Tag Archives: german
#434786 AI Performed Like a Human on a Gestalt ...
Dr. Been Kim wants to rip open the black box of deep learning.
A senior researcher at Google Brain, Kim specializes in a sort of AI psychology. Like cognitive psychologists before her, she develops various ways to probe the alien minds of artificial neural networks (ANNs), digging into their gory details to better understand the models and their responses to inputs.
The more interpretable ANNs are, the reasoning goes, the easier it is to reveal potential flaws in their reasoning. And if we understand when or why our systems choke, we’ll know when not to use them—a foundation for building responsible AI.
There are already several ways to tap into ANN reasoning, but Kim’s inspiration for unraveling the AI black box came from an entirely different field: cognitive psychology. The field aims to discover fundamental rules of how the human mind—essentially also a tantalizing black box—operates, Kim wrote with her colleagues.
In a new paper uploaded to the pre-publication server arXiv, the team described a way to essentially perform a human cognitive test on ANNs. The test probes how we automatically complete gaps in what we see, so that they form entire objects—for example, perceiving a circle from a bunch of loose dots arranged along a clock face. Psychologist dub this the “law of completion,” a highly influential idea that led to explanations of how our minds generalize data into concepts.
Because deep neural networks in machine vision loosely mimic the structure and connections of the visual cortex, the authors naturally asked: do ANNs also exhibit the law of completion? And what does that tell us about how an AI thinks?
Enter the Germans
The law of completion is part of a series of ideas from Gestalt psychology. Back in the 1920s, long before the advent of modern neuroscience, a group of German experimental psychologists asked: in this chaotic, flashy, unpredictable world, how do we piece together input in a way that leads to meaningful perceptions?
The result is a group of principles known together as the Gestalt effect: that the mind self-organizes to form a global whole. In the more famous words of Gestalt psychologist Kurt Koffka, our perception forms a whole that’s “something else than the sum of its parts.” Not greater than; just different.
Although the theory has its critics, subsequent studies in humans and animals suggest that the law of completion happens on both the cognitive and neuroanatomical level.
Take a look at the drawing below. You immediately “see” a shape that’s actually the negative: a triangle or a square (A and B). Or you further perceive a 3D ball (C), or a snake-like squiggle (D). Your mind fills in blank spots, so that the final perception is more than just the black shapes you’re explicitly given.
Image Credit: Wikimedia Commons contributors, the free media repository.
Neuroscientists now think that the effect comes from how our visual system processes information. Arranged in multiple layers and columns, lower-level neurons—those first to wrangle the data—tend to extract simpler features such as lines or angles. In Gestalt speak, they “see” the parts.
Then, layer by layer, perception becomes more abstract, until higher levels of the visual system directly interpret faces or objects—or things that don’t really exist. That is, the “whole” emerges.
The Experiment Setup
Inspired by these classical experiments, Kim and team developed a protocol to test the Gestalt effect on feed-forward ANNs: one simple, the other, dubbed the “Inception V3,” far more complex and widely used in the machine vision community.
The main idea is similar to the triangle drawings above. First, the team generated three datasets: one set shows complete, ordinary triangles. The second—the “Illusory” set, shows triangles with the edges removed but the corners intact. Thanks to the Gestalt effect, to us humans these generally still look like triangles. The third set also only shows incomplete triangle corners. But here, the corners are randomly rotated so that we can no longer imagine a line connecting them—hence, no more triangle.
To generate a dataset large enough to tease out small effects, the authors changed the background color, image rotation, and other aspects of the dataset. In all, they produced nearly 1,000 images to test their ANNs on.
“At a high level, we compare an ANN’s activation similarities between the three sets of stimuli,” the authors explained. The process is two steps: first, train the AI on complete triangles. Second, test them on the datasets. If the response is more similar between the illusory set and the complete triangle—rather than the randomly rotated set—it should suggest a sort of Gestalt closure effect in the network.
Machine Gestalt
Right off the bat, the team got their answer: yes, ANNs do seem to exhibit the law of closure.
