Tag Archives: signal

#432884 This Week’s Awesome Stories From ...

ROBOTICS
Boston Dynamics’ SpotMini Robot Dog Goes on Sale in 2019
Stephen Shankland | CNET
“The company has 10 SpotMini prototypes now and will work with manufacturing partners to build 100 this year, said company co-founder and President Marc Raibert at a TechCrunch robotics conference Friday. ‘That’s a prelude to getting into a higher rate of production’ in anticipation of sales next year, he said. Who’ll buy it? Probably not you.”

Also from Boston Dynamics’ this week:

SPACE
Made In Space Wins NASA Contract for Next-Gen ‘Vulcan’ Manufacturing System
Mike Wall | Space.com
“’The Vulcan hybrid manufacturing system allows for flexible augmentation and creation of metallic components on demand with high precision,’ Mike Snyder, Made In Space chief engineer and principal investigator, said in a statement. …When Vulcan is ready to go, Made In Space aims to demonstrate the technology on the ISS, showing Vulcan’s potential usefulness for a variety of exploration missions.”

ARTIFICIAL INTELLIGENCE
Duplex Shows Google Failing at Ethical and Creative AI Design
Natasha Lomas | TechCrunch
“But while the home crowd cheered enthusiastically at how capable Google had seemingly made its prototype robot caller—with Pichai going on to sketch a grand vision of the AI saving people and businesses time—the episode is worryingly suggestive of a company that views ethics as an after-the-fact consideration. One it does not allow to trouble the trajectory of its engineering ingenuity.”

DESIGN
What Artists Can Tech Us About Making Technology More Human
Elizabeth Stinson| Wired
“For the last year, Park, along with the artist Sougwen Chung and dancers Jason Oremus and Garrett Coleman of the dance collective Hammerstep, have been working out of Bell Labs as part of a residency called Experiments in Art and Technology. The year-long residency, a collaboration between Bell Labs and the New Museum’s incubator, New Inc, culminated in ‘Only Human,’ a recently-opened exhibition at Mana where the artists’ pieces will be on display through the end of May.”

GOVERNANCE
The White House Says a New AI Task Force Will Protect Workers and Keep America First
Will Knight | MIT Technology Review
“The meeting and the select committee signal that the administration takes the impact of artificial intellgence seriously. This has not always been apparent. In his campaign speeches, Trump suggested reviving industries that have already been overhauled by automation. The Treasury secretary, Steven Mnuchin, also previously said that the idea of robots and AI taking people’s jobs was ‘not even on my radar screen.’”

Image Credit: Tithi Luadthong / Shutterstock.com Continue reading

Posted in Human Robots

#432671 Stuff 3.0: The Era of Programmable ...

It’s the end of a long day in your apartment in the early 2040s. You decide your work is done for the day, stand up from your desk, and yawn. “Time for a film!” you say. The house responds to your cues. The desk splits into hundreds of tiny pieces, which flow behind you and take on shape again as a couch. The computer screen you were working on flows up the wall and expands into a flat projection screen. You relax into the couch and, after a few seconds, a remote control surfaces from one of its arms.

In a few seconds flat, you’ve gone from a neatly-equipped office to a home cinema…all within the same four walls. Who needs more than one room?

This is the dream of those who work on “programmable matter.”

In his recent book about AI, Max Tegmark makes a distinction between three different levels of computational sophistication for organisms. Life 1.0 is single-celled organisms like bacteria; here, hardware is indistinguishable from software. The behavior of the bacteria is encoded into its DNA; it cannot learn new things.

Life 2.0 is where humans live on the spectrum. We are more or less stuck with our hardware, but we can change our software by choosing to learn different things, say, Spanish instead of Italian. Much like managing space on your smartphone, your brain’s hardware will allow you to download only a certain number of packages, but, at least theoretically, you can learn new behaviors without changing your underlying genetic code.

Life 3.0 marks a step-change from this: creatures that can change both their hardware and software in something like a feedback loop. This is what Tegmark views as a true artificial intelligence—one that can learn to change its own base code, leading to an explosion in intelligence. Perhaps, with CRISPR and other gene-editing techniques, we could be using our “software” to doctor our “hardware” before too long.

