Google’s AI Bots Invent Ridiculous New Legs to Scamper Through Obstacle Courses

https://gizmodo.com/google-s-ai-bots-invent-ridiculous-new-legs-to-scamper-1829693108


Left: An unmodified walker learning how to navigate challenging terrain. Right: A self-modified walker, using new limbs and walking strategy, working through the same obstacle course.
GIF: David Ha/Google Brain/Gizmodo

Using a technique called reinforcement learning, a researcher at Google Brain has shown that virtual robots can redesign their body parts to help them navigate challenging obstacle courses—even if the solutions they come up with are completely bizarre.

Embodied cognition is the idea that an animal’s cognitive abilities are influenced and constrained by its body plan. This means a squirrel’s thought processes and problem-solving strategies will differ somewhat from the cogitations of octopuses, elephants, and seagulls. Each animal has to navigate its world in its own special way using the body it’s been given, which naturally leads to different ways of thinking and learning.

“Evolution plays a vital role in shaping an organism’s body to adapt to its environment,” David Ha, a computer scientist and AI expert at Google Brain, explained in his new study. “The brain and its ability to learn is only one of many body components that is co-evolved together.”

This phenomenon has been observed in the real world, but Ha wanted to know if similar processes might also apply to the digital realm. To that end, Ha conducted a series of experiments to see if reinforcement learning could coax virtual robots, called walkers, into designing their body plans to better accommodate their environment and the challenges confronting them. Reinforcement learning is a tool used in artificial intelligence to steer agents toward a desired goal or direction, by awarding them points for “good” behavior.

Using the OpenAI Gym framework, Ha was able to provide an environment for his walkers. This framework looks a lot like an old-school, 2D video game, but it uses sophisticated virtual physics to simulate natural conditions, and it’s capable of randomly generating terrain and other in-game elements.

As for the walker, it was endowed with a pair of legs, each consisting of an upper and lower section. The bipedal bot had to learn how to navigate through its virtual environment and improve its performance over time. Researchers at DeepMind conducted a similar experiment last year, in which virtual bots had to learn how to walk from scratch and navigate through complex parkour courses. The difference here is that Ha’s walkers had the added benefit of being able to redesign their body plan—or at least parts of it. The bots could alter the lengths and widths of their four leg sections to a maximum of 75 percent of the size of the default leg design. The walkers’ pentagon-shaped head could not be altered, serving as cargo. Each walker used a digital version of LIDAR to assess the terrain immediately in front of it, which is why (in the videos) they appear to shoot a thin laser beam at regular intervals.

Using reinforcement-learning algorithms, the bots were given around a day or two to devise their new body parts and come up with effective locomotion strategies, which together formed a walker’s “policy,” in the parlance of AI researchers. The learning process is similar to trial-and-error, except the bots, via reinforcement learning, are rewarded when they come up with good strategies, which then leads them toward even better solutions. This is why reinforcement learning is so powerful—it speeds up the learning process as the bots experiment with various solutions, many of which are unconventional and unpredictable by human standards.

Left: An unmodified walker joyfully skips through easy terrain. Right: With training, a self-modified walker chose to hop instead.
GIF: David Ha/Google Brain/Gizmodo

For the first test (above), Ha placed a walker in a basic environment with no obstacles and gently rolling terrain. Using its default body plan, the bot adopted a rather cheerful-looking skipping locomotion strategy. After the learning stage, however, it modified its legs such that they were thinner and longer. With these modified limbs, the walker used its legs as springs, quickly hopping across the terrain.

The walker chose a strange body plan and an unorthodox locomotion strategy for traversing challenging terrain.
GIF: David Ha/Google Brain/Gizmodo

The introduction of more challenging terrain (above), such as having to walk over obstacles, travel up and down hills, and jump over pits, introduced some radical new policies, namely the invention of an elongated rear “tail” with a dramatically thickened end. Armed with this configuration, the walkers hopped successfully around the obstacle course.

