The Feedback Loop

Welcome to The Feedback Loop! Here, we celebrate all things robots and the intricate tech behind them. Come learn about the latest robo-advancements and key concepts about how machines think and learn. Brought to you by Julian Taub

[feedbackloopnews-blog]

Surrogate Humans Could Replace Animals in Testing

The road from the lab to pharmacy is a long and treacherous one for any drug. There are animal tests, government agency approval (FDA in the US), and a series of human trials.

It can take many years and it’s very expensive to push through the process. Furthermore, animal trials only mimic human reactions to toxicity to a certain extent, and are not always cost-effective to maintain, especially for longer studies.

Now, a research collaboration between Los Alamos National Laboratory and Vanderbilt University is building ATHENA, a miniature human organ network model. The model would lie on top of a lab bench and monitor how a drug affects the cell health and metabolism of the “organs” in real time. This style of screening could be an effecient way to test drug toxicity, especially for long term effects.

The researchers already bioengineered a miniature liver, and are now working on a heart, lungs, kidney and synthetic blood. They are also developing mathetical models for the to analyze results. The project comes out of the moment in bioengineering to grow cells in a bioreactor that form structures that work a cohesive whole and sense their environment.

As with all models, it’s only as good as it’s what it can detect. Instead of creating replicas that try to resemble the human organ as close as possible, the researchers have opted to focus on mimicking key metabolic and enzyme functions.    

Hopefully, as this surrogate human evolves, it will narrow down drugs heading to market and even help scientists hone in on side effects and issues with potential drugs.

Photo: Artist's depiction of ATHENA

[feedbackloopnews-blog]

Robot Clam Digs for People

A tasty seafood treat holds the key to robotic anchors. It’s called the razor clam, or American Jackknife clam, and it lives in holes in tidal areas on the coast of the North Atlantic.

Many of the best designs for robots are inspired by the animal kingdom. After all, evolution has time-tested these biological strategies on a scale we can hardly imagine. The razor clam, for example, can dig at 1 centimeter a second and burrow over 1,600 feet while expending very little energy. Controlling its retractable shell and ‘foot’ that sticks out on one side, it can outpace a human with a shovel, fleeing deeper into the earth when threatened by birds or clam hunters.

Below is a video of the razor clam in action:

Amos Winter and Anette Hosoi, mechanical engineering researchers at MIT, studied the clam to figure out how it's able to dig so well. They realized that when the foot starts digging, the shell begins closing to tighten its volume, localizing the area, and reducing drag. Simultaneously, the clam pours water into the soil or sand beneath it, turning it into ‘quicksand.’ The shell has to close at the right rate, in order to get the right consistency of quicksand.

Drawing from their data, the researchers created a mechanical clamshell than expands and contracts like an accordion, attached to a rod that opens and closes the shell and moves it up and down. It’s currently powered by a compressed air system and they are working on an electronic version.

The 'RoboClam' can be used to burrow anchors into the ocean floor, even in rough waters, or to detonate underwater mines. It can also efficiently lay cables in shallow seas, which has previously been accomplished by divers.     

Photo Credit: Donna Coveney

Video Credit: Wild Work Productions on Vimeo

[feedbackloopnews-blog]

Spotting the Fakers

 Could you lie to a face like this?

Your best poker face may be no match for a computer system created at the University of Toronto.

Researchers there tested the system against people to see who was better at telling the difference between real and faked facial expressions (in this case pain). The computer outdid the human participants, determining who was faking 85% of the time. People, on the other hand, did not do better than random chance. Even when people were trained to spot fakes, they improved to a measly 55%.

Basically, we’re terrible at recognizing if someone is trying to deceive us.

The researchers are planning to use the computer to characterize the ‘behavioral fingerprint’ of voluntary and involuntary facial expressions. There are also wide applications for a computer that can tell if you’re lying and how you’re actually feeling (i.e. homeland security, medicine, job screening, civil and criminal court, and psychology).

If the robots ever become our overlords, we’re screwed.

To read the scientific paper, click here: http://www.cell.com/current-biology/abstract/S0960-9822(14)00147-X

[feedbackloopnews-blog]

Bouncing Bots Can Go Where Others Can't

Robots that navigate remote and hard-to-reach areas usually have sensors to detect and avoid obstacles, like trees or windows. However, if it hits one of them, the fragile sensors can break, and the robot is out of commission or missing in action. And let's face it, accidents happen.

Instead of trying to find better ways around objects, researchers at Ecole Polytechnique of Lausanne, Switzerland are designing robots that take obstacles head on, and absorb the impact.

Their two models, the Gimball and the Airburr, both have rotating outer cages that spin independently from the robots inner parts, allowing it to ricochet off of surfaces. Here, a bumpy ride is better than a smooth one - less sensors means a lighter load for the bot.

