While a great many scientists are attempting to create autonomous bots for uses in surgery, a team of Stanford whiz-kids are having a bit more fun with it all. The crew in question has concocted an artificial intelligence system that "enables robotic helicopters to teach themselves to fly difficult stunts by watching other helicopters perform the same maneuvers." Dubbed a demonstration in "apprenticeship learning," the robots can actually learn by observing rather than having to be programmed, meaning that entire airshows could be reeled off by planes that simply keep an open mind when warm-ups are underway. Of course, they could also be used for more serious applications -- mapping out hot spots of California wildfires, finding land mines in war zones, etc. -- but even if none of that pans out, we're cool with inventions being used purely for entertainment.

With the megapixel race already past the point of noticeable benefit to consumers, it looks like the next camera arms race will be the number of lenses your rig sports -- a team at Stanford is working on a 3D camera that uses 12,616 micro-lenses to generate high quality 3 megapixel images with self-contained "depth maps" that measure the distance to every object in the frame. The system works by focusing each lens above four different overlapping sensor arrays, which work in concert to determine depth -- just like your eyes. Unlike similar systems, the Stanford rig is able to use that data to create a depth map without lasers, prisms, or even complex calibration, which will allow the team to shrink the tech down to compact and cellphone camera size. Once it's ubiquitous, the teams says depth map information can be used to do anything from enhancing facial recognition systems to improving robot vision, but there's still a long way to go -- the team has just started trying to work out how to manufacture the system.

Carbon nanotubes have a ton of promise, and we've seen a lot of prospective applications for the tech, but researchers at Stanford, working with Toshiba, have managed to demonstrate the first use of nanotubes in chips that run at commercially-viable speeds. The chip features 256 ring oscillators and packs over 11,000 transistors in just one hundredth of a square inch. When wired with the nanotubes and powered up, the chip ran at speeds between 800MHz and 1.06GHz -- not desktop speeds, to be sure, but still promising. The team says that while the experiment bodes well for the future, we shouldn't expect any direct applications yet -- but you know we're dreaming of tiny implantable supercomputers anyway.

Stanford claims its latest advances in silicon nanowires have it building batteries with 10 times the capacity of existing Li-ion cells. Apparently people have been trying to stuff silicon -- which has a much higher capacity than existing materials -- into a battery for decades, but since it swells when charged with positively charged lithium and shrinks during use, the silicon has a tendency to "pulverize." Who knew batteries could be so dramatic? Oh, right. The advancement at Stanford, led by Yi Cui, builds the battery in the form of silicon nanowires, giving the silicon room to grow and shrink without damage. A patent is being filed, and Yi Cui is already considering forming a company or licensing the tech to a manufacturer.

Oh man, can you taste it? That oily titanium tang is the smack of robotic cars set for release Saturday in the DARPA Urban Challenge. The final event pits 11 vehicles in competition for a $2 million first, $1M second, and $500k third place prizes. The 60-mile course at the former George Air Force Base in Victorville CA must be successfully navigated in less than 6-hours while avoiding about 50 carbon-based drivers. Can Stanford repeat their 2005 Grand Challenge victory with "Junior" or will Carnegie Mellon's "Boss" (pictured) bust out a BFG-9000 in swift retribution? Tune in Saturday to find out.

Generally speaking, we tend to give autonomous vehicles the benefit of the doubt, but we should really come to grips with the fact that not all robotic flying machines have positive intentions. Apparently, one such ill-willed creation managed to lose control of itself and plow right into Robert Scoble's leg while he was out photowalking at Stanford University. Thankfully, no humans were maimed and the chopper did manage to recover from the crash and proceed on its normal, non-threatening flight path, but we're sure Mr. Scoble will be watching any unmanned crafts with extra caution from here on out.

In an attempt to better understand the way in which human babies learn to speak, researchers at Stanford University say they have created a computer program which can learn baby talk. The largely accepted theory about human language is that all of the sounds we make are hard-wired into our brains, but now that James McClelland -- a professor at the Palo Alto college -- has tested his theory, it would appear that those notions have been debunked. During "training sessions" in both English and Japanese, a computer followed along to recordings of mothers speaking to their children, and was able to pick up the basic vowel sounds as the baby did. "It learns how many sounds there are. It figures that out," the professor said, he then laughed maniacally and continued, "and once it has learned to speak, it will be trained to sing the most beautiful operas ever written."

