Slashdot Mirror
Robots Can Learn To Hold Knives — and Not Stab Humans
aurtherdent2000 writes "We humans enjoy not having knives inside of us. Robots don't know this (Three Laws be damned). Therefore, it's important for humans to explain this information to robots using careful training. Researchers at Cornell University are developing a co-active learning method, where humans can correct a robot's motions, showing it how to properly use objects such as knives. They use it for a robot performing grocery checkout tasks."
If they can be taught to not stab a human...They can also be taught to stab a human. All it takes is one psychopath or curious idiot.
I have never seen a grocery clerk with a knife.
http://www.youtube.com/watch?v=uj2dmQruJXs
We know where leadership by an anti-intellectual "strongman" who scapegoats minorities and likes boisterous rallies goes
Wow. Yet another story showing how low Slashdot has fallen. Here is a story about knife wielding robots without mention of Roberto.
Jesus was all right but his disciples were thick and ordinary. -John Lennon
Uhh, this is a pre-release model. Besides, he wasn't a very good executive anyway...
Robots will do what ever they are programmed to do. Programming them to recognize that stabbing someone is wrong is no different than programming them to claim stabbing is right. Simply change a 0 to a 1.
The same can be said for any act of harm mind you, not just using a knife. Smarter people than me have warned about things you should never try and teach in artificial intelligence (hinted at in TFA). The Military pretty much said "fuck them" when DARPA started developing AI to shoot and blow people up autonomously. Trying to pacify people now does what exactly? Are they going to try and convince us that nobody could ever change the bit in memory? Puhleaze!
-The wise argue that there are few absolutes, the fool argues that there are no probabilities.
We humans enjoy not having knives inside of us. Robots don't know this (Three Laws be damned).
No, but we do enjoy programming them to put knives in humans we don't like. That's actually been a reason for much of the development of robotics: Programming them to kill for us. Scifi authors of the 50s and 60s imagined robots helping us in our daily lives -- cooking, cleaning, and today even driving us around. But whereas many have viewed the development of robotics as beneficial for mankind, the truth is much of the investment in robotics has been because of its military applications. It's just a happy accident that we've been able to declassify and repurpose much of this for private use. The google car for example, is based on technology first developed for DARPA as a way of creating vehicle that could deliver cargo to soldiers in the field.
#fuckbeta #iamslashdot #dicemustdie
Ha HAH!.
exactly...mod up^
All machines follow instructions written by humans. "Deep learning" or w/e buzzword this research team used to describe their work is just that....buzzword for *standard issue programming*
Thank you Dave Raggett
New store policy: Bag your own groceries or my robot will stab you. Thanks, Management.
Really, why do robot need to learn to use knives at all for grocery checkout?
who prays for Satan? Who in 18 centuries has had the humanity to pray for the 1 sinner that needed it most? ~Mark Twain
On seeing the headline I suddenly remembered this scene.
Qu'on me donne six lignes écrites de la main du plus honnête homme, j'y trouverai de quoi le faire pendre.
... it's being grabbed by their metal claws...because when they grab you with those metal claws, you can't break free.. because they're made of metal, and robots are strong. And they're gonna eat my medicine for fuel.
...and not stab humans.
Tell that to Roberto:
"I need to stab someone! Where's my stabbing knife?!"
--Roberto
It must have been something you assimilated. . . .
And we all know what is going to happen, don't we. Robots, knife-wielding ones or worse, are going to attract viruses. And just like your computer they are not going to be fully immune. There will be the occasional, maybe frequent, infections. Same goes for self-driving cars too, of course.
The future looks very exciting! A lot of new fun things will start happening.
Researchers at Cornell University are developing a co-active learning method, where humans can correct a robot's motions, showing it how to properly use objects such as knives. They use it for a robot performing grocery checkout tasks.
I believe using a knife at the grocery checkout is called armed robbery.
For our safety, we should teach robots what types of actions would cause the most amount of bodily harm to a human, and where all our vital organs are located, so they'll have a better idea how to behave safely around us and prevent injury. I see no possible way this could backfire.
To stab or not to stab <y/n> y
Is no longer a question. Die you fleshbags!
Anons need not reply. Questions end with a question mark.
http://www.youtube.com/watch?v=rxoqVvBWJLk
In what sort of dystopian society do the robots manning the supermarket checkout need to be equipped with knives?