Programmable matter extends this analogy to the things in our world: what if your sofa could “learn” how to become a writing desk? What if, instead of a Swiss Army knife with dozens of tool attachments, you just had a single tool that “knew” how to become any other tool you could require, on command? In the crowded cities of the future, could houses be replaced by single, OmniRoom apartments? It would save space, and perhaps resources too.

Such are the dreams, anyway.

But when engineering and manufacturing individual gadgets is such a complex process, you can imagine that making stuff that can turn into many different items can be extremely complicated. Professor Skylar Tibbits at MIT referred to it as 4D printing in a TED Talk, and the website for his research group, the Self-Assembly Lab, excitedly claims, “We have also identified the key ingredients for self-assembly as a simple set of responsive building blocks, energy and interactions that can be designed within nearly every material and machining process available. Self-assembly promises to enable breakthroughs across many disciplines, from biology to material science, software, robotics, manufacturing, transportation, infrastructure, construction, the arts, and even space exploration.”

Naturally, their projects are still in the early stages, but the Self-Assembly Lab and others are genuinely exploring just the kind of science fiction applications we mooted.

For example, there’s the cell-phone self-assembly project, which brings to mind eerie, 24/7 factories where mobile phones assemble themselves from 3D printed kits without human or robotic intervention. Okay, so the phones they’re making are hardly going to fly off the shelves as fashion items, but if all you want is something that works, it could cut manufacturing costs substantially and automate even more of the process.

One of the major hurdles to overcome in making programmable matter a reality is choosing the right fundamental building blocks. There’s a very important balance to strike. To create fine details, you need to have things that aren’t too big, so as to keep your rearranged matter from being too lumpy. This might make the building blocks useless for certain applications—for example, if you wanted to make tools for fine manipulation. With big pieces, it might be difficult to simulate a range of textures. On the other hand, if the pieces are too small, different problems can arise.

Imagine a setup where each piece is a small robot. You have to contain the robot’s power source and its brain, or at least some kind of signal-generator and signal-processor, all in the same compact unit. Perhaps you can imagine that one might be able to simulate a range of textures and strengths by changing the strength of the “bond” between individual units—your desk might need to be a little bit more firm than your bed, which might be nicer with a little more give.

Early steps toward creating this kind of matter have been taken by those who are developing modular robots. There are plenty of different groups working on this, including MIT, Lausanne, and the University of Brussels.

In the latter configuration, one individual robot acts as a centralized decision-maker, referred to as the brain unit, but additional robots can autonomously join the brain unit as and when needed to change the shape and structure of the overall system. Although the system is only ten units at present, it’s a proof-of-concept that control can be orchestrated over a modular system of robots; perhaps in the future, smaller versions of the same thing could be the components of Stuff 3.0.

You can imagine that with machine learning algorithms, such swarms of robots might be able to negotiate obstacles and respond to a changing environment more easily than an individual robot (those of you with techno-fear may read “respond to a changing environment” and imagine a robot seamlessly rearranging itself to allow a bullet to pass straight through without harm).

Speaking of robotics, the form of an ideal robot has been a subject of much debate. In fact, one of the major recent robotics competitions—DARPA’s Robotics Challenge—was won by a robot that could adapt, beating Boston Dynamics’ infamous ATLAS humanoid with the simple addition of a wheel that allowed it to drive as well as walk.

Rather than building robots into a humanoid shape (only sometimes useful), allowing them to evolve and discover the ideal form for performing whatever you’ve tasked them to do could prove far more useful. This is particularly true in disaster response, where expensive robots can still be more valuable than humans, but conditions can be very unpredictable and adaptability is key.

Further afield, many futurists imagine “foglets” as the tiny nanobots that will be capable of constructing anything from raw materials, somewhat like the “Santa Claus machine.” But you don’t necessarily need anything quite so indistinguishable from magic to be useful. Programmable matter that can respond and adapt to its surroundings could be used in all kinds of industrial applications. How about a pipe that can strengthen or weaken at will, or divert its direction on command?

We’re some way off from being able to order our beds to turn into bicycles. As with many tech ideas, it may turn out that the traditional low-tech solution is far more practical and cost-effective, even as we can imagine alternatives. But as the march to put a chip in every conceivable object goes on, it seems certain that inanimate objects are about to get a lot more animated.

