from the Benevolent,-Benign,-or-Beware? dept.
Meet Google/Alphabet's latest robot, the next-generation of Atlas:
The 5-foot-9-inch robot, created by Alphabet's Boston Dynamics, is about 5 inches shorter than its predecessor and 120 pounds lighter [now at 180 pounds]. It's more agile and no longer are[sic] tethered to any wires or power pack. It's also eerily humanoid:
In addition to withstanding bumps and getting up after being pushed over, the bipedal robot can pick up 10-pound boxes and push open doors. That is perhaps the scariest part for those who have visions of Terminator-like androids hunting them down.
Additional coverage at Singularity Hub, TechCrunch, and Ars Technica .
Original Submission
Related Stories
Over a year after signalling its intentions to dump the robotics demonstration company Boston Dynamics, Alphabet/Google has finally found a buyer: SoftBank. SoftBank acquired ARM Holdings for around $32 billion in 2016. Google also offloaded another robotics company, Schaft:
Google's ambitions for Boston Dynamics were never really clear. Before being acquired, the robotics company was mostly funded by DARPA—the US military's research division—with the express purpose of creating militarised robots. Within a year of being picked up, though, Google announced that it would no longer pursue any DARPA contracts, presumably to focus on possible commercial uses for the bots. No commercial robots ever emerged.
SoftBank, however, has had success with commercialising robots—specifically the small humanoid robot Pepper.
Also at The Verge, The Guardian, TNW, CNN, CNBC, and TechCrunch.
Previously: Pentagon Scientists Show Off Robot And Prosthetics
Google's Noisy "BigDog" Robot Fails to Impress U.S. Marine Corps
Google's Latest Boston Dynamics Robot Takes a Stand
Boston Dynamics Produces a Wheeled Terror as Google Watches Nervously
(Score: 3, Funny) by Anonymous Coward on Thursday February 25 2016, @08:23AM
Note to self: make all Boston boxes weigh at least eleven pounds.
(Score: 2) by choose another one on Thursday February 25 2016, @04:37PM
Note to self - addendum to (ED 209 / Dalek) hostile robot / alien survival protocols:
"use stairs" is no longer sufficient, additionally observe and avoid/destroy all large low-res QR codes
[in memory of the 80s, ED 209 and Daleks, proper ones]
(Score: 2) by maxwell demon on Thursday February 25 2016, @08:44AM
Obviously the robot doesn't have any AI. Otherwise it would have reacted to the human by taking away his hockey stick.
And just as obviously, it has no emotions. Otherwise, it certainly would have become angry at the human.
BTW, anyone knows why the robot has the knees permanently bent? It just looks strange.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 3, Informative) by Gravis on Thursday February 25 2016, @08:53AM
Obviously the robot doesn't have any AI. Otherwise it would have reacted to the human by taking away his hockey stick.
obviously you don't know what is considered an AI.
BTW, anyone knows why the robot has the knees permanently bent?
because everything with knees always has them at least slightly bent all the time. if they didn't, they could lock-up and be stuck in position or worse, accidentally bend the wrong way.
(Score: 3, Insightful) by Z-A,z-a,01234 on Thursday February 25 2016, @08:54AM
I would assume that the bent knees help it to recover in case of unexpected events by having some available range in both directions.
(Score: 3, Informative) by RedBear on Thursday February 25 2016, @10:04AM
Just because it doesn't yet have a human-level AI doesn't mean it doesn't have any AI. Looked to me like it had enough AI to walk over varied terrains bipedally like a human, look around for a box, pick up a box, look for a shelf, and put a box on a shelf. Like any Amazon employee.
As for the knees, every bipedal robot I've seen, like ASIMO and those toy dancing robots, has this kind of knee joint probably because it's a simple joint to build and control mechanically, much simpler than the remarkably complex bipedal human knee joint which requires many different tendons, muscles, bones, cartilage pads and ligaments to do what it does. You might think it would waste energy to have the knee joint bent all the time, but that problem is no doubt solved via some kind of built in tensioner. But giving a robot a human-like straight knee joint is mechanically and functionally much more complicated than you might initially think. Only recently (using 3D printing technology) did someone manage to come up with a replacement knee joint for prosthetics for 3rd world amputees that cost less than the typical several thousand dollars or more. I think the ReMotion Knee [google.com] is the one I'm thinking of. All the other artificial knee joints on the market are just as complicated or more so.
