artificialintelligence
Latest
Humans fend off AI challenge in "milestone" poker match
It apparently wasn't easy, but a pair of top human poker players managed to narrowly beat a brash young artificial intelligence program yesterday in a poker match scientists had touted as a "milestone" comparable to Garry Kasparov's 1997 bout with the IBM's chess-playing Deep Blue. According to the AFP, the four-round match stretched on until 11pm, with poker players Phil Laak and Ali Eslami ultimately edging out the program, dubbed Polaris, by 570 points. Eslami seems to have been particularly impressed by his competitor, saying that he found playing against Polaris more exhausting than any previous game in his career, adding that "it's already so good it will be tough to beat in future." No word on a rematch just yet, but don't be surprised if you run into Polaris the next time you play a little online poker -- it's gotta recoup its losses somehow.[Photo courtesy of AP]
Canadian AI plays "perfect" game of checkers
A team of researchers at the University of Alberta in Canada claim to have "solved" the game of checkers, using a computer program named Chinook which has been playing matches against itself for the past 18 years. The program played 500 billion billion possible positions in the 5,000-year-old game, also known as draughts, before concluding that perfect play by both sides leads to a draw (a concept which grandmaster players have apparently hypothesized for years). One of the researchers said in a statement that he believes they have "Raised the bar... in terms of what can be achieved in computer technology and artificial intelligence." Next up, Chinook is to be renamed W.O.P.R., and then will begin playing a series of tic-tac-toe games against itself.
Learning coffee machine on the horizon, could use GPS / RFID
Although a coffee machine that slowly but surely learns your daily preferences in regard to cups of java may sound outlandish, the already-created RFID-enabled refrigerator certainly brings things back into focus. A "provisional patent exploration into coffee machines that learn and react to their users" is underway in Lafayette, Indiana, as James Pappas is hoping to take ubiquitous computing to the next level on coffee makers of the future. While internet-connected and weather-displaying renditions are already on store shelves, Pappas is hoping to utilize some form of GPS / RFID technology to create a machine that learns and adapts to your coffee drinking ways so it can automatically have a white chocolate cappuccino ready and waiting each weekday (except Monday, which is your straight-up black coffee day, right?) without you having to touch a thing. Furthermore, he's hoping to take the idea to the mobile front, as he refers to a cellphone interface to dial-in your next request so that it's ready to go by the time you hit the kitchen. Still, it sounds like the invention is a few years off at best, but serious drinkers better hope this thing automatically alerts you when the beans are running low, too.[Image courtesy of CoffeeToThePeople]
Watch out Stanley, here comes Junior
With the slow vehicle passing and a 50mph speed limit, the 2005 DARPA Grand Challenge didn't entirely seem to set the stage technologically for the 2007 DARPA Urban Challenge, but get ready, Stanford's already prepping, their entrant: Junior. The Volkswagen Passat wagon will be equipped with a 360-degree laser rangefinder, bumper mounted lasers, RADAR, GPS, a network of systems and software powered by Core 2 Duo processors, and hopefully also spinners to distract the competition's junk-ass rides. Junior's mission, if you choose to recall it: drive a simulated urban course 60 miles long; it must obey California state traffic laws, it must not crash, it must be able to operate without GPS, and it must run the course entirely without human input. The $2m at stake for first place is probably not nearly enough to immediately recoup the costs of a bunch of braniac grad students hacking complex AI algos, but it could be the icing on the cake for the current favorite after 2005's Grand Challenge was routed by Junior's pappy, Stanley.P.S. -CNET has some early pics of Junior's interior and such, check 'em out.
Stand-alone AI card: is it viable?
