Robotic Slime: Problem Solving in Unusual Places
January 31, 2007
Permalink ||

Back in 2000, Nature published a study by Nakagaki, et al. where slime molds, simple amoebae-like organisms, were able to find the shortest route through a maze. How does slime-mold solve a maze? Well, originally it just spread out and filled the entire maze, but since food was only found at the two exits, it shrunk back its 'body' from the parts of the maze that did not connect to food or connected more indirectly, and soon it had contracted into one thick tube that inhabited only the path that directly connected the food (finding the shortest path and avoiding three longer possibilities).

Then in a 2006 study, Tsuda, et al. hooked a slime mold up to a six-legged robot in order to help control the robot in real-life situations. The mold, for example, is light sensitive, so it can help the robot hide in dark corners. This is, of course, an interesting solution to robotics: integrating biology and engineering to solve complex design problems that evolution has already created many elegant solutions for. In the short term, this will likely be the most promising direction for creating robotic machines to solve problems; directly programming in algorithms, even learning algorithms, is harder than one might imagine. But if you can harness natural solutions, integrated into robotics, then you could further evolve better solutions and establish deeper integration, a sort of symbiosis.

What's more, in 2004 a British university developed a robot ("EcoBot II") that generates its own electricity by catching and eating flies (lured by an odor), removing the need for humans to provide regular power. The line between a biological organism and a non-biological 'organism' (to apply a term that seems appropriate despite the oxymoron) will soon be blurred, and we may someday find a whole new ecosystem of robotic creatures, fully or partially under our control. Of course, we humans have shown a remarkable lack of foresight when it comes to altering delicate ecosystems with our technology, so who knows where that will lead.

Regardless of that issue, it is interesting to look at slime molds and bio-engineered robots and see how they solve problems presented to them. We may tend to think of problem solving as something more complex organisms do - at least something on the complexity scale of ants, not an individual cell (largest on Earth though it may be). But perhaps even organisms like cells can be said to have goals (not conscious goals, by any means, but outcomes they have evolved to work toward), so it is interesting to see how they go about attaining those goals, and how efficiently.

It should not be surprising that slime mold can solve a maze problem - finding the shortest route, rather than just continuing to take up the whole maze - because they have evolved to be efficient foragers. Cells in our bodies have evolved to be efficient coworkers in a larger meta-organisms (multi-cellular organism). And even robots are given goals, which they attempt to solve either through hard-programming or learning-based programming or through integration with a biological system that has evolved previously to solve the same problem.

The real question is how soon robots will be given more complex goals (or less straightforward and more vague goals), and how long it will take for them to develop the ability to solve those problems, and to do so more flexibly than they can right now. Eventually they might make use not only of slime mold, but countless other species, adapting useful solutions and throwing out inapplicable details, then evolving more seamless integration and efficiency. At which point we humans will have to take care, because such a development will no doubt radically reshape our world. Not that we aren't ourselves adaptable to drastic changes - indeed, we may have integrated biotechnology into our own bodies by then - but it won't take long before someone tries to use robots against humans.

Indeed, the military is already designing them to assist in all manner of battlefield operations. DARPA, the research arm of the Pentagon, has funded many such proposals, and more funding will no doubt come as these technologies become further developed and more manageable. Of course, these are also the people behind all manner of questionable and privacy invasive technologies. Perhaps we will soon create a problem that even humanity - king of the problem solvers - cannot solve.

The universe extends out from us for billions of light-years in every direction - at least 13 billion, the rough age of the universe (and that ignores expansion). Now, light is the fastest thing we know of in the universe (the speed of light is like a cosmic speed-limit), and a light-year is the distance light travels in one year. So billions of light years is such a huge distance that going as fast as possible, it will takes billions of years before light from that far away reaches us. That's a long, long time, but it should be expected for such a long, long distance. Of course, anything traveling slower than the speed of light (and all mass must travel slower) will take even longer.

Taking this into consideration, it seems hopeless to travel between galaxies as we see in science fiction all the time. I mean, the nearest galaxy - Andromeda - is over 2 million light years away, so traveling slower than light means it would take much longer than 2 million years before a ship we sent over there arrived.

