As for many other things in life, money is one of the reasons...

Robots used to be very expensive, and only few companies or governmental institutions with deep pockets had the economic firing power to research, prototype, and build them.

Things have changed thanks to lower barriers to entry in robotics, such as the rise of cheap rapid-prototyping tools like 3-D printers, and mobile computing power enabled by the cell phone and tablet industry.

All these factors combined are dramatically lowering the barriers to cheap, powerful robots. Read the rest of the story on CNN.

Read the article here.

An article on Information Week proposes that the next big hit in robotics will come from Telepresence. With a shower of remote telepresence robots hitting the market over the past few years and more to come (MantaroBot, Vgo, Anybots, Suitable Technology, Double Robotics, Revolve), and with price points below $10K (and sometimes below $2K!), the positive repercussion of telepresence on productivity, or simply the ability to chat hands-free with your friends and family, would become apparent.

Telepresence is surely one of the next big sectors in robotics to be booming in the next few years: new, inexpensive gadgets will become cheaper and more powerful. And more useful.

It is surely a legitimate question. There have been endless reports in virtually every single media we know of telling us the time is ripe for the robot invasion to begin. And indeed, there is tremendous amount of private and public funding and research that is pushing the boundary of robotics forward. So, where are they? Why are robots still relegated mostly to clean our floors? Provided that this is per se a nobile job which save us tremendous amount of time and back pain, when are the robots going to pop out from Youtube videos and invade our world (be sure what we wish for!). And why have they not already?

For robots to occupy a spot in our daily life and fulfill their promises, three main things have to materialize. I usually split these three "miracles" into three categories, for simplicity. I call them MIND, BRAIN, and BODY. Each one has its challenges, and each is indispensable to achieving the goal of robots escaping Youtube. Let's see why.

Of course, robots need body, which is the most immediate manifestation of their presence. Robotic bodies are getting cheaper and cheaper: nowadays you can buy a fairly good robot for a few hundreds dollars. Even components that once were extremely bulky, expensive, and power-hungry, such as arms, are being designed in cleaver ways, such as inflatable arms.

How about BRAINS? What is a robot brain, anyway? Well, to design a robot with a meaningful set of behaviors, you need to include an equally significant subset of skills exhibited by natural intelligence. This requires considerable computing power, to a degree that just a few years ago was unthinkable in a power envelope, size, and cost amenable to widespread and affordable mobile robots. All this has changed, mostly thanks to advances in mobile computing (e.g., cell phones and related technology). Nowadays, considerable computing power is packed in cell phone based devices, with dual and quad core low power processors and associated mobile graphic processing cards providing juice for computing challenging algorithms.

How about MIND? I call MIND smart algorithms. These are software programs simulating aspects of perception, decision making, navigation, and motor control which enable intelligent behavior (e.g., see here). As you can intuit, the design of a MIND is tightly coupled to the availability of both a BRAIN to run on, and a BODY to be tested on.

As with many scientific discoveries and technological advances, synergy is the key. The combination of powerful algorithms (MIND), cheap, ~$100 mobile computing hardware enabling these algorithms to run (BRAIN), and affordable robots (BODY) are all enabling technologies for the growing robotic industry. Remember how the introduction of the microprocessor slashed the cost, increased the density of circuits and the performance of computer hardware, resulting in new applications and designs?

A similar paradigm change is happening now, which will ultimately allow robots to leave Youtube and enter the real world.

Actually, a brilliant critique of Gary Marcus on The New Yorker says it all. I agree with all the arguments that Gary presents. The idea illustrated in the book, and in other books (e.g., On Intelligence by Jeff Hawkins) are, in different proportions, re-chewed ideas that have been around in A.I., psychology, and neuroscience for decades.

Not only there is nothing new to the theory of the brain as a hierarchical patter recognition machine. Other than a new fancy name (once you name it, you own it!).

But here is a tip for Neurdon readers. Simple, and plain: do not buy a book that claims to tell you how the brain works. Why? It's obvious: if somebody really knew that, he/she would build it and do billions with it, since as Ray himself acknowledges “reverse-engineering the human brain may be regarded as the most important project in the universe.”

So, somebody really knows how to build a brain, the next thing to do is... go build it! There are plenty of good programmers out there in search of jobs, and wealthy individuals who know how to build a brain would have no issue in hiring a handle of them to implement the latest hierarchical pattern recognition machine.

Here we go, $20 saved! Merry Christmas!