Image Credit: PeterVrabel / Shutterstock.com Continue reading

Posted in Human Robots

#431836 Do Our Brains Use Deep Learning to Make ...

The first time Dr. Blake Richards heard about deep learning, he was convinced that he wasn’t just looking at a technique that would revolutionize artificial intelligence. He also knew he was looking at something fundamental about the human brain.
That was the early 2000s, and Richards was taking a course with Dr. Geoff Hinton at the University of Toronto. Hinton, a pioneer architect of the algorithm that would later take the world by storm, was offering an introductory course on his learning method inspired by the human brain.
The key words here are “inspired by.” Despite Richards’ conviction, the odds were stacked against him. The human brain, as it happens, seems to lack a critical function that’s programmed into deep learning algorithms. On the surface, the algorithms were violating basic biological facts already proven by neuroscientists.
But what if, superficial differences aside, deep learning and the brain are actually compatible?
Now, in a new study published in eLife, Richards, working with DeepMind, proposed a new algorithm based on the biological structure of neurons in the neocortex. Also known as the cortex, this outermost region of the brain is home to higher cognitive functions such as reasoning, prediction, and flexible thought.
The team networked their artificial neurons together into a multi-layered network and challenged it with a classic computer vision task—identifying hand-written numbers.
The new algorithm performed well. But the kicker is that it analyzed the learning examples in a way that’s characteristic of deep learning algorithms, even though it was completely based on the brain’s fundamental biology.
“Deep learning is possible in a biological framework,” concludes the team.
Because the model is only a computer simulation at this point, Richards hopes to pass the baton to experimental neuroscientists, who could actively test whether the algorithm operates in an actual brain.
If so, the data could then be passed back to computer scientists to work out the next generation of massively parallel and low-energy algorithms to power our machines.
It’s a first step towards merging the two fields back into a “virtuous circle” of discovery and innovation.
The blame game
While you’ve probably heard of deep learning’s recent wins against humans in the game of Go, you might not know the nitty-gritty behind the algorithm’s operations.
In a nutshell, deep learning relies on an artificial neural network with virtual “neurons.” Like a towering skyscraper, the network is structured into hierarchies: lower-level neurons process aspects of an input—for example, a horizontal or vertical stroke that eventually forms the number four—whereas higher-level neurons extract more abstract aspects of the number four.
To teach the network, you give it examples of what you’re looking for. The signal propagates forward in the network (like climbing up a building), where each neuron works to fish out something fundamental about the number four.
Like children trying to learn a skill the first time, initially the network doesn’t do so well. It spits out what it thinks a universal number four should look like—think a Picasso-esque rendition.
But here’s where the learning occurs: the algorithm compares the output with the ideal output, and computes the difference between the two (dubbed “error”). This error is then “backpropagated” throughout the entire network, telling each neuron: hey, this is how far off you were, so try adjusting your computation closer to the ideal.
Millions of examples and tweakings later, the network inches closer to the desired output and becomes highly proficient at the trained task.
This error signal is crucial for learning. Without efficient “backprop,” the network doesn’t know which of its neurons are off kilter. By assigning blame, the AI can better itself.
The brain does this too. How? We have no clue.
Biological No-Go
What’s clear, though, is that the deep learning solution doesn’t work.
Backprop is a pretty needy function. It requires a very specific infrastructure for it to work as expected.
For one, each neuron in the network has to receive the error feedback. But in the brain, neurons are only connected to a few downstream partners (if that). For backprop to work in the brain, early-level neurons need to be able to receive information from billions of connections in their downstream circuits—a biological impossibility.
And while certain deep learning algorithms adapt a more local form of backprop— essentially between neurons—it requires their connection forwards and backwards to be symmetric. This hardly ever occurs in the brain’s synapses.
More recent algorithms adapt a slightly different strategy, in that they implement a separate feedback pathway that helps the neurons to figure out errors locally. While it’s more biologically plausible, the brain doesn’t have a separate computational network dedicated to the blame game.
What it does have are neurons with intricate structures, unlike the uniform “balls” that are currently applied in deep learning.
Branching Networks
The team took inspiration from pyramidal cells that populate the human cortex.
“Most of these neurons are shaped like trees, with ‘roots’ deep in the brain and ‘branches’ close to the surface,” says Richards. “What’s interesting is that these roots receive a different set of inputs than the branches that are way up at the top of the tree.”
This is an illustration of a multi-compartment neural network model for deep learning. Left: Reconstruction of pyramidal neurons from mouse primary visual cortex. Right: Illustration of simplified pyramidal neuron models. Image Credit: CIFAR
Curiously, the structure of neurons often turn out be “just right” for efficiently cracking a computational problem. Take the processing of sensations: the bottoms of pyramidal neurons are right smack where they need to be to receive sensory input, whereas the tops are conveniently placed to transmit feedback errors.
Could this intricate structure be evolution’s solution to channeling the error signal?
The team set up a multi-layered neural network based on previous algorithms. But rather than having uniform neurons, they gave those in middle layers—sandwiched between the input and output—compartments, just like real neurons.
When trained with hand-written digits, the algorithm performed much better than a single-layered network, despite lacking a way to perform classical backprop. The cell-like structure itself was sufficient to assign error: the error signals at one end of the neuron are naturally kept separate from input at the other end.
Then, at the right moment, the neuron brings both sources of information together to find the best solution.
There’s some biological evidence for this: neuroscientists have long known that the neuron’s input branches perform local computations, which can be integrated with signals that propagate backwards from the so-called output branch.
However, we don’t yet know if this is the brain’s way of dealing blame—a question that Richards urges neuroscientists to test out.
What’s more, the network parsed the problem in a way eerily similar to traditional deep learning algorithms: it took advantage of its multi-layered structure to extract progressively more abstract “ideas” about each number.
“[This is] the hallmark of deep learning,” the authors explain.
The Deep Learning Brain
Without doubt, there will be more twists and turns to the story as computer scientists incorporate more biological details into AI algorithms.
One aspect that Richards and team are already eyeing is a top-down predictive function, in which signals from higher levels directly influence how lower levels respond to input.
Feedback from upper levels doesn’t just provide error signals; it could also be nudging lower processing neurons towards a “better” activity pattern in real-time, says Richards.
The network doesn’t yet outperform other non-biologically derived (but “brain-inspired”) deep networks. But that’s not the point.
“Deep learning has had a huge impact on AI, but, to date, its impact on neuroscience has been limited,” the authors say.
Now neuroscientists have a lead they could experimentally test: that the structure of neurons underlie nature’s own deep learning algorithm.
“What we might see in the next decade or so is a real virtuous cycle of research between neuroscience and AI, where neuroscience discoveries help us to develop new AI and AI can help us interpret and understand our experimental data in neuroscience,” says Richards.
Image Credit: christitzeimaging.com / Shutterstock.com Continue reading