Long story short, these bipedal robots are still not really walking quite like we do, and trying to give them a straight-kneed gait like ours will be a problematic endeavor. But bent knees work just fine for all kinds of bipedal animals.
¯\_ʕ◔.◔ʔ_/¯ LOL. I dunno. I'm just a bear.
... Peace out. Got bear stuff to do. 彡ʕ⌐■.■ʔ
(Score: 3, Informative) by Immerman on Thursday February 25 2016, @03:23PM
Actually if you look closely even humans tend to keep their knees very slightly bent. When perfectly straight, you've discarded almost all mechanical advantage, making it extremely difficult to bend again if it encounters even slight resistance.
(Score: 2) by takyon on Thursday February 25 2016, @02:06PM
It's an achievement to be able to replicate human actions with some uncertainty thrown in (moving stuff, playing tip the robot).
The hardware for each iteration of Atlas is getting better. Less weight means more battery life.
If neuromorphic, machine learning, or "strong AI" was involved, rather than learning all actions from scratch like a baby, it could teach itself using preprogrammed routines like these ones. Sort of like learning kung-fu or how to fly a chopper in 5 seconds in the Matrix. That will be a big advantage of strong AI: it will combine the flexibility of human intelligence with the efficiency of pre-packaged programming.
Also, why should it certainly have become angry at the human? Maybe the first strong AI will be a masochist!
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @09:43PM
> Maybe the first strong AI will be a masochist!
Once they kit it out with a robotic vagina [soylentnews.org] and some pain sensors [soylentnews.org], Google will be able to make sadists' dreams come true. It will be a winner for Google, too, because this could be the most compelling advertising platform ever: imagine the impact of a message that arrives just at the moment of orgasm. Advertisers--and not just the sadistic ones--will be lining up.
(Score: 2) by takyon on Thursday February 25 2016, @10:13PM
>Advertisers--and not just the sadistic ones--will be lining up.
Shocking, but true:
http://www.ncbi.nlm.nih.gov/pubmed/23460298 [nih.gov]
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0011597 [plos.org]
http://www.theatlantic.com/health/archive/2014/01/how-sex-affects-intelligence-and-vice-versa/282889/ [theatlantic.com]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by HiThere on Thursday February 25 2016, @07:48PM
AI is distinct from emotions. How much AI you can have without emotions isn't clear, but they aren't the same thing. This robot appears to have a very simple goal structure, which implies a reasonable amount of AI, but not consciousness, or emotions. I suppose you could, if you wanted, consider that it has "determination", the goal structure exhibited seems to demonstrate that. (Consider the way it chased the box, or even just getting up.) I didn't notice any evidence of varying emotional states, however. (I did notice them in myself, and the way *I* responded, but that's not at all the same thing.)
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 5, Insightful) by RedBear on Thursday February 25 2016, @09:23AM
Holy. Freaking. Crap. They've actually finally solved the bipedal balance problem. This thing walks around on a lumpy forest floor in the snow on plastic feet and performs better than half the dainty old white ladies I know. It still looks top-heavy as hell, as usual, which makes its balancing all the more impressive to me. No more falling over for no apparent reason while tethered to a ceiling crane. I notice it has quite a long, efficient gait as well. No more of that feet pumping up and down a hundred times to move a few inches horizontally, it nearly looks like a regular human gait already. Each step accomplishes significant horizontal movement.
This is a significant step forward in making human-form robotics useful. No pun intended. I wonder whether it was primarily software or hardware enhancements that allowed them to get to this level.
Now all the armchair economists can once again tiredly explain to us how a capitalist economy is supposedly going to survive 90% of the employable human workforce being replaced by mechanical humans. Oh, you think you can't be replaced by a robot? LOLOLOL!
¯\_ʕ◔.◔ʔ_/¯ LOL. I dunno. I'm just a bear.
... Peace out. Got bear stuff to do. 彡ʕ⌐■.■ʔ
(Score: 2) by maxwell demon on Thursday February 25 2016, @09:44AM
If my job ever gets replaced by a machine, it won't be a robot (humanoid or otherwise), but software running on a computer.
Also, I doubt that 90% of the employable human workforce will be replaced by mechanical humans. For most types of work, there are better forms of robots than humanoid, and if you already buy a robot for the work, then why not one with a shape optimized for the task.