Read more: http://www.newyorker.com/online/blogs/books/2012/11/ray-kurzweils-dubious-new-theory-of-mind.html#ixzz2G65UW4ut

A recent articles on Wired, "Almost Being There: Why the Future of Space Exploration Is Not What You Think", makes a very nice job disentangling what would be the role of humans and robots in space exploration, and how their complementary capabilities can nicely interlock to provide economically viable ways to explore the Moon, Mars, and beyond, without putting astronaut's life at huge risks.

Economy is also a very important factor: if the mission is too expensive, it won't just happen. But humans, with their extremely advanced brains and motor skills, are still above what robots can provide at the present day in terms of autonomy and dexterous manipulation. Well, at least that's the point of the article, which suggests how teleoperation of robots by humans orbiting a planet or a satellite can be more viable economically than landing humans on the surface, while reducing communication delays to a fraction of what currently scientists deal with from Earth.

This is all true, but the article assumptions are still that autonomy is fairly limited, and therefore teleoperation is required. While teleoperation from close ranges would be a step forward with respect to state-of-the-art, this solution would not scale up with dozens or robots simultaneously roaming the surface. Autonomy, and with it cognitive-perceptual-planning-motor capabilities in robots are still a huge desiderata, and inescapable need for space exploration (see for instance here). The truth will most likely lie in the middle, with a mixture of teleoperation and autonomy becoming more pervasive in future years.

from the article:

Attempts to simulate the brain usually involve programming software to behave like groups of neurons. A new "neuromorphic" design instead tries to recreate the brain's hardware, using analogue components last seen in the early days of computing. "On our system, you can physically point to the neuron," says Karlheinz Meier of the University of Heidelberg in Germany.

Read more: http://www.newscientist.com/article/mg21628925.100-brainlike-chip-outstrips-normal-computers.html

The old way...

But, like the Texas penal system,what we've got is an execution problem. There is no easy way to get those electrodes to the cells. And once they're in, there are still risks of infection or rejection. The signal is also subject to degradation by movement, electrical noise, or electrode deterioration. These are problems routinely faced by researchers working with animals, and are only magnified tenfold when the possibility of human applications arises. Right now, non-invasive techniques such as EEG and fMRI are the most common methods of reading brain activity in humans. But none of those options have the spatial or temporal resolution to provide meaningful data about real-time information processing, which is critical for the goals of any BCI. For useful data, we need those tiny electrodes. But as long as implanting them remains such an invasive, messy, and potential dangerous procedure, it won't be available to the masses.

Now, it should be noted, there is the minor matter of knowing what those neural signals that we record actually mean. Or figuring out how to make a pattern of stimulation that has the intended effect. Cracking the neural code is obviously key to interfacing with the brain; it's hard to have a conversation when you don't speak the language. And right now we are far from fluent. But it is an area of huge research and I believe the breakthroughs are coming. However without the ability to utilize that knowledge in a physically plausible way, we don't stand much to gain from it. And that is why our lack of a good implementation is so problematic.

So, what then is the future of BCI execution? I'd put my money, if I had any, into nanotechnology. And not (entirely) because it just sounds futuristic. The fact is, the process of opening up the skull, placing a hard and very foreign object onto the brain, and closing it back up again is never going to be a clean one. We need our BCI devices to be more natural, and to deliver them in a non-invasive way. And the only way to do that is if they are very small, and made of some more bio-friendly materials. And that is exactly what nanotechnology researchers such as John Rogers are working on. His group's recent Science paper (and his appearance on NPR) describes a biodegradable electrode, which could theoretically be implanted in the brain and utilized for diagnostic purposes and then be allowed to dissolve away safely. Longer lasting versions could be used for months or even years. Furthermore, these kinds of 'soft' electrodes can fit better to the curves of the brain, leading to better signal quality. Another nanotech company has developed a coating for traditional electrodes the enhances the signal quality and longevity by reducing the response of the immune system to the electrode. So, through these measures some of the danger of neural implantation can be reduced and the quality of recordings increased.

The future!

There is still, however, the issue of delivery. But, do not fear; nanotechnology has the answer to that too! Well, potentially. Because nanomembranes are so small, thin and pliable, they can easily survive being delivered via injection (as this study looking at nano-scaffolds for bone regeneration shows). There is also already precedence for the injection of neural-stimulating electrodes. The BION system involves injecting traditional electrodes into peripheral nerves via a hypodermic needle. Those electrodes are then powered and controlled wirelessly, and allow for stimulation and recording of motor neurons. Now to translate this method into a useful BCI application would require getting the electrodes into the brain, not just the peripheral nervous system. And that is a problem of another magnitude, not easily solved even with nanomembranes in your toolbox. The blood-brain barrier is just annoyingly particular about what it lets thorough. And even if you could get the nanomembrane electrode in, you'd need to have a way to control where it plants itself. Trying to move a prosthetic arm with neural signals coming from your visual cortex is not advisable. So the injection would probably have to be targeted, meaning there are still some serious obstacles to a completely non-invasive procedure. But we have to leave some problems for future scientists...

