From new scientist, quote: the infant crawls across a floor strewn with blocks, grabbing and tasting as it goes, its malleable mind impressionable and hungry to learn. it is already adapting, discovering that striped blocks are yummy and spotted ones taste bad. Its exploration is driven by instincts: an interest in bright objects, a predilection for tasting things, and an innate notion of what tastes good. This is how babies explore the world and discover that pink, perky objects exist, and that they produce milk. Hands-on exploration moulds their billions of untrained brain cells into a fully functioning brain…

so begins an interesting story i came across today in new scientist and allusions to perky nipples aside it’s not about biology per se. This particular curious baby isn’t a quivering dollop of lovable milk-fat it’s a trash-can shaped robot a with a mere 20,000 brain cells to call it’s own. it’s a bit more complex than an EG-6 “Gonk” but not quite a FX-7... as far as trash-can esthetics go. its name is darwin 7, another in a line of bots being tinkered with at the neurosciences institute (NSI) in la jolla, california. humble though their looks and brain cell count may be these darwin bots (and their cousins) are pretty fascinating.

here take a gander-

what sets darwin 7 and his kin apart from the gazillion other robotic projects out there is the philosophy behind their a.i.  Essentially the focus is on emulating both the structure and the function of living brains in detail, using a neural-style of processing in combination with robotic movement rather than the elaborately modeled systems most common for the last couple of decades.

quote: The key to Darwin’s abilities is its brain. This is an amalgam of rat and ape brains, encoded in a computer program that controls its actions. Darwin tastes blocks by grabbing them with its metal jaws to see if they produce electricity. It likes the ones that do and dislikes the ones that don’t. Within half an hour of being switched on it learned to find the tasty blocks and has managed to master the abilities of a 18-month-old baby - a pretty impressive feat for a machine.

The dream is that these new brains, embedded in robotic bodies of silicon and steel, will go to a level beyond today’s artificial intelligence systems. By sensing their environments as they explore and learn, they will develop the ability to survive in the constantly changing real world of imperfect information that we navigate so effortlessly, but which computers have yet to master. They will learn to do anything from mundane household chores we’d rather not do, to driving the kids to school, and even autonomously explore Mars or run nuclear waste facilities, all without human intervention. All you would have to do is teach them.

(click thumb for larger version) it may sound like the same ol’ thing, neural nets, etc, but in fact it’s different in a key aspect. these bots are not programmed by instructions like most computer a.i. systems, but instead, like biological systems, they operate according to selectional principles that allow them to adapt to the environment. in this way they form categorical memory, associate these categories with innate value, and adapt. in a certain sense they are more akin to automata than typical robots- simple instructions leading to complex behavior. Most A.I. robots of the past two decades use elaborate modeling systems to describe the real world around them, but these types of a.i. have tend to fail at simple “common-sense” tasks. To make them function, programmers have to tackle the monumental task of anticipating all the likely objects in a robot’s environment and how they might change. the neural robots learn for themselves, develop appetites through conditioning (eventually showing neural “recognition” of these appetites by sight of desired objects alone!) and interestingly continually activate different neural connections to achieve the same task.

quote: Similar to biological organisms, different Darwin VII subjects (i.e. instantiations of the simulated nervous system with slight variations in initial conditions) and cloned Darwin VII subjects with different experiences never displayed identical patterns of neural activity, even during repetitions of the same behavior. However, the adaptive behaviors tend to remain similar. In this respect, Darwin VII is an example of a degenerate system: different circuits and dynamics can yield similar behavior. Degeneracy is a ubiquitous property of biological systems and is necessary for natural selection.

the adaptive, flexible behavior these early “brain-based” systems seem to be exhibiting, which traditional neural nets do not, makes you wonder wether this is the avenue which might eventually lead to true machine intelligence or even sentience? not a fast track to the singularity i wouldn’t think in that “brain-based” robot systems are limited to our own understanding of the real -biological- deal, and it just so happens the brain remains one of the least understood gadgets going.

as with all such stories its import is by no means clear yet. interesting in any case.

the new scientist piece is subscription only but if you’d like to read some more on the subject the (extremely) technical aspects are covered in detail at oxford’s cerebral cortex journal.