So how is it that something could cross the universe in a matter of days? How is it that we humans could theoretically get to Andromeda in a human lifetime?

In 1991, a cosmic ray detector in Utah registered a proton called the Oh My God Particle. Based on its enormous energy, they deduced its speed:

0.9999999999999999999999951 c (where c is the speed of light)

In other words, it was traveling incredibly close to the speed of light without quite reaching that speed.

Here's the interesting part. When you plug that velocity into the equations of relativity, you find that the particle was traveling so fast that from its frame of reference it could travel to the nearest star in .43 milliseconds. Which may not seem all that weird until you remember that the nearest star is over 4 light years away (it takes 4 years for light to arrive here from there). So how does something traveling slower than the speed of light cross over 4 light years in a fraction of a second?

Well, it turns out that when you are moving, time passes slower than when you are still. It is such a tiny effect that you could not possibly notice it in day to day life, because even our fastest travels are nothing compared to the speed of light. But if you can accelerate something up close to the speed of light, the effect becomes quite dramatic. Time passes more slowly for the traveling mass and length contracts in the direction of motion (meaning the distance to your target becomes shorter - for distance is basically just a measure of how long it would take to get somewhere at a particular speed).

What this means is that if we sent a spaceship out at the speed of this particle, from our perspective watching it back on Earth, it would take about X years to travel X light years. It would not arrive at the nearest star until over 4 years past the launch date. However, from the perspective of the people onboard the ship, time would pass slower, distance would contract, and sure enough they'd reach the star in the same second they got up to full speed. They'd reach the nearest galaxy (billions of light-years away) in a few minutes. They could, in fact, travel the distance of the known universe in a matter of weeks.

Of course, we can't propel mass to those speeds, and if we did the ship would not be able to hold together and survive the first leg of the trip. But it is still mind-opening to plug in these numbers and think about the effects of relativity and to realize that time may not work like we generally think it does. We have no intuitive grasp of relativistic time because we are confined to such slow speeds, but in turn reality is by no means confined to our intuitions.

This site has always been mainly text, because I want to keep it simple (for myself and my readers), keep load time low, appeal to patient, reasoned thought more than attention-grabbing graphics. Thus, I have long been able to get by on minimal space and bandwidth through my host. But recently I've been nearing my bandwidth max (even with mainly text content) and I've gone over the limit on space, so I finally upgraded.

I can now hold a gig of data and transfer 10 GB of data a month, which should be sufficient for a long time to come. It also gives me the option to add more pictures, video and PDFs where those additions would actually enhance the content. I also will probably redesign the general look of the site in the coming month or two, so if you have any suggestions, leave a comment. My goal is still to maintain simplicity as much as possible.

Brain-Computer Interfaces

Follow the link for a video-lecture by researcher Brendan Allison on current work in brain-computer interfaces (BCI), technology for reading brain signals and translating them into messages or commands (e.g. typing words by thought alone). It's aimed at the intelligent layman, not neuroscientists in the field, but it's a nice little overview. I was surprised by some of the developments in quick, clean, gel-less EEG hardware. I still see my peers using huge, ugly EEG scalp caps today, so hopefully that will change soon. More importantly, it's exciting to see the ways they've gotten better at cutting out noise in the signals.

Now, EEGs still don't have the spatial resolution of something like fMRI scans, though they benefit from great temporal resolution. The extent to which EEG's spatial resolution limitations come into play for finer-grained brain readings will probably be offset by advancements in technology before it becomes a huge issue. But it is certainly part of the reason we can only detect simple signals right now. With EEG, we can't tell what word a person is thinking, just things like when their attention changes or when they relax or tense up. The 20-second video below gives a demonstration: two people play Pong without moving any joystick or mouse.

Of course, the technology is being aimed at more than just sending language-based messages on a computer. Not too long ago, scientists succeeded in getting a monkey to control a robotic arm with its thoughts by first matching the robot arm's movements to the brain signals when the animal moves its own arms, and then by training the monkey not to move its own arms but simply control the robot arms. Human experiments along these lines are coming along incredibly fast, and soon quadriplegics may be able to manipulate objects almost as easily as a human with normal biological hands.

