I've been thinking for some time about how to get people to walk in a virtual environment. My idea was to have a floor of rollers, like large ball bearings, with the user supported by a hanging seat, akin to those in stand-up roller coasters. The movement of the ball on the floor could then be translated into movement, like upside-down ball mice. But these researchers came up with an even better idea.

Plus, the article touches on one of my favorite new words: proprioception. Because man cannot live by exteroception alone.

This article has been bouncing around the nerdosphere, so I thought I'd give it my two cents. There's definitely a creepy "Minority Report" vibe with this technology, but I think it speaks to really interesting and more positive trend. For centuries, we've been using machines to augment our senses--spectacles, telescopes, earhorns, microphones, electon microscopes, etc. In fact, it can be said that the advent of literacy and mathematics were among the first cases of technological augmentations to our cognitive abilities--extending our memory and logical capacity beyond what we can process from moment to moment or over the course of a lifetime.

The DARPA project takes the next logical step, culling our brain activity to find the patterns that will aid image recognition. Imagine, in the near future, getting an instant read out of an object or person you're looking at before you actually "know" what you've seen. What will that do to us as humans and as a society? I haven't the slightest idea. Will it help us become ultra-sensitive to our environment? Will it be information overload (or is there even such a thing, given how malleable we keep proving to be)? Will it have a negative effect, reducing the amount of actual memorization we do and "softening" our brains? Or is human memory no great shakes to begin with and, like the invention of writing, will "precog" computer-aided senses expand our abilities to point we can not yet even imagine?

I feel like I say this all the time: I have no idea where we're going or if I'm going to like being there, but it's coming, and I'm excited to see how it all plays out.

Via Slashdot comes what seems to be yet another of my pet links on robotics and AI. Researchers in Italy have gotten AIBO "dogs" to talk about a ball and some of the things it does.

Talking robo-dogs are cool and all, but so what? Well, the dogs came up with the words. On their own. And the grammar to describe it to each other. After a while of "babbling" to each other, they formed a consensus on what they saw and what it did.

The consequences of this are deceptively profound and widespread. For one, I know from the resident psych expert that one of the schools of thought on the purpose of education is the "Interpretationist" perspective. Interpretationists, as opposed to functionalists and those descended from Marxists, see the purpose of education as forming a shared interpretation between students who initially have wildly different subjectivities. What this AI research might suggest is that the Interpretationists are correct in ways that go deeper than expected. If learning even at a relatively simple machine level can be effectively modelled and, in so doing, show how much of the "deep structure" of learning might be this consensus building exercise, could AI be on the right track at last to figuring out the right analog between machine and biological learning and, tentatively, thought?

There's a lot more going on here, too, particularly with what the scientists are doing with what appears to be a variation of subsumption architecture, which I've written about recently. But I need to let this sink in for a while.

A couple of interesting stories from one of my new favorite blogs, "we make money not art." First, there's the robo-gecko with super sticky gecko feet. More about it at New Scientist. Of particular interest is its use of synthetic setae that increases the surface area of its feet, which is the secret of the "sticky" quality.

Another interesting robotic story is about "Scooby-Doo," the nickname for a bomb-defusing robot that served in Iraq. Its technician grieved for it when it was blown up after completing 35 missions.

Humans have a long history of forming emotional bonds with nonhuman objects and creatures. Religious expression is, arguably, an example of this, though, really, so is the core of our ability to think metaphorically. Having a connection with something involves us identifying enough of the human in the Other to be able to establish that.

Roboticist Masahiro Mori's term, the Uncanny Valley, is an attempt to quantify how much "humanity" is required to make this connection...and how much is too much. My feeling is that the "valley," where people start to distance themselves from a simulated human, is due to the simulacrum interfering with our ability to make metaphors for the object that create the emotional connection. That is, if it's too real, our minds won't let us close the gap between what we see, which is not human, and what we need to experience in order to make the object familiar.

The fact the "Scooby Doo" received an emotional connection from its technician may mean that it's not just looking human that matters. The soldier said that the robot had saved his life on several occasions--and, ultimately, died doing so again. The machine itself has no capacity for nobility or courage, but it was able, in the course of its life, to demonstrate a simulation of those qualities--enough of them, at least, to be mourned when it was destroyed.

Where does this leave us, then? I think that, when we design the next generation of robots, software, interfaces, etc., we should consider and research the human-like qualities that attract people to an object, beyond just the appearance, like a Star Wars droid, and a few canned phrases, like an ATM. I'm not sure what those qualities are just yet, but I think it's an area of research that crosses psychology and technology at what might finally be the right time to start getting some good answers.

As it now appears that humanity's collective goal is to eventually produce hot, hot cylons, we welcome another step along the path--Babybot! I'm especially excited that their next project, Robocub, will be open source, including the hardware designs.

I'm a big fan of what I understand to be the approach Rodney Brooks takes to artificial intelligence, which is to eschew the more classical avenues to machine intelligence and focus on simple sensory and motor tasks that can built into more complex behaviors. I'd liken it to the behaviorist/cognitive division in psychology (the latter of which was influence by early AI work, among other things), though my wife would have much more to say about that in a much more informed way.

By the way, if you haven't already seen Errol Morris's stunning look at life and eccentricity, Fast, Cheap, and Out of Control, which features Brooks, definitely check it out. The title comes from something Brooks came up with for a fleet of tiny Mars exploration robots.

Very cool engineering solution for a hikers' bridge where the only two connection points have very different elevations. Check out more pictures of its construction.

The roboblogging continues. Check out this biped from Tokyo's Waseda University. Just remember to keep evil penguins away from it.

Holy synaesthesia! The military is working out a sensor system that would be attached to a diver's tongue, allowing him or her to "see" sonar, in addition to what he or she can see visually, by stimulating nerves with 144 microelectrodes.

I really, really want to play with this. Just imagine the video game possibilites! A whole new addition to the sensorium--you could even create a "Spidey sense" for the player!