When trained on natural images, the networks better classified the illusory set as triangles than those with randomized connection weights or networks trained on white noise.
When the team dug into the “why,” things got more interesting. The ability to complete an image correlated with the network’s ability to generalize.
Humans subconsciously do this constantly: anything with a handle made out of ceramic, regardless of shape, could easily be a mug. ANNs still struggle to grasp common features—clues that immediately tells us “hey, that’s a mug!” But when they do, it sometimes allows the networks to better generalize.
“What we observe here is that a network that is able to generalize exhibits…more of the closure effect [emphasis theirs], hinting that the closure effect reflects something beyond simply learning features,” the team wrote.
What’s more, remarkably similar to the visual cortex, “higher” levels of the ANNs showed more of the closure effect than lower layers, and—perhaps unsurprisingly—the more layers a network had, the more it exhibited the closure effect.
As the networks learned, their ability to map out objects from fragments also improved. When the team messed around with the brightness and contrast of the images, the AI still learned to see the forest from the trees.
“Our findings suggest that neural networks trained with natural images do exhibit closure,” the team concluded.
AI Psychology
That’s not to say that ANNs recapitulate the human brain. As Google’s Deep Dream, an effort to coax AIs into spilling what they’re perceiving, clearly demonstrates, machine vision sees some truly weird stuff.
In contrast, because they’re modeled after the human visual cortex, perhaps it’s not all that surprising that these networks also exhibit higher-level properties inherent to how we process information.
But to Kim and her colleagues, that’s exactly the point.
“The field of psychology has developed useful tools and insights to study human brains– tools that we may be able to borrow to analyze artificial neural networks,” they wrote.
By tweaking these tools to better analyze machine minds, the authors were able to gain insight on how similarly or differently they see the world from us. And that’s the crux: the point isn’t to say that ANNs perceive the world sort of, kind of, maybe similar to humans. It’s to tap into a wealth of cognitive psychology tools, established over decades using human minds, to probe that of ANNs.
“The work here is just one step along a much longer path,” the authors conclude.
“Understanding where humans and neural networks differ will be helpful for research on interpretability by enlightening the fundamental differences between the two interesting species.”
Image Credit: Popova Alena / Shutterstock.com Continue reading
#434569 From Parkour to Surgery, Here Are the ...
The robot revolution may not be here quite yet, but our mechanical cousins have made some serious strides. And now some of the leading experts in the field have provided a rundown of what they see as the 10 most exciting recent developments.
Compiled by the editors of the journal Science Robotics, the list includes some of the most impressive original research and innovative commercial products to make a splash in 2018, as well as a couple from 2017 that really changed the game.
1. Boston Dynamics’ Atlas doing parkour
It seems like barely a few months go by without Boston Dynamics rewriting the book on what a robot can and can’t do. Last year they really outdid themselves when they got their Atlas humanoid robot to do parkour, leaping over logs and jumping between wooden crates.
Atlas’s creators have admitted that the videos we see are cherry-picked from multiple attempts, many of which don’t go so well. But they say they’re meant to be inspirational and aspirational rather than an accurate picture of where robotics is today. And combined with the company’s dog-like Spot robot, they are certainly pushing boundaries.
2. Intuitive Surgical’s da Vinci SP platform
Robotic surgery isn’t new, but the technology is improving rapidly. Market leader Intuitive’s da Vinci surgical robot was first cleared by the FDA in 2000, but since then it’s come a long way, with the company now producing three separate systems.
The latest addition is the da Vinci SP (single port) system, which is able to insert three instruments into the body through a single 2.5cm cannula (tube) bringing a whole new meaning to minimally invasive surgery. The system was granted FDA clearance for urological procedures last year, and the company has now started shipping the new system to customers.
3. Soft robot that navigates through growth
Roboticists have long borrowed principles from the animal kingdom, but a new robot design that mimics the way plant tendrils and fungi mycelium move by growing at the tip has really broken the mold on robot navigation.
The editors point out that this is the perfect example of bio-inspired design; the researchers didn’t simply copy nature, they took a general principle and expanded on it. The tube-like robot unfolds from the front as pneumatic pressure is applied, but unlike a plant, it can grow at the speed of an animal walking and can navigate using visual feedback from a camera.