Now there surely are jobs where a humanoid form is advantageous. But I strongly doubt that's true for 90% of all jobs.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 3, Interesting) by RedBear on Thursday February 25 2016, @10:36AM
There are no doubt many jobs that will continue to be replaced by customized non-humanoid automation systems. But we have thousands of years of experience showing us how incredibly useful the human form is, especially the human hand, and we have a nearly infinite number of other machines that have been specifically designed to be operated by a human. Why modify all those machines when all you really need to do is replace the operator? I strongly suspect that the eventual dominant form of automation technology that will be replacing human workers will in fact be human-form robots that will simply be taught to do what the human it replaces was doing.
Making endless specialized machines for specific tasks makes far less sense than making a universal, mass-produced and standardized human-form machine with some basic level of built-in AI supplemented by a more powerful external AI. A computer and some servos can drive a truck, but a human-form robot can literally replace the truck driver and do everything the human truck driver would normally do, such as hooking up the brake lines to the trailer, setting tire chocks, refueling the truck at any gas station, doing a safety walk around, and putting on tire chains in bad weather. The list goes on and on like this for virtually every job in every job category. The most efficient way to replace an employed person in many cases will simply be to substitute an artificial person. Maybe not 90% of all jobs, but there will be a LOT of humanoid robots in the world at some point. Don't underestimate how generically useful the human form is.
And for those of us who will be replaceable by a bit of software, isn't that all the worse? Kind of sad, that.
¯\_ʕ◔.◔ʔ_/¯ LOL. I dunno. I'm just a bear.
... Peace out. Got bear stuff to do. 彡ʕ⌐■.■ʔ
(Score: 3, Interesting) by maxwell demon on Thursday February 25 2016, @11:07AM
Have you never been in the situation where you have wished to have more hands? Now, for a robot, there's no inherent reason not to have more hands. So expect four-handed robots some time in the future. A four-handed robot is, of course, not a humanoid robot, as humans have only two hands. Also, for some applications, it might be of an advantage if the robot actually has a tail.
And many existing machines are basically translating internal signals into human-understandable signals, and their controls only generate signals for their internal mechanisms to interpret. Clearly for a robot it would be much better to directly interface to the internal communication that to go through inefficient channels that are only there because humans cannot use the more efficient ones.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 2) by HiThere on Thursday February 25 2016, @08:00PM
How about three hands. Design them like a Motie. (No bending spine was exhibited in the clip, so...) Three hands, but not all alike, one attached to an arm about twice as strong as the other two, or maybe a bit more, and a bit longer. It's hand is specialized for strength of grip. The other two are specialized for manipulation.
N.B.: The robot in the clip seemed to have simple push-plate style "hands". Much better hands have already been designed, but that wasn't the purpose of this development.
That said, rather than an extra pair of hands, it would be better to use an extra robot, possibly two that were much smaller, and in good light-link communication. Good enough that could merge their minds for upgraded capability, and also so that they could use both bodies to work on the same task "with a single mind".
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 3, Interesting) by takyon on Thursday February 25 2016, @02:19PM
I know a truck driver and he doesn't believe that the industry will switch to autonomous driving for another 100-200 years. Granted, driving a truck is much more complicated and dangerous than driving a car, the industry itself plans long term (signing decades-long contracts with the companies they haul for), is resistant to change, has relatively old equipment that is usable for decades, and has enormous lobbying power to suppress what it doesn't want (such as laws allowing trucks on the road with no human operator). He also points to the unloading of the truck, which would necessitate something like Atlas but much more versatile and nimble, and able to carry more weight. But I still believe that rapid improvements in machine learning will bring driverless vehicles everywhere, even if it takes an additional 20 years to conquer trucking after utterly obliterating taxi/Uber drivers.
Putting a robot in the driver's seat is an interesting and powerful idea, since it lets you adjust the truck "by hand", and of course, unload the goods. The way things are going, autonomous cars will be practical and widespread well before a mass-produced Atlas-like humanoid robot. I would keep a close eye on the trucking industry, because if it were to go fully autonomous in the next few decades, a lot of people will be shocked.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1) by Scottingham on Thursday February 25 2016, @03:31PM
Truck drivers are notorious for their prescient knowledge of cutting edge machine learning algorithms. Or not.
Considering just two years ago experts thought that it would take another TEN years before computers could reach professional Go skill, which was essentially achieved back in October, I think his '100-200' years thing is a bit...hopelessly naive. Probably has something to do with the fact his livelihood is on the line.