Gimball, a giant spinning soccerball of a robot, was let loose in the Swiss forest and arrived at given coordinates with ease, even after hitting a few trees and ferns.

Alternatively, AirBurr, which looks a bit like a flying pentagram, has moving legs that let it either land when it hits something or perch on random ledges.

The robots can be used in war zones and for search and rescue missions.

Below are videos of both robots in action: 

     Image Credit: A. Herzog, EPFL

[feedbackloopnews-blog]

Robots Take Over The EU

In the coming days, robots, big and small, will invade Europe. No, I’m not talking about Terminator 5: The Paris Chronicles. I’m talking about EU Robotics Week, the continent-wide festival that celebrates all things robo. From November 26th to December 2nd, a slew of universities, companies, government institutions, and educational programs will come together to showcase, discuss, and educate about the latest robot advancements and robot enthusiasm. (In reality, the first event starts on Nov. 20th and the last one ends on Dec. 7th.) You can attend a workshop on roboethics, get a tour of a robotics lab, compete in a Lego challenge, or even see a concert given by robots!

For a list of events, visit the EU Robotics Week website here.  

Unfortunately, some of you, like me, will be stuck in another hemisphere. Have no fear — The University of Twente, in Holland, has set up a website that will let you peek into any event and see it live-streamed. You can go to this site by clicking on http://www.robotics-labs.eu.  

[feedbackloopnews-blog]

Mind Controlled Robots

Everyone, at some point, wishes at some point that they could control things with their mind. Who wouldn’t want that kind of power? Luckily, this is not complete science fiction: Researchers in a cross-collaboration between France and Japan (the CRNS-AIST Joint Robotics Laboratory) are working on a robot that acts according to your thoughts.

Well, sort of. The researchers use a little trick with flickering lights to know what your thinking. When you look at a light that flashes at a speed of more than 3.5 times a second, your brain neurons create a signal that matches the light’s frequency. These signals are easily measured by electrodes placed on your scalp. In the video below, the four arrows on the computer screen are flashing at different frequencies and many of these frequencies are picked up at the same time by the cyborg-like helmet the subject wears. The machine then weighs the strength of each frequency signal and figures out the location of your gaze. It tries to infer what object you are looking at and what you might be interested in. Objects that the subject sees on the screen can have their own frequency as well, to choose more accurately.

The next trick they use is an AI concept called affordance. This basically means that an object’s properties (size, shape, compactness, etc.) suggest what you can, and probably should, do with it. For example, a baseball fits into the palm of your hand. It would be much easier to throw it than to try to sit on it or wear it as a hat. Measuring how suited an object is for a certain task helps the robot respond accordingly when the system picks up that you are interested in it. In a few cases, the robot already has the instructions of how to deal with a certain object.

This advancement in robotics can help people who are bedridden or cannot fully move about interact with the world in ways that they may have never imagined.

Note: The man standing behind the robot is not influencing its actions. He's protecting the power button and the robot's balance.

[feedbackloopnews-blog]

Dismantling Transhumanity's Sex-Bots (NSFW)

A few days ago, the futurist blog, Transhumanity, posted an article about how sex robots will give us mind-bending orgasms, which will help us live longer. The robots' sexual prowess will supposedly outdo anything “meat-bag” humans can perform. 

According to the article: “They’ll offer us quadruple-tongued cunnilingus, open-throat silky fellatio, deliriously gentle kissing, transcendent nipple tweaking, g-spot massage & prostate milking dexterity, plus 2,000 varieties of coital rhythm with scented lubes — this will all be ours when the Sexbots arrive.”

Their article was prompted by the book, “Love and Sex with Robots,“ written by artificial intelligence expert Dr. David Levy, saying that robots and humans will in the future fall in love and even marry.

Let’s tackle the love aspect first: there are already robots out there that can fall in love with humans. They’re called Lovotics, and were created by Hooman Samani and his team at National Taipei University. This shaggy robot has synthetic versions of hormones and responds to being touched by whirring and changing color. The reward mechanism from the stimuli makes it come back for more — and be very clingy. Here’s a video of the love robot in action:

Emotions are actually very important for the robot’s learning mechanism. Humans work on the sub-optimal level— not everything is perfect, but they still go forward, and improve the situation along the way. When robots run an algorithm, on the other hand, they search for the best way to do something. Emotions help them decide quicker, because it eliminates some of the possible options.