Maybe the notion of tiny, injectable brain-controlling devices sounds crazy (or terrifying) to you. But it is the general direction we must go in if we want to really utilize all the knowledge that we're painstakingly gathering about how the brain works. Potential applications aren't limited just to disease treatment or prosthetic limb control. Even perfectly healthy people could find BCI beneficial. The gaming industry, for example, has already wet its feet in the BCI pool, using EPOC headsets to add another element of control to games. But these superfluous applications are only sensible if the BCI is incredibly low-risk. No one would endanger their health to make World of Warcraft slightly more entertaining... Ok, maybe some people would, but the FDA isn't going to approve it. So we are a long way off from BCI impacting the everyday life of the average person. But all crazy technology had to start somewhere. I'm excited to see how this particular one develops, and how quickly we can get ourselves into the future.

via Neurdiness

The idea is pretty simple, though it is remarkable: “neurons represent information and compute as they form cell assemblies”. This notion is quite old, going back to early- and mid-twentieth century. The first evidences probably came from observations of muscles activities, once increasing or decreasing the number of active motor units changes the amount of force produced by a muscle.

In 1949, Donald O. Hebb suggested that co-activation of ‘cell assemblies’ can be responsible for representing concepts. Thus, the concept is old and the neuroscience literature has plenty of examples about neural cell assemblies.

One neuron alone can be thought as a dynamical system, which behaves as an instable and noisy computational unity. When neurons fire in groups such ‘weaknesses’ disappear. In this sense, the ‘sparse coding’ is a well-accepted concept in which neurons ‘codify’ external and internal states of the world by firing in gathered coalitions.

But the question is: how do cell assemblies represent, memorize and compute such information in order to control behaviors?

This is  what the Neural Assembly Computing (NAC) approach tries to explain. In a brief overview it is possible to resume the concept as follows:

1. Spikes do not propagate instantaneously along the axons, so there are delays that must be considered in spiking neural networks;

2. The conjunction of propagation delays, synaptic weights, and network interconnections (the topology) makes a single spike to spread and reach many other neurons at different instants;

3. As spikes reach other neurons at different instants with different strengths, sets of neurons naturally fire together. They can fire as synfire chains (synchronously) or as polychronous groups, firing time-locked. Note that the cell assembly is an ephemeral phenomenon.

Based on these principles, previously theorized by Izhikevich and Hoppensteadt in Polychronous Wavefront Computations, the NAC framework proposes that:

4. As such neural coalitions happen they interact with other coalitions and logical functions are performed:

4.1 A single assembly is able to trigger another assembly, or a single assembly can trigger more than one assembly creating parallel processes (this is called branching);

4.2 Sometimes an assembly A or an assembly B can independently trigger a third assembly C. It means that A OR B is able to trigger C (or both). This is equivalent to the logical function OR (we have used the Boolean notation C=A+B, which is read C is caused by spikes from A or from B);

4.3. In other situations, the spikes from an assembly A are not strong enough for triggering C alone, and the same may occur with the spikes from an assembly B. However, when spikes from A AND B occur coincidently they trigger C. It means that this interaction is performing the logical function AND (in Boolean notation C=A.B, which is read C is caused by spikes from A and B simultaneously).

Considering that assemblies stand for something, i.e. they ‘represent’ external or internal objects or states, nervous systems would not let such ephemeral events disappear. Once assemblies represent some (important) information, how could nervous systems let these events to extinguish?

Hence, cell assemblies must interact with other assemblies in order to retain important representations, events, states, etc.

5. Thus, assemblies reverberate with other assemblies creating memory loops. It means that one bit of information can be retained by a chain of assemblies with feedback: A triggers B that triggers C that triggers back the assembly A. We call these reverberating loops Bistable Neural Assemblies (BNA). Note that such loops have not to do with plasticity mechanisms, and it is not necessary to change synaptic weights for instantiating this kind of memory.  In thesis, such loop would remain firing indefinitely.

6. Therefore, it becomes necessary to dismantle branches and established BNAs. Note that the role of such inhibitory assemblies is similar to the NOT logical function. Therefore, when an assembly D inhibits a branch or a BNA we say that D is executing the NOT logical function and dismantling the established branch or BNA.

7. It is possible that two assemblies (A AND B) execute an inhibition of a branch or a BNA. It means that singly neither A nor B can inhibit the assembly C, but together they can do that. So they are performing the NAND logical function (an AND associated to a NOT function). On the other hand, an assembly A OR an assembly B may be capable to perform inhibition independently, so they are performing the NOR logical function.

