive Ms. Pac-Man or the wiliest ghost gang. The participants coded routines for the titular hero or Blinky, Pinky, Inky, and Sue, with organizers then running the programs against one another on the Java-based playing field. The highest single-game score went to Atif, who racked up 69240 points versus DarkRodry's ghosts, while ghost team Legacy2TheReckoning held RandomMsPacMan to a mere 410 points. Another competition will take place in August, so limber up your coding fingers, Pac-Maniacs. In the meantime, strap on your headgear and cheer on your round yellow hero in the video below.
Steve Wozniak calls us all dogs, in a nice way
You can stop worrying about the robot apocalypse now. Steve Wozniak has weighed in on the matter, and it turns out we've pretty much lost. The Apple co-founder / dancing star discussed the subject with an Australian business crowd, mapping out a future in which artificial intelligence equals our own, and mankind's own input is meaningless. In other words, "We're going to become the pets, the dogs of the house." Woz added that his take on the whole war thing was, in part, a joke -- it's the part that wasn't that we're worried about. Though if our own dogs' existences are any indication, things could be a lot worse. [Thanks, Shaun]
Schizophrenic computer may help us understand similarly afflicted humans
Although we usually prefer our computers to be perfect, logical, and psychologically fit, sometimes there's more to be learned from a schizophrenic one. A University of Texas experiment has doomed a computer with dementia praecox, saddling the silicon soul with symptoms that normally only afflict humans. By telling the machine's neural network to treat everything it learned as extremely important, the team hopes to aid clinical research in understanding the schizophrenic brain -- following a popular theory that suggests afflicted patients lose the ability to forget or ignore frivolous information, causing them to make illogical connections and paranoid jumps in reason. Sure enough, the machine lost it, and started spinning wild, delusional stories, eventually claiming responsibility for a terrorist attack. Yikes. We aren't hastening the robot apocalypse if we're programming machines to go mad intentionally, right?
Robots learn to share, try to repair bad rep (video)
We've been told time and time again to fear our mechanical friends, so imagine our relief when we heard that some Swiss scientists had a batch of bots that displayed altruism. What's more, these little two-wheeled foragers weren't programmed to share, they evolved the trait. Researchers at EPFL infused Alice microbots with digital "genes" that mutated over time as well as color sensors that allow them to navigate their environment. The robots were tasked with collecting "food" and given the option to keep it for themselves or split it amongst their silicon-brained relatives. The more they decided to give to others with similar genetic makeup the more those virtual genes were passed on to future generations -- including the one for altruism. The experiment is an example of Hamilton's Rule, an evolutionary model for how the seemingly counter-intuitive trait of selflessness could arise through natural selection. Don't let your guard down just yet, though -- the robots are only sharing with each other for now.
Researchers build synthetic synapse circuit, prosthetic brains still decades away
Building a franken-brain has long been a holy grail of sorts for scientists, but now a team of engineering researchers have made what they claim to be a significant breakthrough towards that goal. Alice Parker and Chongwu Zhou of USC used carbon nanotubes to create synthetic synapse circuits that mimic neurons, the basic building blocks of the brain. This could be invaluable to AI research, though the team still hasn't tackled the problem of scope -- our brains are home to 100 billion neurons, each of which has 10,000 synapses. Moreover, these nanotubes are critically lacking in plasticity -- they can't form new connections, produce new neurons, or adapt with age. All told, the scientists say, we're decades away from having fake brains -- or even sections of it -- but if the technology advances as they hope it will, people might one day be able to recover from devastating brain injuries and drive cars smart enough to avert deadly accidents.
Quadrocopters juggle balls cooperatively, mesmerize with their lethal accuracy (video)
You've seen one quadrocopter juggle a ball autonomously while gliding through the air, but how's about a pair of them working cooperatively? Yeah, we've got your attention now. The Zurich-based lab that brought us the piano-playing and ball-bouncing quadrocopter is back with a simply breathtaking display of robotic dexterity and teamwork. Like all mad scientists, they call their Flying Machine Arena research "an experiment," though we see it a lot more as a Pong-inspired dance of our future overlords. We all know how far video games have come since two paddles batted a ball between one another, right?
Japan's space agency considers using rockets with artificial intelligence
The keyword here is obviously "considers," but it looks like Japan's space agency, JAXA, is indeed seriously thinking about using artificial intelligence to improve their rocket launches. As JAXA scientist Yasuhiro Morita explains, as opposed to simply being "automatic" as rockets are today, an "artificially intelligent" rocket would be able to keep watch on its condition, determine the cause of any malfunction, and potentially even fix it itself. According to JAXA, that would not only make rocket launches more efficient, but more cost-effective as well given the reduced manpower needs. That's not the only new measure being explored to cut costs, though -- as Space.com reports, JAXA's new Epsilon launch vehicle is also being built using fewer, but more advanced components, which promises to let it be moved to the launch pad nearly fully assembled. It's currently set to launch sometime in 2013, although it's not yet clear how much it will actually be relying on AI if such a system is put in place.