Which brings us to the next point: satisfying robo-sex. I don’t know about you, but for me, someone (or something) that’s a good lay needs to be adaptive, spontaneous, and not to mention have fluidity of motion. At this point, autonomous robots (we are talking about autonomous, right? Since having some guy in a lab remote control the sexbot would be creepy) would take their sweet time figuring out the best way to please you. In a painfully slow and semi-clumsy way. To strengthen this point, here’s a clip of robots playing soccer at the 2009 Robocup:

Look how much time the robots spent shuffling their feet in front of the ball! In a sexual scenario, the robot would act more like a shy middle-schooler than a sexpot, staring at his naked counterpart, trying to find the optimal solution to make the first move and initiate “the sex.” Even five Hitachi motors and a bucket of pheromones wouldn’t do it for me at that point.

Robots also have a really hard time extrapolating and intuiting sensory data. For example, at New York University's Computer Science Department, researchers sent Coureuse, a state-of-the-art navigation robot, to ride through central park and label objects. When it came across light hitting the asphalt, it mistook it for sand. What will a robot think went it sees the wall cast a shadow on your naked body while you’re lying in bed?

Hopefully, the future will provide robots with smoother moves. I’m all for human-robot couples. After all, humans have been finding connections and meaning since the beginning of time. Just one word of caution: if you decide to have sex with a robot — disable its camera first. You really don’t want someone hacking into it and putting that stuff on YouTube.

For more/background info:

Transhumanity's Robo-Sex Article 

Review of "Love and Sex with Robots"

Love and Sex with Robots on Amazon

Lovotics

Robocup Website

Coureuse Robot Homepage

[feedbackloopnews-blog]

Robot Rules the Kitchen

Wouldn’t you love to come home, after a long day’s work, and have your robot cook you dinner? Well, that’s exactly what researchers at the Korean Institute of Science and Tech were thinking when they created CIROS. This bot can move around a kitchen on its own, using items like microwaves, dishwashers, cups, and cutting boards. It can clean up after you, mix your drinks, and recognize voices or keywords.

Admittedly he doesn’t have the quickness of a human, but he is a step forward for household robots, as well as being extremely entertaining. Here is a video of him making a salad at Robot World 2012 — all to an epic mealtime soundtrack!

[feedbackloopnews-blog]

Super Glider Takes on Sandy

Living in downtown New York City, I sat last week in the darkness, listening to Sandy’s howl and rippling winds that wreaked havoc all across the Eastern seaboard. There was, however, a robot who took on the brunt of the storm like the best of New Yorkers (or New Jerseyites) did. Her name was Mercury.

Mercury is a Wave Glider, an autonomous robot from the company Liquid Robotics that uses the energy in waves to navigate the seas. Waiting one hundred meters off the coast of New Jersey, it danced with Hurricane Sandy, keeping her balance and staying intact, even in the presence of 81 mile per hour winds. All the while, Mercury continued to take sensory data of the storm.  

Wave Gliders utilize the motion of water at different depths of the wave to push itself forward. In every wave, water moves in a loop: the top of the wave has the biggest loops and the most motion. As you go deeper into the wave, the loops become smaller and smaller, causing less of a disturbance. The Wave Glider has two parts: the “Float,” which sits on the rough-riding wave surface, and the “Sub,” which stays in the calmer depths. They are separated by a 20-foot umbilical cord. When the float is rocked on the surface, it tugs on the sub. The sub, in turn, uses that motion to pull down on the float and propel it forward. Scale-like “wings” on either side of the Sub rotate with the motion to help the water move past. Here's an image of this mechanism below:

Other parts of the glider include solar panels to power the sensors, a GPS system, computers, an antenna, and a safety device that detaches the cord if animals get tangled in it.

This is not the first hurricane a Wave Glider has been caught in. Last summer, another glider stood its ground during Hurricane Isaac off the coast of Louisiana. It also recorded invaluable weather and pressure data within that storm.

Other gliders were set aside for a longer and more harrowing journey. In November 2011, four gliders started from San Francisco across the Pacific Ocean, dealing with storms that would frighten Jason and the Argonauts. After stopping briefly in Hawaii, they split up into two groups, one heading to Japan, and the other heading to Australia, and both groups should arrive at each destination within the next few months. (You can check out their current positions on Google Maps here.)

Not only can these bots withstand incredible conditions, but they can also pick up data from previously unexplored parts of the ocean’s surface, and even warn us of natural disasters.      

Image credit: Liquid Robotics

[feedbackloopnews-blog]

Before We Turn On…

Welcome to The Feedback Loop!

This is a site about robot news and robot culture, but as its name suggests, it is a lot more. I will periodically introduce AI and machine learning concepts, distilled and dejargonized for the every(man/woman) reader. With robotics on the rise, it’s important to empower ourselves and understand how to navigate the science behind it.

As always, I am open to any feedback you may have, either here or on the JuliansScienceCorner.com mother site.

Cheers,

Julian