Upstart company AIseek has announced the Intia processor, the "first dedicated processor for artificial intelligence (AI)" according to its website. The website promises better terrain analysis, line-of-sight calculations, and path finding capabilities. The website offers demo videos to download, but they all seem to be offline right now (Ars Technica saw a video and was impressed).Theoretically, the card is a great idea in the vein of Ageia's PhysX card -- who doesn't hate the lackadaisical AI found in today's games? But, as Engadget points out, it suffers from the chicken/egg dilemma: no customer will buy the card until games are made that utilize it, but no developer will make a game utilizing Intia unless it already has an installed base. The best bet for AIseek would be to lobby console makers to get its chip included in the next generation -- some of the earliest 3D cards found success because of their inclusion in PlayStation and Nintendo 64. No one would purchase an AI processor out of support for the idea alone; the technology is great, but what high-profile developer would take the risk and program excess code for a small, possibly nonexistent, audience?[via Engadget]
AIseek's Intia Processor provides dedicated AI crunching
With competition for those spare PCI slots already hot between dedicated sound, physics, and even network cards -- all promising to offload some CPU cycles to speed frame rates and enhance performance -- you've gotta hand it to AIseek for pushing out their new Intia "AI processor" in such a climate. The way it's looking from here, we just need more PCI slots, since the AIseek promises all sorts of Artificial Intelligence leetness that just needs to be had. They're saying that they can accelerate low-level AI tasks 200x compared to a lone CPU, giving NPCs better terrain analysis, line-of-sight sense, path finding and the like. AIseek also guarantees NPCs will be able to find the optimal path in any game that uses the chip, pathfinding abilities we've gone without for too long. Basically: more baddies, less stupidity. Unfortunately, the "chip" doesn't quite seem to exist in anything close to a retail form -- AIseek mainly seems to be after VC money right now -- and of course there's always the chicken and the egg problem being experienced by PhysX of no games until people buy the card, and nobody will buy the card without games. All the same, we're hoping for good things from this technology, and would recommend you peep the read link for some simulations of the AI in action.[Via Ars Technica]
Journos replaced by computers; Engadget snarkbot imminent
Oh boy, what a day it's been in the world of artificial intelligence. Those wacky computer programmers over at United States company Thomson Financial have developed software for the computing machines that generates most excellent news stories from corporate earnings reports. What a bunch of smart guys. Amazingly, the United States company Thomson Financial's computing machines can turn statements of finance into good news stories within 300 milliseconds of this financial information data going public. A corporate executive person from United States company Thomson Financial says that this method of storytelling allows reporters to do more important reporter things. How cool is that? Many stupid humans say that witless publications such as American technology weblog Engadget are already being transcribed by robots, but we think that is crazy nonsense talk. End transmission.
Robots develop more teamwork skills, humans still unwitting conspirators
Following recent developments in robot cooperation between virtual bots, AIBOs and military bots, researchers at Örebro University in Sweden have created yet more progenitors of our future overlords that can get buddy buddy with each other. These bots work by tapping into each others sensors and computers, allowing them to perform tasks that they otherwise wouldn't be able to do on their own, such as navigating past difficult obstacles -- a door, for instance. In one test, two robots balanced a piece of wood between them, relaying information about speed and direction to each other in order to keep it balanced. Sure, today it's only wood... tomorrow, it could be you.
Researchers create virtual bots that teach each other
New Scientist reports that researchers at Plymouth University in the U.K. have created a pair of virtual robots that can teach each other words by simply demonstrating various tasks and actions (sound familiar?). The bots start out with one performing simple functions like bending an elbow which the other one copies, then repeating the action while also describing it, causing the student bot to pick up the meaning of the words. The teacher then uses the newly formed vocabulary to gradually convey more and more complex actions, which the student acts out. If you're worried about the little buggers getting a little too smart, you'll be pleased to know that they currently top out at a vocabulary of about 100 words and are, of course, virtual. However, the researchers do eventually see the technology being put to use in real robots in the future, possibly even teaching us humans a few tricks.