By this point in the experiment, Ha could see that reinforcement learning was clearly working. Allowing a walker “to learn a better version of its body obviously enables it to achieve better performance,” he wrote in the study.

Not content to stop there, Ha played around with the idea of motivating the walkers to adopt some design decisions that weren’t necessarily beneficial to its performance. The reason for this, he said, is that “we may want our agent to learn a design that utilizes the least amount of materials while still achieving satisfactory performance on the task.”

The tiny walker adopted a very familiar gait when faced with easy terrain.
GIF: David Ha/Google Brain/Gizmodo

So for the next test, Ha rewarded an agent for developing legs that were smaller in area (above). With the bot motivated to move efficiently across the terrain, and using the tiniest legs possible (it no longer had to adhere to the 75 percent rule), the walker adopted a rather conventional bipedal style while navigating the easy terrain (it needed just 8 percent of the leg area used in the original design).

The walker struggled to come up with an effective body plan and locomotion style when it was rewarded for inventing small leg sizes.
GIF: David Ha/Google Brain/Gizmodo

But the walker really struggled to come up with a sensible policy when having to navigate the challenging terrain. In the example shown above, which was the best strategy it could muster, the walker used 27 percent of the area of its original design. Reinforcement learning is good, but it’s no guarantee that a bot will come up with something brilliant. In some cases, a good solution simply doesn’t exist.

“By allowing the agent’s body to adapt to its task within some constraints, it can learn policies that are not only better for its task, but also learn them more quickly,” wrote Ha in the paper. His experiment showed that embodied cognition can apply to the virtual realm, and that agents can be motivated to devise body structures more suitable for a given task.

More practically, this application of reinforcement learning could be used for machine learning-assisted design, in which computers are tasked with designing aerodynamic shapes, testing materials under stressful conditions, or building super-agile robots (the corporeal kind). It could also help with computer graphics and improved video gameplay—imagine having to face off against an AI-enabled adversary that can continually redesign itself as it learns from its mistakes and your strengths.

Best of all, reinforcement learning requires minimal human intervention. Sure, many of the solutions conceived by these virtual bots are weird and even absurd, but that’s kind of the point. As the abilities of these self-learning systems increase in power and scope, they’ll come up with things humans never would have thought of. Which is actually kind of scary.

[Google Brain via New Scientist]

via Gizmodo https://gizmodo.com

October 11, 2018 at 04:48PM

“Like A Hot Knife Through Butter”: SVANKi, A Wireless Heated Ice Cream Scoop

https://geekologie.com/2018/10/like-a-hot-knife-through-butter-svanki-a.php


This is the somehow already funded Kickstarter campaign for SVANKi, a wireless heated ice cream scoop ($59 early bird special, $75 after *spit-takes rainbow sprinkles*). The wireless scoop charges on a wired base, and is good for 30 minutes of 140°F -150°F (60°C – 65°C) use on a charge. It comes in three finishes: white, black, and fake bamboo. Don’t have the $60+ to blow on another kitchen gadget? I’ve got you: just leave the ice cream on the counter for five to ten minutes (or run your scoop under scalding water for a few seconds). That’s what I do, and I’m a firm believer that the simplest solution is often the best one. “You tried electrifying your entire apartment floor to kill cockroaches.” Come over, we’ll play the floor is lava.

Keep going for the Kickstarter video while I invented a super-chilled fork that prevents pizza bites from melting the roof of your mouth.