Posted in Human Robots

#431424 A ‘Google Maps’ for the Mouse Brain ...

Ask any neuroscientist to draw you a neuron, and it’ll probably look something like a star with two tails: one stubby with extensive tree-like branches, the other willowy, lengthy and dotted with spindly spikes.
While a decent abstraction, this cartoonish image hides the uncomfortable truth that scientists still don’t know much about what many neurons actually look like, not to mention the extent of their connections.
But without untangling the jumbled mess of neural wires that zigzag across the brain, scientists are stumped in trying to answer one of the most fundamental mysteries of the brain: how individual neuronal threads carry and assemble information, which forms the basis of our thoughts, memories, consciousness, and self.
What if there was a way to virtually trace and explore the brain’s serpentine fibers, much like the way Google Maps allows us to navigate the concrete tangles of our cities’ highways?
Thanks to an interdisciplinary team at Janelia Research Campus, we’re on our way. Meet MouseLight, the most extensive map of the mouse brain ever attempted. The ongoing project has an ambitious goal: reconstructing thousands—if not more—of the mouse’s 70 million neurons into a 3D map. (You can play with it here!)
With map in hand, neuroscientists around the world can begin to answer how neural circuits are organized in the brain, and how information flows from one neuron to another across brain regions and hemispheres.
The first release, presented Monday at the Society for Neuroscience Annual Conference in Washington, DC, contains information about the shape and sizes of 300 neurons.
And that’s just the beginning.
“MouseLight’s new dataset is the largest of its kind,” says Dr. Wyatt Korff, director of project teams. “It’s going to change the textbook view of neurons.”