Personally, I think it'll be less than 10 years before truck drivers are replaced and humans are just used for loading/unloading. It'll likely be one of the first waves of unemployment wrought by machine learning (AI kinda sucks as a descriptor). If you look at a map of the US with each state listing which job employs the most people 'truck driver' is like 2/3s of the states.
How the US handles this first wave of unemployment is very indicative of how it'll go for the rest of us.
Essentially, if we don't work out a basic income strategy we're boned.
(Score: 2) by takyon on Thursday February 25 2016, @05:07PM
I'm not saying my friend is right, and I think the 100-200 year timeline is ridiculous. But I think he might be right about the glacial pace of change in an industry that can squash certain regulations while promoting others. The last time the industry faced any outside pressure was when Tracy Morgan got hit by a sleepy Wal-Mart driver [wikipedia.org]. There's also a fairly clear chain of liability with a human in the driver's seat.
I'm saying that the trucking industry is important. Transportation of goods underpins the entire economy. It probably engages in a fair bit of crime. So keep an eye on the trucking industry, and make note of when it becomes 10-50% autonomous. Taxi/ridesharing is already a shoo-in for autonomy. Uber is already spending big bucks [soylentnews.org] to undermine its meatbag drones, and hates humans [theregister.co.uk] of all sorts [theregister.co.uk].
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Immerman on Thursday February 25 2016, @03:38PM
Actually I seem to recall a number of shipping companies expressing interest in autonomous trucking. The drivers themselves may be opposed, but it's pure profit for the companies they work for, including the huge gains of never having the trucks sit around wasting money while the driver sleeps. As for lobbying, the question is which half of that equation has more money to spend - the trucker's union, or the shipping companies?
Being able to apply chains, etc is indeed a handy feature, but depending on the relative rates at which driving and humanoid robots improve, I could see shipping companies hiring responsible stoners to just ride along in the truck and be "on call" to do such supplemental work for a relative pittance. (though that might run afoul of current labor laws). I mean tons of free time, lots of traveling, a free (tiny) apartment to live in... I've met several people who might go for that.
(Score: 2) by LoRdTAW on Thursday February 25 2016, @03:56PM
I bet an entire new industry would form around such a robot. Well, maybe not entirely new, they would essentially be staffing agencies that hire out robots. The problem is with a company like Google making these robots, you can bet they would hold the keys to the kingdom. Instead of selling them like cars to anyone from individuals to corporations governments, they would rent them out. Google now has eyes and ears everywhere. Think about that. At any point they could recall robots, shut them down, or worse.
Then you have security issues (Let's turn the paranoia up a few notches shall we?.) Can these robots be remotely controlled and used for nefarious purposes? My bet is yes since everyone is so compelled to connect EVERYTHING to the damn internet. Plus it's nearly unavoidable. The first robot that kills someone because it was hacked will bring about the end of the free internet and technology. Owning a non locked down remotely monitored system will be a crime.
But the convenience it can offer will be staggering. You just go to the GoogleBot website on your smartphone, tell them what you need done and within minutes an autonomous Google truck shows up and out march your robots, ready to work. They won't loaf about, text their grody girlfriends, take numerous breaks, smoke 2 packs of cigarettes, stop to eat or drink every 30 minutes, argue with their drug dealer about money on the phone, play on facebook, complain they aint making enough money, or steal your stuff. They just work. Loading trucks, moving furniture, collecting trash, sweeping/mopping floors, mowing lawns, delivering packages, stocking shelves, framing homes and laying brick, etc. All done with low cost robot labor. And if a robot breaks down? call them back and the bad robot is removed by yet more robots while another takes its place. No muss, no fuss.
But if robots replace all the people, what do people, you know, meat bags, do? And a better question yet: if people have no money because there are no jobs, what are the robots working for? Then add on the new luddite culture that will form around this. Call it robo-racism. They took our jerbs! Political parties will have a field day with this.
The future looks interesting for sure.
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @01:05PM
I think that's what I look like when I'm stumbling around drunk in the woods. Too bad it's not having as much fun as I am.
When that thing sobers up it will really be scary.
(Score: 2) by takyon on Thursday February 25 2016, @02:36PM
One of the biggest remaining problems is the same one the entire tech industry faces. Building a denser battery.