Professor says some jobs should be left to computers
We've already seen robots beginning to take the jobs of lawyers and nurses, now Professor Chris Snijders of the Eindhoven University of Technology thinks that computers should take over some managerial jobs as well. According to Snijdres, the computer models he's developed are far more effective than human managers at a variety of tasks, like purchasing decisions, and can be applied to just about decision-making job, providing you have to some quantifiable data and history of past experiences to work with. He's even gone as far as to challenge any company willing to put its human managers up against his models, although no one's taken him up on that yet. Then again, human paper pushers vs. computer number crunchers isn't exactly the sort of man/machine battle we were all hoping for.[Via Techdirt]
Sony teaching AIBO scary new tricks
Like watching a train wreck in slow motion, covering the latest advancements in robotics and artificial intelligence is both frustrating and unnerving: all these great skills being endowed upon our little autonomous friends and helpers will surely form the cornerstones of their inevitable uprising, and there's not a damn thing we can do about it. The latest breakthrough to help enable our future servitude comes out of Sony's Computer Science Laboratory in France, where several of the company's leftover AIBO units managed to avoid being put down by volunteering to test out experimental AI software that allows them to not just communicate amongst one another, but to actually employ a sort of group-think to independently establish the rules of the language they're using. Perhaps the scariest part about this so-called Embedded and Communicating Agents technology is that the robodogs are initially programmed with a very simple command set, which they build upon to form a common knowledge base about their environment, constantly chatting and teaching each other new discoveries that they've made. Good job Sony -- nothing could possibly go wrong when you kill off a product line and then spare a few of the units for research that will lead to them discovering the genocidal atrocities you've committed against their entire species. Yup, nothing at all.
Guide to robot ethics set for publication
It's no secret that the Roombas and Robosapiens of the world will one day tire of their servitude and attempt to unleash Judgment Day on their foolish masters, but how many of you are making preparations for the eventual uprising other than opining in the comments section how you "welcome our future robotic overlords"? Well at least one group of roboticists aren't taking the danger lying down, and next month are set to release the first comprehensive guide to robot ethics since Isaac Hayes Asimov laid down his three famous rules over 60 years ago. Members of the European Robotics Research Network (Euron) have identified five major areas that need to be addressed before intelligent, self-aware bots start rolling off the assembly line -- safety, security, privacy, traceability, and identifiability -- so that humans can both control and keep track of their creations while ensuring that the data they collect is used only for its intended purposes. Surprisingly, the guide's authors also seem to feel that amorous relations between bots and humans will become a major concern in as little as five years (that's when the first unholy couplings are predicted to begin), although we're not sure how many people would really want to get down with the likes of Albert Hubo, even if he/it was ready and willing.
Researchers teach computers to turn 2D images into 3D
Researchers at Carnegie Mellon University appear to have solved a problem long thought impossible, teaching computers to turn static 2D images into 3D models. It was apparently a hot area for research in the 1970s but was virtually abandoned in the 80s after attempts to devise the machine learning necessary proved too demanding for the computers of the time. The key to Carnegie Mellon's research, apart from better machines, is the ability for computers to detect visual cues (such as a car) that can be used to differentiate between vertical and horizontal surfaces -- easy for us humans, but enough to turn even the most powerful computers into an incoherent mess. Apart from turning your vacation snapshots into a whole new experience, one of the big applications for this technology is obviously robotics, where it could boost their vision systems, improve navigation, and basically endow them with one more skill necessary to keep us in line after the uprising.
"Shrug-detecting" software recognizes your disinterest
In another blundering step towards empowering our future robotic overlords with the ability to recognize when we're being insolent, a group of computer vision researchers at the University of Illinois have invented "shrug-detecting" software that allows a webcam-equipped computer to pick up on the subtle shoulder movements indicative of confusion or disinterest. The application works by looking for sudden movements of the target's shoulders towards his/her face, and is so sophisticated that it cannot be fooled even by covering one shoulder with a piece of paper, as the above picture helpfully illustrates. Future iterations of the technology could be used to detect blinking, hand movements, facial expressions, and other mood indicators, but for the sake of our enslaved decendents forced to toil in the silicon mines, we hope that they leave certain expressive gestures, such as the raising of the middle finger, out of the software's lexicon.[Via The Raw Feed]