http://mouselight.janelia.org/assets/carousel/ML-Movie.mp4
Brain Atlas
MouseLight is hardly the first rodent brain atlasing project.
The Mouse Brain Connectivity Atlas at the Allen Institute for Brain Science in Seattle tracks neuron activity across small circuits in an effort to trace a mouse’s connectome—a complete atlas of how the firing of one neuron links to the next.
MICrONS (Machine Intelligence from Cortical Networks), the $100 million government-funded “moonshot” hopes to distill brain computation into algorithms for more powerful artificial intelligence. Its first step? Brain mapping.
What makes MouseLight stand out is its scope and level of detail.
MICrONS, for example, is focused on dissecting a cubic millimeter of the mouse visual processing center. In contrast, MouseLight involves tracing individual neurons across the entire brain.
And while connectomics outlines the major connections between brain regions, the birds-eye view entirely misses the intricacies of each individual neuron. This is where MouseLight steps in.
Slice and Dice
With a width only a fraction of a human hair, neuron projections are hard to capture in their native state. Tug or squeeze the brain too hard, and the long, delicate branches distort or even shred into bits.
In fact, previous attempts at trying to reconstruct neurons at this level of detail topped out at just a dozen, stymied by technological hiccups and sky-high costs.
A few years ago, the MouseLight team set out to automate the entire process, with a few time-saving tweaks. Here’s how it works.
After injecting a mouse with a virus that causes a handful of neurons to produce a green-glowing protein, the team treated the brain with a sugar alcohol solution. This step “clears” the brain, transforming the beige-colored organ to translucent, making it easier for light to penetrate and boosting the signal-to-background noise ratio. The brain is then glued onto a small pedestal and ready for imaging.
Building upon an established method called “two-photon microscopy,” the team then tweaked several parameters to reduce imaging time from days (or weeks) down to a fraction of that. Endearingly known as “2P” by the experts, this type of laser microscope zaps the tissue with just enough photos to light up a single plane without damaging the tissue—sharper plane, better focus, crisper image.
After taking an image, the setup activates its vibrating razor and shaves off the imaged section of the brain—a waspy slice about 200 micrometers thick. The process is repeated until the whole brain is imaged.
This setup increased imaging speed by 16 to 48 times faster than conventional microscopy, writes team leader Dr. Jayaram Chandrashekar, who published a version of the method early last year in eLife.
The resulting images strikingly highlight every crook and cranny of a neuronal branch, popping out against a pitch-black background. But pretty pictures come at a hefty data cost: each image takes up a whopping 20 terabytes of data—roughly the storage space of 4,000 DVDs, or 10,000 hours of movies.
Stitching individual images back into 3D is an image-processing nightmare. The MouseLight team used a combination of computational power and human prowess to complete this final step.
The reconstructed images are handed off to a mighty team of seven trained neuron trackers. With the help of tracing algorithms developed in-house and a keen eye, each member can track roughly a neuron a day—significantly less time than the week or so previously needed.
A Numbers Game
Even with just 300 fully reconstructed neurons, MouseLight has already revealed new secrets of the brain.
While it’s widely accepted that axons, the neurons’ outgoing projection, can span the entire length of the brain, these extra-long connections were considered relatively rare. (In fact, one previously discovered “giant neuron” was thought to link to consciousness because of its expansive connections).
Images captured from two-photon microscopy show an axon and dendrites protruding from a neuron’s cell body (sphere in center). Image Credit: Janelia Research Center, MouseLight project team
MouseLight blows that theory out of the water.
The data clearly shows that “giant neurons” are far more common than previously thought. For example, four neurons normally associated with taste had wiry branches that stretched all the way into brain areas that control movement and process touch.
“We knew that different regions of the brain talked to each other, but seeing it in 3D is different,” says Dr. Eve Marder at Brandeis University.
“The results are so stunning because they give you a really clear view of how the whole brain is connected.”
With a tested and true system in place, the team is now aiming to add 700 neurons to their collection within a year.
But appearance is only part of the story.
We can’t tell everything about a person simply by how they look. Neurons are the same: scientists can only infer so much about a neuron’s function by looking at their shape and positions. The team also hopes to profile the gene expression patterns of each neuron, which could provide more hints to their roles in the brain.
MouseLight essentially dissects the neural infrastructure that allows information traffic to flow through the brain. These anatomical highways are just the foundation. Just like Google Maps, roads form only the critical first layer of the map. Street view, traffic information and other add-ons come later for a complete look at cities in flux.
The same will happen for understanding our ever-changing brain.
Image Credit: Janelia Research Campus, MouseLight project team Continue reading

Posted in Human Robots

#431385 Here’s How to Get to Conscious ...