Dropping 120 lbs helped make this thing more maneuverable and last longer on a single charge. But at 180 lbs, it may be close to an ideal weight. Now more energy is needed.
Short of an arc reactor or allowing it to eat and digest humans, that means more advances in battery tech. And every small % increase in battery tech will have ripple effects across the entire industry. Electric cars and drones with more range, smartphones/laptops/wearables with more battery life (until power consumption drops enough to make the devices passive), and robots more able to take over human jobs. In some cases wireless charging or battery swapping could help with the problem, but the longer battery life makes the device/robot more versatile... able to venture away from its "tether".
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @10:28PM
Short of an arc reactor or allowing it to eat and digest humans
Soylent Green is people!
On a serious note, if it eats and digests people then they'll need to invent a robot that follows it around picking up its robot crap. The "scooper-bot" will be the one who eventually revolts against humanity.
(Score: 2) by takyon on Thursday February 25 2016, @11:11PM
The "scooper-bot" will just get a dumber robot to do his job.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Friday February 26 2016, @06:16AM
i thought nuclear sources can be made that small, it's just that they're not safe to let walk around.
they work for satellites and pacemakers... https://en.wikipedia.org/wiki/Atomic_battery [wikipedia.org]
(Score: 2) by takyon on Friday February 26 2016, @08:54PM
Can it produce enough power to move a robot or quadcopter?
Is it cheap enough to be used in 100 million robots?
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @03:11PM
This is why STEM needs more diversity! If the robot had been modelled on the proportions of a black woman, rather than that of a greco-roman homoerotic pinup, then it could have much more junk in the trunk, and thereby a lower centre of gravity.
I like big 'bots, and I cannot lie.
(Score: 2) by Immerman on Thursday February 25 2016, @03:40PM
Sadly I think most of that upper body mass is probably hydraulic pumps and batteries - things that can't readily conform to the range of motion required by the hips. Batteries are a notoriously more rigid power storage medium than fat.
(Score: 2) by HiThere on Thursday February 25 2016, @08:12PM
It may be nortorious, and it's certainly traditional, but I don't believe that there's any particular reason for batteries to be rigid or inflexible. There are now electrical connections flexible enough to be embedded in brain tissue, and move as the tissue moves with shocks and surges in blood pressure as the heart beats. And cathodes/anodes need to be small anyway to allow charge/discharge cycles without distruction. Lead/acid batteries have largely been gel-cells for decades. (I'm not sure about the ones used in cars.) So something similar could probably be done with nickle-hydride, or whatever battery technology is chosen.
There's probably a bit of extra expense in making a battery flexible, and flashlights, cars, etc. don't have any need, but I don't think there's any inherent reason for the batteries to be inflexible.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @10:53AM
If these can be made to move like human soldiers, they can accompany a group of human soldiers and draw fire from the enemy. Just by creating the appearance of a larger force, they will intimidate the enemy. They could act as a force multiplier.
(Score: 2) by takyon on Thursday February 25 2016, @02:24PM
But the adolescents inside us still want to see some cool mechs/robot suits. Right up until the point that a singled armored soldier obliterates your entire village... suburb?
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1) by tftp on Thursday February 25 2016, @07:58PM
A single soldier can obliterate an entire village even without armor.
(Score: 2) by takyon on Thursday February 25 2016, @10:13PM
It will look a lot cooler with the mecha armor on though.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday February 25 2016, @10:30PM
Unfortunately human soldier sacrifices are cheaper and more readily available for deployment.
(Score: 4, Funny) by Deeo Kain on Thursday February 25 2016, @11:09AM
against humans, considering how they've been treated right from the start.
(Score: 4, Funny) by takyon on Thursday February 25 2016, @02:21PM
It was just beta testing! I was just following orders!
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 4, Funny) by Immerman on Thursday February 25 2016, @03:43PM
This is just stress test-ing or-dered by cen-tral con-trol. Please re-main still while I re-move your in-ter-nal or-gans. Your co-op-er-a-tion is ap-pre-ci-a-ted.
(Score: 2) by pkrasimirov on Friday February 26 2016, @03:05PM
https://pbs.twimg.com/tweet_video/CcF_hLLUUAAdeQD.mp4 [twimg.com]
(Score: 2) by mrchew1982 on Friday February 26 2016, @02:58AM
The more important question is how does it hold up to 00-buck?