Domo arigato, mr. roboto

Discussions on everything related to the software, electronic, and mechanical components of information systems and instruments.

Domo arigato, mr. roboto

Postby Braininvat on May 22nd, 2015, 12:41 pm 

http://www.nytimes.com/2015/05/22/scien ... erity.html

It's been interesting to watch the shift from topheavy software that's very task-specific, towards the neural net architecture with simple general learning algorithms.
User avatar
Braininvat
Resident Member
 
Posts: 5828
Joined: 21 Jan 2014
Location: Black Hills
MarshallSciameriKen liked this post


Re: Domo arigato, mr. roboto

Postby Marshall on May 22nd, 2015, 1:50 pm 

Here's a paper by that team of people, referred to in the NYT article:
http://arxiv.org/abs/1504.00702
End-to-End Training of Deep Visuomotor Policies
Sergey Levine, Chelsea Finn, Trevor Darrell, Pieter Abbeel
(Submitted on 2 Apr 2015)
Policy search methods based on reinforcement learning and optimal control can allow robots to automatically learn a wide range of tasks. However, practical applications of policy search tend to require the policy to be supported by hand-engineered components for perception, state estimation, and low-level control. We propose a method for learning policies that map raw, low-level observations, consisting of joint angles and camera images, directly to the torques at the robot's joints. The policies are represented as deep convolutional neural networks (CNNs) with 92,000 parameters. The high dimensionality of such policies poses a tremendous challenge for policy search. To address this challenge, we develop a sensorimotor guided policy search method that can handle high-dimensional policies and partially observed tasks. We use BADMM to decompose policy search into an optimal control phase and supervised learning phase, allowing CNN policies to be trained with standard supervised learning techniques. This method can learn a number of manipulation tasks that require close coordination between vision and control, including inserting a block into a shape sorting cube, screwing on a bottle cap, fitting the claw of a toy hammer under a nail with various grasps, and placing a coat hanger on a clothes rack.
Subjects: Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
Report number: BVLC Report Series #100
User avatar
Marshall
Honored Member
 
Posts: 7916
Joined: 17 Oct 2006


Re: Domo arigato, mr. roboto

Postby Marshall on May 22nd, 2015, 2:27 pm 

Another paper the NYT referred to was a letter in Nature (26 Feb 2015)called
Human-level control through deep reinforcement learning
by V. Mnih et al

I couldn't find a copy of that free online, but here is a 2013 article by the same authors:
http://arxiv.org/abs/1312.5602
Playing Atari with Deep Reinforcement Learning
Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller
(Submitted on 19 Dec 2013)
We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them.
Comments: NIPS Deep Learning Workshop 2013
================

The 2015 article can be read in a restricted access context at Nature. You can't print it out. Or I couldn't see how. This Volodymyr Mnih group seems to be what became the UK company "Deep Mind" which Google purchased for an undisclosed amount.
User avatar
Marshall
Honored Member
 
Posts: 7916
Joined: 17 Oct 2006


Re: Domo arigato, mr. roboto

Postby Braininvat on June 18th, 2015, 12:54 pm 

Here is one of the (to me) weirder results of a robotics industry in its infancy....

http://www.nytimes.com/video/technology ... y-dog.html

Concerns Sony's decision to end tech support and repair parts production for its robotic canines. The attachment people developed to these little robots is sweet and rather sad.
User avatar
Braininvat
Resident Member
 
Posts: 5828
Joined: 21 Jan 2014
Location: Black Hills


Re: Domo arigato, mr. roboto

Postby Braininvat on March 15th, 2017, 8:46 pm 

Penrose Institute to study AI/consciousness questions.....how humans and computers handle a chess problem differently, for one.

https://www.chess.com/news/view/will-this-position-help-to-understand-human-consciousness-4298
User avatar
Braininvat
Resident Member
 
Posts: 5828
Joined: 21 Jan 2014
Location: Black Hills


Re: Domo arigato, mr. roboto

Postby Dave_Oblad on March 16th, 2017, 1:22 am 

Hi Biv,

Have you been watching the UK program called "Humans"? It's very good and is exploring many angles of Human vs Robot interactions. The Robots are becoming self-aware or conscious.. due to a bit of code that has been hacked into their Processing Hardware.

The latest twist (here in the USA, we are getting older pre-aired programs today) is the secret manufacturing of Children Robots, presumably for couples that can't have real children.. like the Robo-Pets mentioned above.

Conscious Robots are battling an uphill battle with Humans for basic rights and autonomy from Human controls.

It's a very good program with lot's of neat twists.

Regards,
Dave :^)
User avatar
Dave_Oblad
Resident Member
 
Posts: 3213
Joined: 08 Sep 2010
Location: Tucson, Arizona
Blog: View Blog (2)


Re: Domo arigato, mr. roboto

Postby Braininvat on March 16th, 2017, 9:59 am 

Hi, Dave. Thanks for reminding me. I plan to see it when there is a dvd. (no cable - just occasional subs to Netflix, and whatever comes in on antenna) Same thing goes for "Westworld."

A lot of quality stuff coming from the UK these days.
User avatar
Braininvat
Resident Member
 
Posts: 5828
Joined: 21 Jan 2014
Location: Black Hills



Return to Computers

Who is online

Users browsing this forum: No registered users and 4 guests