Even in its nascent stage, this stuff is really exciting. Not just for its current and near-future implications for, say, physically disabled individuals, and not just because I want to play next-generation virtual reality video games. It's exciting because it suggests our interactions with the world around us will be vastly different after this technology becomes cheap, common, safe and perfected. We may all end up with implanted scanners that can broadcast commands to open doors, turn on lights, or send an instant message or email without having to find a computer terminal and move our hands to type. We may all end up with upgraded semi-robotic arms that can make more precise and powerful movements than a normal human hand, and give more fine-grained feedback. Cochlear implants are so wide-spread for deaf people today that next-generation implants may eventually become standard for even hearing people who wish to hear better or at frequencies normally outside the limits of our perception.

But of course much of that stuff is far away, and today the field is still in its infancy. BCI as Allison presents it is defined in a rather restricted way, as reading brain signals to send messages or commands. However, as he briefly mentions, there is a lot more going on in the field of brain-scanning field and of direct brain-computer connections. For example, legitimate biofeedback techniques (as opposed to over-hyped new-age claims) are becoming more accepted in the medical community for a small number of conditions. They involve patients getting feedback on their physiological state so that they can train themselves to affect their own levels of stress, muscle tension, etc. Neurofeedback, in particular, looks potentially very promising for the future, though as yet it is a young field (and one must always watch out for bad science and pseudo-science attempting to highjack young technologies).

Implications

Of course, despite the positive promises of new brain-interactive technologies, there's also a bothersome and even somewhat frightening element here. If we can scan brains - read data from them - what about writing to them? Will this technology lead to forms of direct mind-control? Recent studies have shown that rats with a few electrodes implanted in their brains can be steered around a maze like remote-controlled cars (they can also be taught to pause when they smell bomb ingredients, a very useful potential application). In 2006, engineers working for the U.S. military announced that they were working on remove-control implants for sharks, which could be used as underwater spies. They've already succeeded in steering spiny dogfish from a laptop across the room. Perhaps radio-controlled squirrels may someday be used to monitor city parks for loiterers, litterers and terrorists, and it may not be obvious which creatures are unaffected and which are spies.

Obviously this is still a very simplistic technology. Leading the rats through a maze consisted in hooking into the cortical cells relating to whisker input and then artificially reinforcing those signals by activating the pleasure center of the rats' brains. Leading the spiny dogfish consisted in stimulating the olfactory center to convince the creature an interesting smell was to one side or the other and make it turn in response.

Yet the technology will get better, and no doubt eventually someone will find a plausible reason why it should be applied to humans. Perhaps they will use it to cease self-mutilating behavior in those with mental disorders, or to induce sleep in insomniacs, or to stimulate the pleasure center in order reinforce alternate, positive behaviors for those with behavioral problems (criminals, children with ADHD, etc.). More likely it will show up first in more subtle ways, as medical treatments for people with brain disabilities. But technology rarely goes back in Pandora's Box once it is released.

A more simple, subtle and plausible threat comes in the form of neuromarketing. Studies at universities and marketing research corporations are using brain scans from fMRIs and other medical technologies to predict buyer behavior and learn how to better influence and manipulate consumers.

Certainly this attempt to get inside the consumer mind is nothing new, but it highlights a trend that has troublesome implications. Along with targeted advertising (that is, spying on consumers and creating/mining databases of their behavior and preferences in order to market to them on a 'personal' level), neuromarketing promises to change the landscape of corporation-consumer interaction, placing the consumer at a disadvantage where their biology is used against them.

Whether current neuromarketing techniques actually lead to more success than simply placing ads in front of people is not clear (they have by no means found a "buy button" in the human brain, so far), but future developments will no doubt keep this a growing field for a while.

The television cartoon Futurama jokingly envisioned a future where advertisements were directly beamed into peoples' dreams while they slept. That may seem implausible today given our current cultural views on privacy, but as technology spreads and changes our culture and our expectations, we may someday see dream-ads as no more strange or off-limits than billboards taking up our visual field or mass-mailed ads showing up in our mailbox. There are significant privacy implications as brain-computer interface technology becomes ubiquitous, pervasive and embedded (i.e. widespread, invisible integration).