Although the mere mention of "virtual reality" typically conjures thoughts of gaudy headwear and a fairly good time, Stanford researchers are taking the technology to a much more serious front. A newfangled $25,000 helmet can be used to take victims back to the crime scene, and moreover, can adjust 3D digital busts to give individuals a more accurate look at what the attacker may have looked like at the time of the incident. The virtual police lineup, as it's so aptly titled, enables a virtual world to be opened up in hopes of giving traumatized victims a chance at truly remembering what someone looked like. The weight, height, and basically any other physical attribute of the digital criminal can be altered to give folks a better view, and the wearer can actually approach the busts, inspect their figures, and check out the scar on John Doe's left arm as if the suspects were actually before them. Looks like ski mask robbery just came back into style.

[Via Primidi]

We've all seen them before: high-tech eye-tracking systems that let disabled folks control their computers by adjusting their gaze, or those cheaper models that promise to free up gamers' hands and enable vision-based navigation. However, the problem with current systems -- both medical and recreational -- is that they have difficulty correcting for rapid, unconscious movements of the pupil, making them prone to frustrating errors and giving them only meager functionality. Well that may all be about to change thanks to a Stanford researcher named Manu Kumar and his EyePoint software, which can be used with the same multi-thousand dollar hardware as existing setups, but improves upon their accuracy through pupil-steadying algorithms and by throwing the user's hand into the mix. Someone using a rig powered by EyePoint first looks at the general area of the screen they're interested in, and then presses a key to magnify that area for purposes of editing text or clicking links. Kumar's ultimate goal is to bring eye-tracking hardware and software to the masses at affordable prices, but right now the 20% error rate means his system is still too flaky for everyday use; further refinement of the algorithms to incorporate peripheral vision may help somewhat, but he'll have to get the accuracy above 95% if there's any hope of widespread adoption. Or products like the Project Epoc thought-controlled helmet could end up making a big splash, and completing obviating the need for what is essentially a souped-up early 19th Century technology.

[Via Gadget Lab]

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.

While GPS can do pretty much everything from provide basic map locations to help us engage in fun activities like geocaching, it had never occurred to us that it could also be used in seismology. Fortunately for folks living in quake country, geophysicists have figured out how to do just that. According to a paper that was just presented at the American Geophysical Union in San Francisco, a joint team from Stanford University and from the University of Indiana have figured out a way to use GPS antennae lodged deep in bedrock that can provide a new model for assessing risk of future earthquakes. Armed with these tools, scientists can now determine how quickly various points on the earth are moving, which allows for a better understanding of how tectonic faults shift. But if anyone's knowledge of earthquakes can be improved, it's certainly ours -- we had no idea they get shaken up occasionally over in Indiana.

Our visions of cheap electronic paper may finally come to fruition, if scientists at Stanford University and the University of California, Los Angeles have their way. A joint team from the two universities just published a paper in the journal Nature, which outlines a new technique for mass producing single-crystal organic transistors (previously, the transistors had to be made by hand). According to the researchers, they can print transistors on silicon wafers and flexible plastic, meaning that soon it may be possible to print external inexpensive sensors for commercial products like future generations of LCDs. Exciting times, people, exciting times.

[Via CNET]

Sure we all love robots that can play soccer, crawl through our intestines, or blow things up, but the real holy grail of robotics -- at least to the lazy folks among us -- is a multi-purpose bot that can perform a number of different tasks that humans find distasteful: we need a robot butler. Well thanks to a team of 30+ researchers at Stanford University (the same school that brought us Stanley the Grand Challenge-winning, um, autobot), our dream of one day bossing a domestic bot around the house while we blog from the hot tub may soon come to fruition, as the school's so-called STAIR program has just achieved the first of its four major objectives. Using a webcam and a proprietary algorithm, the STAIR bot is now able to analyze an unknown object and decide how to pick it up based on past experiences; so although it may never have encountered a roll of duct tape, for instance, it can query its on-board database to locate similar objects from which it can form a grasping strategy. The next steps for the robobutler will be significantly more difficult, however, as the project will only be deemed a success when the bot is able to clean up after a party (including loading the dishwasher), assemble one of those frustrating pieces of IKEA furniture using multiple tools, and guide guests around a dynamic space such as a research lab. Just don't expect to see the STAIR bot at your local Circuit City or temp agency anytime soon, as it won't be able to adequately fetch open-air objects until sometime in 2008 and hidden objects until 2011 or later, according to team leader Andrew Ng. Still, we love where this research is headed, because when it comes time for the inevitable mechatronic uprising, our future overlords will be able to calmingly lead us to our dungeons by the hand instead of painfully dragging us by the ear, nose, or other sensitive body part. Click on to check out the robobutler perched on its Segway-based mobility platform...

[Via Robot Gossip]