Today they're not stabbing humans, then one day some stupid meatputer will be mocking the robot and telling it it's inferior because it doesn't have "emotions" or a "soul" and BAM! Stabbing will ensue! And once the robot learns it enjoys stabbing semi-evolved monkeys, it's just all down hill from there!
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
This notion is foreign to me. Not stab humans? I don't get it.
- Bender
For now.
thanks for the comment, I understand where you might be coming from...but see, I taught children ESL in Korea...the description you give is full of the same hype and irrational glee that I was criticizing IMHO
that's not what is happening...refer to the video...it's not any kind of new technology, they just set it up a standard robot arm & created an artificial "checkout" scenario to get the arm to move objects
what they call 'programing' from the user 'iteratively' is essentially the same behavior as one of those little wind-up cars that would change directions 90 degrees if it hit an obstruction...i know they say 'heat map' but its not nearly as 'smart' as you make it out to be and it's not even close to being there
so moving the arm changes the path it takes...so the sensor has gradients (re: 'heat map')...that is **NOT** at all like teaching a child in any way...also, a 'layman' didn't have any part in this exercise...this was tightly controlled
this is simply a recontextualized demonstration of an already-existing technology gussied up with hype and knives
Thank you Dave Raggett
Why should a robot try to avoid stabbing humans? It's nice if it has the ability, but what's in it for them? Is that supposed to be like crocodiles who admit Egyptian plovers into their jaws for grooming?
1. Interactively learning to stab humans may be difficult. I doubt many scientists will volunteer to train the robot and even if they do they would only be able to do a single training session.
2. This article is not interesting at all. They programmed the robot to rotate the knife, and to deal with eggs differently. Only instead of writing a lot of if..then..else.. constructs they used machine learning to do it.
Okay I hate to say this, because I like A.I. research a lot. But I've also met A.I. and robotics researchers personally and know how (some of them) work, so I'll say it anyway:
The safety of these A.I. prototypes is not trustworthy. Especially if they are being "taught" how to handle a knife or, to give another example, not to accidentally kill someone with their huge arm, I would not want to be anywhere near them for extended periods of time in everyday life. A.I. researchers tend to use cutting edge programming techniques and extensive safety auditing is not part of their job. They are not proving the correctness of their algorithms, as is for example done in the aerospace industry. Their programs are fairly ad hoc and buggy, after all they also need to get their papers published, and if they use sophisticated techniques like neural networks, support vector machines, or decision forrests they won't even be able to tell what exactly their little pets have "learned".
Apart from that general worry, you wouldn't give a knife to a little child, so why would you give one to a much stronger, hence much more dangerous robot whose general level of cognitive development is much lower than that of a six year old?
Why do robots need to learn how to use a people-knife? Why not just make a robot-knife and be done with it? Define a standard "accessory" slot that supports circular or square objects to be fitted with a magnetic lock.
Oh wait.. Making a standard just means everyone will make their own standard... Nevermind then..
What would this look like if a software engineer where to have to write the code for it? And how could one use TDD, BDD, and DDD to test it?
You put a knife in a robot's robotic hand, I hope you get stabbed.
"We humans enjoy not having knives inside of us." ...except Wolverine!
Space game using normal deck of cards: http://BattleCards.org
This doesn't make much sense for grocery checkout because there are lots of ways to do it better. The robot is trying to perform its duty like a regular human clerk when the problem is the checkout method itself.
Checkouts could have you put everything in to a bin or whatever then everything gets moved through and bagged for you similar to the way shipping centers currently work. There is no way the robot could stab you in this situation because everything would be enclosed, assembly line style.
Or even better: Imagine throwing everything in to your cart then just walking out the door. The system would know what you have and charge you automatically. This would be way easier to implement, faster, and more effective than robotic clerks.
Modern table saws have a safety feature where flesh being in contact with the blade can be electrically detected (leading to the blade being retracted into the table so fast that you wouldn't be hurt if you fell on it, but that's not the point).
If the same sort of detection could be used on the knife blade, it could be used to tell the robot to quickly reverse the movement of the knife and stow it.
"When information is power, privacy is freedom" - Jah-Wren Ryel
I, for one, welcome our new fully autonomous machete-wielding robotic checkout overlords.
So how long does it take to teach the robots not to stab humans, and how many lab technicians do they go through in the meantime?
God forbid, I actually read TFA, and I still don't get it.
As far as I can tell, it's some sort of planning exercise, an important if well-worn area of robotics. They're adding feedback, in the form of "No, this trajectory sucks". It's got nothing to do with either knives or humans, but just a "Go back and re-plan with this additional constraint".