These are the elements necessary to create computers!

The logical gates (AND, OR, NOT, NAND and NOR) associated to the memory, which in digital circuits are performed by flip-flops, are the basic elements used by engineers to construct computers. By using these elements engineers are also able to create Finite State Machines (FSM), the first step on constructing serial machines.

The great advantage in NAC is that it is possible to create a large number of parallel FSMs in the same substratum: the spiking neural network. Such parallel FSMs can interact and this opens a new perspective in creating 'real parallel processing' machines in spiking neural networks.

Moreover, note that the neural assemblies are both ‘the representation’ and the ‘control’ element for the computational flux. In other words, in NAC the groups of firing neurons ‘represent’ things and states, and at the same time they ‘control’ how information are processed. The paper in which these ideas are introduced is:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6186825

I've created a blog for publishing correlated works:

http://www.neuralassembly.org/

In this site there is a short animation (2'30'') showing the NAC fundamental ideas, and the video can also be seen at:

Matlab codes are available from this site, so other researchers can reproduce the experiments. I’ll try to let a short tutorial available for each code published there. The code for fundamentals and FSM are available, although the article explaining the FSM on NAC has not been published yet (“Neural Assembly Executing Finite State Machines”).

The NAC framework is quite recent, but I visualize useful machines being created by using this approach. By October 2012, I have made few tries in order to insert STDP and other neural plasticity mechanisms within this framework. Therefore, the machines I have worked on (so far) are mainly deterministic.

The neural ‘tuning’ (for timing and synaptic weights) is obtained experimentally by generating candidate topologies. For instance, FSMs are ‘designed’ by using both Mealy and Moore’s method, so the candidate topologies come from a well-established knowledge and methodology. Then, the final tuning is reached by realizing small changes on synaptic weights and propagation delays, and by selecting the well-succeeded topologies which match propagation delays and synaptic weights for performing the desired computation.

There are lots of issues to be investigated starting from the NAC framework.

The neuromorphics lab and the ECE department at Boston University are collaborating with NASA Langley to use animal visual systems as inspiration for the development of electrical circuits that use a camera to spot distant aircraft and, if necessary, adjust a flight path to avoid any accidents. Such a system could be used as an additional warning system for pilots, similar to a computer assisting a radiologist in identifying suspicious image regions on a CT scan. These systems can also be used to help pilot unmanned aerial vehicles (UAV's), completely robotic aircraft that will only be allowed to share the airspace with people once their flying abilities are comparable to human pilots.

The project's central theme leverages optic flow to allow an aircraft to understand how itself and other objects are moving through the world. Optic flow is visual motion that accompanies body movement; for example, the optic flow produced by driving toward a tunnel is the expansion of the tunnel's entrance over time. Moving in different ways produces different patterns of optic flow; moving to the right makes the whole world shift to the left, which can be distinguished from the expansion of the world experienced when moving forward. Because of this correspondence between body movement and the experience of optic flow, mathematical models can suggest how to run the equation in reverse — how to estimate where my body is moving given the optic flow that I see. While we don't know which ‘equation’ our brains use, we know that we have this capability: a person can easily drive a car in a video game where she sees the screen but lacks the kinematic sensation of having her body move with the virtual car.

From http://opticflow.bu.edu

Honeybees use optic flow for, among other strategies, compensating for wind when searching for flowers and for maintaining a constant height above the ground; these methods have proved equally simple and effective for slow-moving robots navigating tight spaces and for helicopter hovering (YouTube). Quickly and accurately determining optic flow from a camera, however, requires intense computing power. UAV's cannot carry heavy payloads, so the team plans to eventually condense the artificial visual system, including the optic flow component, into FPGA, a lightweight programmable hardware platform.

While some knowledge about the world can be directly programmed into the artificial visual system, recognizing distant objects as potentially dangerous may best be learned through practice. We currently use FlightGear for simulator training, but in the end there's no better training experience (nor a better way to develop a neuromorphic system) than to allow the system to test pilot an actual aircraft, the eventual goal of the project.

NASA may eventually use UAV's to automatically collect weather statistics; more generally, UAV's can be used for a variety of dangerous flight situations like firefighting. Making autonomous flight easier, however, can also encourage illicit and immoral UAV use. Drone aircraft military strikes can breach humanitarian law, and UAV surveillance by police units can erode privacy protections. Drone aircraft still require a remote human operator, but increasingly automated flight breaks new ethical ground that would be better met by proactive ethical engagement throughout society. We hope that, by creating systems that make manned and unmanned flight safer, the world may also become a safer place to live.