“We cannot be conscious of what we are not conscious of.” – Julian Jaynes, The Origin of Consciousness in the Breakdown of the Bicameral Mind
Unlike the director leads you to believe, the protagonist of Ex Machina, Andrew Garland’s 2015 masterpiece, isn’t Caleb, a young programmer tasked with evaluating machine consciousness. Rather, it’s his target Ava, a breathtaking humanoid AI with a seemingly child-like naïveté and an enigmatic mind.
Like most cerebral movies, Ex Machina leaves the conclusion up to the viewer: was Ava actually conscious? In doing so, it also cleverly avoids a thorny question that has challenged most AI-centric movies to date: what is consciousness, and can machines have it?
Hollywood producers aren’t the only people stumped. As machine intelligence barrels forward at breakneck speed—not only exceeding human performance on games such as DOTA and Go, but doing so without the need for human expertise—the question has once more entered the scientific mainstream.
Are machines on the verge of consciousness?
This week, in a review published in the prestigious journal Science, cognitive scientists Drs. Stanislas Dehaene, Hakwan Lau and Sid Kouider of the Collège de France, University of California, Los Angeles and PSL Research University, respectively, argue: not yet, but there is a clear path forward.
The reason? Consciousness is “resolutely computational,” the authors say, in that it results from specific types of information processing, made possible by the hardware of the brain.
There is no magic juice, no extra spark—in fact, an experiential component (“what is it like to be conscious?”) isn’t even necessary to implement consciousness.
If consciousness results purely from the computations within our three-pound organ, then endowing machines with a similar quality is just a matter of translating biology to code.
Much like the way current powerful machine learning techniques heavily borrow from neurobiology, the authors write, we may be able to achieve artificial consciousness by studying the structures in our own brains that generate consciousness and implementing those insights as computer algorithms.
From Brain to Bot
Without doubt, the field of AI has greatly benefited from insights into our own minds, both in form and function.
For example, deep neural networks, the architecture of algorithms that underlie AlphaGo’s breathtaking sweep against its human competitors, are loosely based on the multi-layered biological neural networks that our brain cells self-organize into.
Reinforcement learning, a type of “training” that teaches AIs to learn from millions of examples, has roots in a centuries-old technique familiar to anyone with a dog: if it moves toward the right response (or result), give a reward; otherwise ask it to try again.
In this sense, translating the architecture of human consciousness to machines seems like a no-brainer towards artificial consciousness. There’s just one big problem.
“Nobody in AI is working on building conscious machines because we just have nothing to go on. We just don’t have a clue about what to do,” said Dr. Stuart Russell, the author of Artificial Intelligence: A Modern Approach in a 2015 interview with Science.
Multilayered consciousness
The hard part, long before we can consider coding machine consciousness, is figuring out what consciousness actually is.
To Dehaene and colleagues, consciousness is a multilayered construct with two “dimensions:” C1, the information readily in mind, and C2, the ability to obtain and monitor information about oneself. Both are essential to consciousness, but one can exist without the other.
Say you’re driving a car and the low fuel light comes on. Here, the perception of the fuel-tank light is C1—a mental representation that we can play with: we notice it, act upon it (refill the gas tank) and recall and speak about it at a later date (“I ran out of gas in the boonies!”).
“The first meaning we want to separate (from consciousness) is the notion of global availability,” explains Dehaene in an interview with Science. When you’re conscious of a word, your whole brain is aware of it, in a sense that you can use the information across modalities, he adds.
But C1 is not just a “mental sketchpad.” It represents an entire architecture that allows the brain to draw multiple modalities of information from our senses or from memories of related events, for example.
Unlike subconscious processing, which often relies on specific “modules” competent at a defined set of tasks, C1 is a global workspace that allows the brain to integrate information, decide on an action, and follow through until the end.
Like The Hunger Games, what we call “conscious” is whatever representation, at one point in time, wins the competition to access this mental workspace. The winners are shared among different brain computation circuits and are kept in the spotlight for the duration of decision-making to guide behavior.
