Slashdot Mirror
Unboxing Boston Dynamics' DARPA-Ready Atlas Robot
mikejuk writes with some robot eye candy, in the form of this excerpt: "If you think its cool to video the unboxing of your latest mobile phone — think again. Unboxing a robot has a lot more going for it and reaches a whole new level of sci-fi realized. The Atlas robot is a standard humanoid robot to be used by competitors in the DARPA Robotics challenge. Built by Boston Dynamics, it is in the same line as Petman and BigDog. It is now being delivered to the labs that will take part and the temptation to make an unboxing video has been irresistible They arrive in plain of wooden crates as if they were auto parts. Next it is unwrapped and lifted out of its packing case using a crane. It looks black and threatening — just like a sci-fi movie but watch the videos and see."
If the robot is so advanced, why can't it unbox itself?
I do not fail; I succeed at finding out what does not work.
Intense music with powerful electric guitars set to 1 minute of video of a machine wiggling about a bit while strung up to high-power cables and solid steel scaffolding. It is so pretentious and unimpressive that it can only pass as pure comedy. Boston Dynamics is a black hole for funding.
Signature intentionally left blank.
That's a nice piece of machinery. Much good mechanical engineering went into that robot. It took $120 million to get to this point, via BigDog and the LS3. DARPA is really throwing money at this, and it's working.
The DARPA competition in a simulator in August indicates that the perception software is getting reasonably good, but the movement software from the teams still sucks. The best team had 12 falls in simulation. My guess is that the results in December 2013 won't be very impressive, but by round 2, in 2014, the robots will be moving much better.
Teams are provided with a .so file (no source) from Boston Dynamics which allows control of the robot and has some functions for basic walking behavior. But the code Boston Dynamics provides to teams is not the good stuff they use internally.
Boston Dynamics is a black hole for funding.
No, that's what it costs to play in this game. There are theoretical problems to be solved, for which solutions may not be known. There are also many practical problems to be solved in mechanical design, actuation, and electronics. Those will yield to money and routine engineering effort.
That was the lesson of the DARPA Grand Challenge in 2003-2005. Until then, the typical robotics project was a professor and three to five grad students, and it took years to make minor advances. DARPA had been putting money into automatic driving since the 1960s without getting anything useful out. Dr. Tony Tether, the DARPA director at the time, decided that academic robotics needed a major kick in the ass. The DARPA Grand Challenge did that.
It wasn't the $1M prize which caused major universities to devote big chunks of their CS departments to that project. It was the threat that if they didn't do well, their DARPA funding would be cut off for failure. Fear worked as a motivator.
A side effect of the 2003-2005 Grand Challenge was that many key components, like integrated INS/GPS combos and LIDAR systems, became smaller, cheaper, and better, now that there was some demand. The original CMU INS/GPS combo took 9U of rack space and required air conditioning. Three years later, you could get that in a box the size of a thick book.
[News that matters...]
Robot gets unboxed.
How about someone unboxing a full-size, fully-functional "Atlas" like *this*?
http://imageshack.us/photo/my-images/807/mech.png/
Would that qualify as "news that matters"?
Besides, you'd be better be nice. Some of the TFA's Atlas's close relatives may be armed and patrolling and manning checkpoints in and around your neighborhood before too long.
"In Soviet US, droids look for YOU!"
You know what's even worse than "Skynet"-style AI-controlled armies of "Terminator"-style cyborgs/robots?
Government/human-controlled armies of "Terminator"-style cyborgs/robots.
At least an emotionless AI has no concept of enjoying other's suffering/pain, or blind and illogical hatred, revenge for revenge's sake, lust for power, wealth, and domination, or any of a thousand other similar negative human thoughts/feelings/behaviors.
How long do you think that the US government would continue to even pretend to acknowledge individual rights, Rule of Law, or any limits at all set by the US Constitution, if they had a functional robotic army and didn't have to worry about the human US military fracturing if ordered to attack US citizens en-masse? The speed of their deciding our fate would rival that of Skynet's.
"Give me a plasma rifle in the 40-watt range."
Strat
Progressivism (aka US 'Liberalism'): Ideas so good they need a police/surveillance-state to enforce.