4. 3D printed liquid crystal elastomers for soft robotics
Soft robotics is one of the fastest-growing sub-disciplines in the field, but powering these devices without rigid motors or pumps is an ongoing challenge. A variety of shape-shifting materials have been proposed as potential artificial muscles, including liquid crystal elastomeric actuators.
Harvard engineers have now demonstrated that these materials can be 3D printed using a special ink that allows the designer to easily program in all kinds of unusual shape-shifting abilities. What’s more, their technique produces actuators capable of lifting significantly more weight than previous approaches.
5. Muscle-mimetic, self-healing, and hydraulically amplified actuators
In another effort to find a way to power soft robots, last year researchers at the University of Colorado Boulder designed a series of super low-cost artificial muscles that can lift 200 times their own weight and even heal themselves.
The devices rely on pouches filled with a liquid that makes them contract with the force and speed of mammalian skeletal muscles when a voltage is applied. The most promising for robotics applications is the so-called Peano-HASEL, which features multiple rectangular pouches connected in series that contract linearly, just like real muscle.
6. Self-assembled nanoscale robot from DNA
While you may think of robots as hulking metallic machines, a substantial number of scientists are working on making nanoscale robots out of DNA. And last year German researchers built the first remote-controlled DNA robotic arm.
They created a length of tightly-bound DNA molecules to act as the arm and attached it to a DNA base plate via a flexible joint. Because DNA carries a charge, they were able to get the arm to swivel around like the hand of a clock by applying a voltage and switch direction by reversing that voltage. The hope is that this arm could eventually be used to build materials piece by piece at the nanoscale.
7. DelFly nimble bioinspired robotic flapper
Robotics doesn’t only borrow from biology—sometimes it gives back to it, too. And a new flapping-winged robot designed by Dutch engineers that mimics the humble fruit fly has done just that, by revealing how the animals that inspired it carry out predator-dodging maneuvers.
The lab has been building flapping robots for years, but this time they ditched the airplane-like tail used to control previous incarnations. Instead, they used insect-inspired adjustments to the motions of its twin pairs of flapping wings to hover, pitch, and roll with the agility of a fruit fly. That has provided a useful platform for investigating insect flight dynamics, as well as more practical applications.
8. Soft exosuit wearable robot
Exoskeletons could prevent workplace injuries, help people walk again, and even boost soldiers’ endurance. Strapping on bulky equipment isn’t ideal, though, so researchers at Harvard are working on a soft exoskeleton that combines specially-designed textiles, sensors, and lightweight actuators.
And last year the team made an important breakthrough by combining their novel exoskeleton with a machine-learning algorithm that automatically tunes the device to the user’s particular walking style. Using physiological data, it is able to adjust when and where the device needs to deliver a boost to the user’s natural movements to improve walking efficiency.
9. Universal Robots (UR) e-Series Cobots
Robots in factories are nothing new. The enormous mechanical arms you see in car factories normally have to be kept in cages to prevent them from accidentally crushing people. In recent years there’s been growing interest in “co-bots,” collaborative robots designed to work side-by-side with their human colleagues and even learn from them.
Earlier this year saw the demise of ReThink robotics, the pioneer of the approach. But the simple single arm devices made by Danish firm Universal Robotics are becoming ubiquitous in workshops and warehouses around the world, accounting for about half of global co-bot sales. Last year they released their latest e-Series, with enhanced safety features and force/torque sensing.
10. Sony’s aibo
After a nearly 20-year hiatus, Sony’s robotic dog aibo is back, and it’s had some serious upgrades. As well as a revamp to its appearance, the new robotic pet takes advantage of advances in AI, with improved environmental and command awareness and the ability to develop a unique character based on interactions with its owner.
The editors note that this new context awareness mark the device out as a significant evolution in social robots, which many hope could aid in childhood learning or provide companionship for the elderly.
Image Credit: DelFly Nimble / CC BY – SA 4.0 Continue reading