Because of these features, C1 consciousness is highly stable and global—all related brain circuits are triggered, the authors explain.
For a complex machine such as an intelligent car, C1 is a first step towards addressing an impending problem, such as a low fuel light. In this example, the light itself is a type of subconscious signal: when it flashes, all of the other processes in the machine remain uninformed, and the car—even if equipped with state-of-the-art visual processing networks—passes by gas stations without hesitation.
With C1 in place, the fuel tank would alert the car computer (allowing the light to enter the car’s “conscious mind”), which in turn checks the built-in GPS to search for the next gas station.
“We think in a machine this would translate into a system that takes information out of whatever processing module it’s encapsulated in, and make it available to any of the other processing modules so they can use the information,” says Dehaene. “It’s a first sense of consciousness.”
Meta-cognition
In a way, C1 reflects the mind’s capacity to access outside information. C2 goes introspective.
The authors define the second facet of consciousness, C2, as “meta-cognition:” reflecting on whether you know or perceive something, or whether you just made an error (“I think I may have filled my tank at the last gas station, but I forgot to keep a receipt to make sure”). This dimension reflects the link between consciousness and sense of self.
C2 is the level of consciousness that allows you to feel more or less confident about a decision when making a choice. In computational terms, it’s an algorithm that spews out the probability that a decision (or computation) is correct, even if it’s often experienced as a “gut feeling.”
C2 also has its claws in memory and curiosity. These self-monitoring algorithms allow us to know what we know or don’t know—so-called “meta-memory,” responsible for that feeling of having something at the tip of your tongue. Monitoring what we know (or don’t know) is particularly important for children, says Dehaene.
“Young children absolutely need to monitor what they know in order to…inquire and become curious and learn more,” he explains.
The two aspects of consciousness synergize to our benefit: C1 pulls relevant information into our mental workspace (while discarding other “probable” ideas or solutions), while C2 helps with long-term reflection on whether the conscious thought led to a helpful response.
Going back to the low fuel light example, C1 allows the car to solve the problem in the moment—these algorithms globalize the information, so that the car becomes aware of the problem.
But to solve the problem, the car would need a “catalog of its cognitive abilities”—a self-awareness of what resources it has readily available, for example, a GPS map of gas stations.
“A car with this sort of self-knowledge is what we call having C2,” says Dehaene. Because the signal is globally available and because it’s being monitored in a way that the machine is looking at itself, the car would care about the low gas light and behave like humans do—lower fuel consumption and find a gas station.
“Most present-day machine learning systems are devoid of any self-monitoring,” the authors note.
But their theory seems to be on the right track. The few examples whereby a self-monitoring system was implemented—either within the structure of the algorithm or as a separate network—the AI has generated “internal models that are meta-cognitive in nature, making it possible for an agent to develop a (limited, implicit, practical) understanding of itself.”
Towards conscious machines
Would a machine endowed with C1 and C2 behave as if it were conscious? Very likely: a smartcar would “know” that it’s seeing something, express confidence in it, report it to others, and find the best solutions for problems. If its self-monitoring mechanisms break down, it may also suffer “hallucinations” or even experience visual illusions similar to humans.
Thanks to C1 it would be able to use the information it has and use it flexibly, and because of C2 it would know the limit of what it knows, says Dehaene. “I think (the machine) would be conscious,” and not just merely appearing so to humans.
If you’re left with a feeling that consciousness is far more than global information sharing and self-monitoring, you’re not alone.
“Such a purely functional definition of consciousness may leave some readers unsatisfied,” the authors acknowledge.
“But we’re trying to take a radical stance, maybe simplifying the problem. Consciousness is a functional property, and when we keep adding functions to machines, at some point these properties will characterize what we mean by consciousness,” Dehaene concludes.
Image Credit: agsandrew / Shutterstock.com Continue reading

Posted in Human Robots