Elements S4 • E82

Why This Virtual Human Is Being Injured by Scientists





The interactive transcript could not be loaded.


Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Aug 22, 2019

Researchers in South Korea trained a neural network to control a simulated human body, which could shape the future of physical therapy, surgery, and robotics.
» Go to https://NordVPN.com/seeker to get 75% off a 3 year plan and use code SEEKER for an extra month for free!

» Subscribe to Seeker! http://bit.ly/subscribeseeker
» Watch more Elements! http://bit.ly/ElementsPlaylist

The human body is complex; it’s got 206 bones, more than 600 muscles, and learning to control it all takes years of practice.

And so when scientists decided to train a neural network to control a simulated futuristic human body, they simplified the body a bit. Researchers figured they didn’t need all 600+ muscles, so they narrowed it down to 346 that contribute to how our joints move.

After building out the simplified digital human body, researchers started training it, teaching an algorithm to control the skeleton through various tasks ranging from walking to cartwheeling. And the AI learned fast, quickly understanding how to coordinate the muscles.

Once the artificial intelligence had the movements down, researchers started to change the parameters to see how it would respond: making the weights a little heavier, having it jump a little higher, and pelting it with simulated balls that the digital human shook off until finally toppling over.

Researchers tweaked muscles to simulate ailments and simulated surgeries to see how the AI would react.

Whatever the researchers threw at it, the AI adapted.

Learn more about why and how this work could shape the future of physical therapy, surgery, robotics, and overall, the future of mankind in this episode of Elements.

#Robotics #ArtificialIntelligence #HumanBody #Therapy #Seeker #Science #Elements

What Happens When Bone Regeneration Goes Wrong - https://youtu.be/lApqklFkv9c

Read More:
Scalable Muscle-Actuated Human Simulation and Control
"This work aims to build a comprehensive musculoskeletal model
and its control system that reproduces realistic human movements
driven by muscle contraction dynamics."

What is the strongest muscle in the human body?
"Most sources state that there are over 650 named skeletal muscles in the human body, although some figures go up to as many as 840. The dissension comes from those that count the muscles within a complex muscle."

Scalable Muscle-actuated Human Simulation and Control
"The key technical contribution is a scalable, two-level imitation learning algorithm that can deal with a comprehensive full-body musculoskeletal model with 346 muscles. We demonstrate the predictive simulation of dynamic motor skills under anatomical conditions including bone deformity, muscle weakness, contracture, and the use of a prosthesis."


Elements is more than just a science show. It’s your science-loving best friend, tasked with keeping you updated and interested on all the compelling, innovative and groundbreaking science happening all around us. Join our passionate hosts as they help break down and present fascinating science, from quarks to quantum theory and beyond.

Seeker explains every aspect of our world through a lens of science, inspiring a new generation of curious minds who want to know how today’s discoveries in science, math, engineering and technology are impacting our lives, and shaping our future. Our stories parse meaning from the noise in a world of rapidly changing information.

Visit the Seeker website https://www.seeker.com/videos

Elements on Facebook https://www.facebook.com/SeekerElements/

Subscribe now! http://www.youtube.com/subscription_c...

Seeker on Twitter http://twitter.com/seeker

Seeker on Facebook https://www.facebook.com/SeekerMedia/

Seeker http://www.seeker.com/


When autoplay is enabled, a suggested video will automatically play next.

Up next

to add this to Watch Later

Add to

Loading playlists...