Thanks to Coleen, who agrees astronauts don’t have this problem with their ice cream.

blog comments powered by Disqus

via Geekologie – Gadgets, Gizmos, and Awesome https://geekologie.com/

October 11, 2018 at 02:19PM

So, This Is What The End Looks Like: Boston Dynamics’ Latest Shockingly Agile Atlas Humanoid Robot

https://geekologie.com/2018/10/so-this-is-what-the-end-looks-like-bosto.php


This is a short video from Boston Dynamics of the latest iteration of their ATLAS humanoid robot doing a little untethered parkour. First it hops over a log to warm up, then jumps up three steps with shocking grace and agility. So, if you thought owning a two-story house was going to save you from the robots, it’s time you burn the paper placement/kid’s menu you drew your robot apocalypse survival plan on and start taking the threat seriously. We need to build a bunkbed. “You mean bunker?” That’s an even better idea. I mean yes of course that’s what I meant. *discreetly ripping up bar napkin*

Keep going for the full video.

Thanks to cjcjcjcj, who agrees the robots are evolving much, much faster than we ever will.

blog comments powered by Disqus

via Geekologie – Gadgets, Gizmos, and Awesome https://geekologie.com/

October 11, 2018 at 04:41PM

Richard Branson Claims Virgin Galactic Will Be in Space “Within Weeks”

http://blogs.discovermagazine.com/d-brief/?p=27759

According to Sir Richard Branson, founder of the space tourism company Virgin Galactic, their latest spaceplane, SpaceShipTwo, will be in space in a matter of weeks.
“We should be in space within weeks, not months. And then we will be in space with myself in months and not years,” Branson told CNBC on Tuesday in Singapore. “We will be in space with people not too long after that, so we have got a very, very exciting couple of months ahead,” he added. Branson has not yet officially expande

via Discover Main Feed https://ift.tt/1dqgCKa

October 10, 2018 at 07:59PM

Amazon patents Alexa tech to tell if you’re sick, depressed and sell you meds

https://arstechnica.com/?p=1392695


Amazon's Echo smart speaker with its blue light ring illuminated.

An Amazon Echo.

Amazon has patented technology that could let Alexa analyze your voice to determine whether you are sick or depressed and sell you products based on your physical or emotional condition.

The patent, titled “Voice-based determination of physical and emotional characteristics of users,” was issued on Tuesday this week; Amazon filed the patent application in March 2017.

The patent describes a voice assistant that can detect “abnormal” physical or emotional conditions. ”For example, physical conditions such as sore throats and coughs may be determined based at least in part on a voice input from the user, and emotional conditions such as an excited emotional state or a sad emotional state may be determined based at least in part on voice input from a user,” the patent says. ”A cough or sniffle, or crying, may indicate that the user has a specific physical or emotional abnormality.”

It’s not clear what ads would be sent based on a user’s emotional state, but someone who is sick might be asked if they want to buy cold medicine.

“A current physical and/or emotional condition of the user may facilitate the ability to provide highly targeted audio content, such as audio advertisements or promotions, to the user,” the patent said.

If the Amazon voice assistant determines that you have a sore threat, the system would “communicate with the audio content server(s)” to select the appropriate ad. “For example, certain content, such as content related to cough drops or flu medicine, may be targeted towards users who have sore throats,” the patent says.

Alexa might then ask, “would you like to order cough drops with 1 hour delivery?” After the order is made, the voice assistant “may append a message to the audible confirmation, such as well wishes, or ‘feel better!'”

System could raise privacy concerns

Companies get patents all the time for technologies that never make it to market, so there is no guarantee this capability will be implemented in future versions of Alexa.

Amazon would have to consider the privacy implications of letting its voice assistant analyze the emotional and physical states of Amazon customers. Amazon and other tech companies last month were called to a Senate Commerce Committee hearing to testify about consumer data privacy, and senators are considering whether to write a new privacy law.

Besides analyzing your physical or emotional states, Amazon’s patent says the system would take into account the user’s browsing history and purchase history:

Embodiments of the disclosure may use physical and/or emotional characteristics of a user in combination with behavioral targeting criteria (e.g., browse history, number of clicks, purchase history, etc.) and/or contextual targeting criteria (e.g., keywords, page types, placement metadata, etc.) to determine and/or select content that may be relevant for presentation to a user.