That is not to say that these technologies are bad. Tools in general aren't easy to label good or bad on their own, but only when they are applied in certain ways. There will no doubt be both positive and negative effects of these technologies, and we will just have to do what we can to mitigate the latter. Our best resistance right now is to become aware and informed, and plan ahead. But one thing is clear: change is coming, and however fast or slow, it will reshape our society.

I added a new section on Learning and Behavior Analysis to the science section, with a couple articles. The first, The Rat in Your Slot Machine, deals with how gambling establishments take advantage of animal behavior research to manipulate peoples' behavior. The second, Controlling Behavior, deals with reward and punishment: how effective are they, should we only use positive reinforcement, and are reward and punishment even in practice distinct?

This video may be long, but it is well worth listening to. I suspect if this lecture was given in every intro philosophy, politics or history course in college, it would reshape America pretty quick. Sam Harris takes on the sociological effects of religion in modern American life, looking both at the fundamentalists and moderate believers.

His argument about moderates is frightening in how convincing it is, because that is not how I've tended to see things. He makes interesting points, at least. I'm curious what other people think about this part of his argument.

At any rate, his larger, more general points about religion and reason, religion and science, are pretty well-reasoned and well-spoken. Food for thought at least.

Added a review of Dhalgren, Samuel Delany's frustrating and wonderful labyrinth-book, to the fiction section.

From John Gribbon's Little Book of Science:

If you hang a bucket of water from a long twisted rope, and let it start to spin as the rope untwists, at first the bucket rotates but the water stays still. Gradually, friction makes the water rotate, and as it does so it forms a concave surface, produced by centrifugal force. Now grab the bucket and stop it moving. At first the water keeps moving, and stays concave. How does the water 'know' when it is rotating, and ought to have a concave surface? It certainly doesn't measure its rotation relative to the bucket, because it can be concave when both bucket and water are moving, or when the water's moving and the bucket is not.

This highlights the puzzle of inertia, which the Irish philosopher, mathematician and bishop George Berkeley explained in the early eighteenth century by saying that objects measure their motion (somehow) relative to the most distant objects in the Universe. He referred to the distant stars, but today we would think in terms of distant galaxies.

The idea was developed further by Ernst Mach in the 1860s, and is often known as Mach's principle, but didn't quite succeed. Even so, there seems little doubt that when I try to push a child on a swing, and have to use energy to make it start to move, somehow the influence of all the matter in the Universe, in all the stars and galaxies, resists my efforts and provides the inertia that I have to do work against. The coffee in your cup rises up the side of the cup when you stir it because it knows that the distant galaxies are there.

Which at this time of night leads my thoughts down the following track. Does gravity act across all that distance instantly? I mean, matter essentially bends space-time, so all the mass of far-off galaxies is bending space-time such that it tries to bring our planet closer in space-time to it (literally a force that pushes it toward being in our future). Presumably these balance out on the largest scales, but our solar system, and even galaxy, and even supercluster of galaxies is moving, whether from gravitational pull, or more likely from rotation imposed by earlier (eventually) cosmological events.

But anyway, back to my original question: does the bending of space-time happen instantly? I guess I'm asking: if a huge new mass instantly came into being off to one side of our solar system but close enough to exert a significant pull, would our solar system feel the pull (register different movement paths than they were on a moment before) right away? Or only after some (short) time related to the distance? After all, we know the strength of the gravitational force is inversely proportional to the square of the distance, so maybe the force would only exert its pull after enough time had passed.

I guess my example - matter suddenly coming into existence - only works if we're somehow converting energy into matter (conservation and all). Energy doesn't exert a gravitational pull, does it? So is the setup even theoretically plausible by any stretch of the imagination? Or did I just set myself up with a nonsensical question?

Either way, it's mind-numbingly cool to keep coming back to the vast idea that all the matter in the universe is exerting a tugging force on you in every direction.