NASA's Global Hawk completes unmanned airborne refueling simulation, will do it for real next year (video)
While some bot makers are busying themselves designing AI to simulate humans' natural and distinct lack of intelligence, it's nice to see there are still old-fashioned researchers out there keeping the Skynet dream alive. Northrop Grumman's aeronautics gurus have paired together a Global Hawk unmanned aircraft with a manned Proteus ship way up in the skies -- 45,000 feet, to be precise -- with the vessels of ingenuity managing to fly in tandem at a distance as short as 40 feet. Unsurprisingly, this is the first time such intimacy has been reached between UAVs (the Proteus had a monitoring crew on board to ensure the insurance bill wasn't through the roof) in high altitude, and the ultimate goal of having two Global Hawks doing the deed without any human intervention is said to be within reach by next year. That's when these light and agile air drones will be able to refuel themselves and go on for a mighty 120 hours in the air... plenty of time to complete a well planned extermination down below, if one were so inclined.
Meka's M-1 Mobile Manipulator, a cuter Cody the spit bath robot (video)
Remember Cody? The robot from Georgia Tech designed to give spit baths to the elderly and crippled? Well, Cody's got an attractive younger cousin named M-1, and for $340,000 this fine piece of machinery could be all yours. Built by San Francisco-based Meka Robotics, the M-1 Mobile Manipulator (based on Cody) runs on a combination of ROS and proprietary software and sports a Kinect-compatible head with a five megapixel Ethernet camera, arms with six-axis force-torque sensors at the wrist, force controlled grippers, and an omnidirectional mobile base. If the standard features don't fit your needs, Meka offers various upgrades, including four-fingered hands and humanoid heads, complete with expressive eyelids (à la Meka's Dreamer), ears, and additional sensor compatibility. These add-ons will of course cost you, but we think its worth it to have those big translucent eyes staring back at you. A rather touching demo after the jump.
IBM demonstrates Watson supercomputer in Jeopardy practice match
We're at IBM's HQ in upstate NY, where IBM will pit its monstrous Watson project (in the middle buzzer spot) against two Jeopardy greats, Ken Jennings and Brad Rutter. Watson has been in development for four years, and this is its first big public practice match before it goes on national TV in February for three matches against these giants of trivia. Unlike IBM's Deep Blue chess project in the 90s, which was pretty much pure math, Watson has to deal with the natural language and punny nature of real Jeopardy questions. IBM, ever the salesman, has thrown gobs of its fancy server hardware at the project, with 10 racks full of IBM Power 750 servers, stuffed with 15 terabytes of RAM and 2,880 processor operating at a collective 80 teraflops. IBM says it would take one CPU over two hours to answer a typical question, so this massive parallel processing is naturally key -- hopefully fast enough to buzz in before Ken and Brad catch on to the human-oriented questioning. We'll update this post as the match begins, and we'll have some video for you later in the day.
USC Institute for Creative Technologies gets new building to amp up its military VR work
University of Southern California's Institute for Creative Technologies has just moved into 72,000-square feet of shiny new facilities in Playa Vista, California -- on the same grounds which once served as the headquarters for Howard Hughes' aircraft company. Funded by the US Army to develop virtual reality technology, the ICT's work is now found on 65 military sites across the country. Before your brain starts wandering towards thoughts of Call of Duty on military-grade steroids though, keep in mind that much of the institute's innovations revolve around simulating surrogate interactions with so-called "virtual humans". For example, thanks to advanced AI language programming, soldier patients projected on life size semi-transparent screens help teach doctors about treating combat trauma, while virtual Army personnel characters such as Sergeant Star can interact naturally with soldiers in leadership training exercises. Still, that shouldn't imply ITC doesn't dabble in good ole' fashion combat simulation work. In fact, it's currently running a training exercise on three military bases designed to prepare soldiers for an insurgent ambush within a highly-realistic virtual town reconstructed from satellite imagery. No, it may not sound as wild as robotic exoskeletons, flying Humvees or ultrasound-based mind control, but it does make your life on The Sims seem totally fake. To judge for yourself, check out the video overview on the next page.
Robo-nurse gives gentle bed baths, keeps its laser eye on you (video)
When they're not too busy building creepy little humanoids or lizard-like sand swimmers, researchers at the Georgia Institute of Technology like to concern themselves with helping make healthcare easier. To that end, they've constructed the Cody robot you see above, which has recently been demonstrated successfully wiping away "debris" from a human subject. The goal is simple enough to understand -- aiding the elderly and infirm in keeping up their personal hygiene -- but we'd still struggle to hand over responsibility for granny's care to an autonomous machine equipped with a camera and laser in the place where a head might, or ought to, be. See Cody cleaning up its designer's extremities after the break.