The system would use a “voice processing algorithm” to determine a user’s emotional state. The voice analysis would be able to detect “happiness, joy, anger, sorrow, sadness, fear, disgust, boredom, stress, or other emotional states.” It would make those determinations “based at least in part on an analysis of pitch, pulse, voicing, jittering, and/or harmonicity of a user’s voice, as determined from processing of the voice data.”

The system would apply tags to each physical or emotional characteristic. Those tags may be “associated with or linked to a data file of the voice input,” and “used to determine content for presentation to the user.”

The emotion-detecting system would be tailored to each user, determining the user’s “default or normal/baseline state” so that it can detect changes that indicate that “the emotional state of the user is abnormal,” the patent says.

Amazon’s analysis would presumably be more accurate when tailored to a specific user, but the patent says the technology can also determine the emotional state of “any user” regardless of whether they normally use that device.

via Ars Technica https://arstechnica.com

October 11, 2018 at 12:18PM

A Soyuz crew makes an emergency landing after rocket fails

https://arstechnica.com/?p=1392427


Article intro image
Enlarge /

A Soyuz rocket launches with Expedition 57 Flight Engineer Nick Hague of NASA and Flight Engineer Alexey Ovchinin of Roscosmos, Thursday, October 11, 2018.

NASA

On Thursday in Kazakhstan, at 4:40am EDT, a Soyuz rocket took off carrying NASA astronaut Nick Hague and Russian cosmonaut Aleksey Ovchinin toward the International Space Station. The ascent proceeded normally until the separation of one of the rocket’s booster stages, by which point the crew had already experienced microgravity.

Because the Soyuz spacecraft did not reach orbit at the point of this booster failure, the crew was forced to make a rapid ballistic descent likely under high g-forces. After about 20 minutes of uncertainty, Russian officials confirmed the crew were ok and had landed about 20km east of Dzhezkazgan, a city in central Kazakhstan. As rescue crews arrived, Hague and Ovchinin were reported in “good condition” and found out of the capsule.

via Ars Technica https://arstechnica.com

October 11, 2018 at 05:32AM

Xbox One October Update Rolling Out With Alexa Functionality, New Avatars

https://www.gamespot.com/articles/xbox-one-october-update-rolling-out-with-alexa-fun/1100-6462458/


The Xbox One October update is rolling out today, offering a few new bells and whistles for the Microsoft console. The features in this update allow you to use more voice interactions, more easily watch HDR movies with streaming services, and personalize your Avatar.

The Xbox Skill feature lets you interact with your Xbox using Alexa or Cortana devices, with a greater degree of precision than the previous Kinect voice commands. As an example, Microsoft says you can tell Alexa to “launch Forza Horizon 4” and in response the Xbox One will start up, sign you in, and launch the game, all without additional commands needed. Other commands include starting Mixer broadcasts, capturing screenshots, and adjusting the volume. You’ll need to download the Xbox Skill in the respective stores to get started.

To promote the new functionality, Amazon will be offering a bundle that includes an Amazon Echo Dot with new Xbox One S or Xbox One X purchases. That deal is coming soon, but supplies will be limited.

The newly revised Avatars are available for all users today as well. These have a focus on inclusiveness, with a greater range of body types, more accessories, and the ability to dress in any kind of clothing you’d like. You’ll see the new Avatars on your own profile, in the activity feed and Gamerscore leaderboard, and more. You can use a Photobooth to take a picture of your Avatar to use as your profile image. Plus, if you want to stick with your original Avatar, that option is available too.

The update also adds Dolby Vision HDR streaming across supported televisions. For now it supports Netflix only, but Dolby Vision will expand to more apps in the coming months. You can enable it through the Settings menu. Finally the Narrator has added Spanish, Portuguese, Polish, Swedish, Dutch, and Australian English.

Got a news tip or want to contact us directly? Email news@gamespot.com

via GameSpot’s PC Reviews https://ift.tt/2mVXxXH

October 11, 2018 at 01:49PM