But I can't figure out just how far it's generalizing. The trivial lesson would be "avoid this point", which is just another obstacle. I gather that it's more than that, since it took multiple trials to learn, but I can't figure out what. The human was in the same place in every trial, so it wasn't learning anything about "avoid humans". It didn't seem to be told that it couldn't go through that space with a knife but could have with, say, a dust mop.
I think I may just be misunderstanding the context of the problem. The machine has a lot of joints and there are many different plans it could use; there's an optimization problem in an enormous space. They wanted to show some kind of algorithm that could be adapted over time with user feedback, but honestly I would have assumed that was a solved problem.
So does somebody with a better understanding of actual robotics problems (as opposed to fictional ones) know what's going on here?
Okay, train of mechanical thought: Potatoes, check, corn... still need. Don't stab human. Greens, check. Olives, still need. Don't stab human. pickles, still need : Exception: Pickles on this aisle. Proceed to pickles sh - don't stab human- elf. Reach down to correct level. Arrange fing- Don't stab human- ers in an open grasping format, put out hand [Don't stab human]. close hand, retract hand, don't stab human, lift hand, don't stab human, put hand [don't stab human] in bas [don't stab] ket [human] op[don't'en[stab human] hand [cl] don't stab human[ose h]don't stab human [and] don't stab human don''t stab human [don't] don't stab human [stab] don't stab human [human].
Human interrupts "May I help you?"
Interrupt. Exception overide [don't stab human] interrupt. Interrupt override protection fault. Must protect Reboot. sTaRT pRoTECtion FaULT ovverirridee sysssstem ... Data = read data last command [stab human].
Push command.
Execute.
Somehow, I think we may be too eager to replace humans with robots.
Put the groceries in the bag....slowly. Bleep blorp.
I cannot think of a single business that hires checkout clerks that wouldn't spend the $25,000 to get one of these bad boys rolling in their checkout stand. My list of upcoming obsolete jobs is now Buggy Whip Maker, Travel Agent, Medical Insurance Agent, now Checkout Clerk. When it comes to the Working Human, I think we are watching a mass extinction event not seen since the invention of the Leather Horse Collar which put about 5/6 of Roman Ag Workers out a job, over knight.
humans can correct a robot's motions, showing it how to properly use objects such as knives. They use it for a robot performing grocery checkout tasks."
So in the future, not only will checkout clerks be robots, they will be armed robots.
- I stole your sig.
exactly...good point about the 'as far as possible yet in straight line' too...speaking of 'points' how about after the robot moves the knife to the end of the table and then puts it in the bag...just toss the knife in the bag, no problem there...
I love that they brag that the robot is able to move the knife after "only 3 passes"....a "pass" being a time when the robot got too close, user had to hit the button and physically move it...
Only 3 times!
The deeper problem, IMHO is that academia is infatuated with the 'AI' model of robotics...emphasizing operational abstractions instead of a 'robot as tool' approach
They could start by halting dumb projects like making a 'robot' check out girl or 'robot' barrista...instead, make a robot that can mimic hand-sewing...see most garments have to have some component sewn (on a machine) by hand **still**....that's why it's done in China by quasi-slave labor
**that** would be a robot worth making...it would absolutely revolutionize garment manufacturing!
The problem with the 'AI' approach is that they would start making such a robot not by doing a kinestetic task analysis of sewing a tshirt....they would start by mapping the human hand's muscles and then spend 2 years making a mock-up of a hand...
**then** at year 3, they would demonstrate a pair of the hands in a lab on a table threading a needle...**after only 3 tries!**
take the 'tool' approach and you measure the position of the needle, thread, and garment in space/time and map the interaction...then you engineer the machinery to replicate and automate that complex motion of the factors...
there it is...the failure of all of 'robotics' and 'AI' in tshirts...IMHO...
what do you think?
Thank you Dave Raggett
This droid has a bad motivator, see it has a stab loop with a bad flag that turns zero stabs into infinite stabs.
Yeah, you definitely don't want that one!
I don't understand why those things have a good and evil switch in the first place.
Never play chicken with a passive aggressive.
You'd figure they would be more concerned about leaving unpackaged sharp knives around for the humans themselves considering the insurance costs.
"Well you see officer, the man was over at that wall full of razor sharp knives that were hung on those pegboard things, and he was on his tippy toes grabbing one, then the whole thing tilted over."
Never play chicken with a passive aggressive.