Robo Brain Project Wants To Turn the Internet Into a Robotic Hivemind

malachiorion writes Researchers are force-feeding the internet into a system called Robo Brain. The system has absorbed a billion images and 120,000 YouTube videos so far, and aims to digest 10 times that within a year, in order to create machine-readable commands for robots—how to pour coffee, for example. From the article: "The goal is as direct as the project’s name—to create a centralized, always-online brain for robots to tap into. The more Robo Brain learns from the internet, the more direct lessons it can share with connected machines. How do you turn on a toaster? Robo Brain knows, and can share 3D images of the appliance and the relevant components. It can tell a robot what a coffee mug looks like, and how to carry it by the handle without dumping the contents. It can recognize when a human is watching a television by gauging relative positions, and advise against wandering between the two. Robo Brain looks at a chair or a stool, and knows that these are things that people sit on. It’s a system that understands context, and turns complex associations into direct commands for physical robots."

Comments

You can almost see it comming:

[balaband]

1. Post video doing some pretty innocent action (watching tv, pouring coffee, sweeping the house) as having sex with a lamp post
2. Rick roll the video, so it gets plenty of hits and Robo Brain takes it into consideration
3. ???
4. Profit!

Re:You can almost see it comming:

[iluvcapra]

1. Post video doing some pretty innocent action (watching tv, pouring coffee, sweeping the house) as having sex with a lamp post

Just don't let WALL-E see any Harmony Korine movies.

Re:You can almost see it comming:

[malachiorion]

Awesome. It does make me wonder whether the various researchers watch any of the videos in their entirety, before feeding them into Robo Brain. I should have asked the guy I spoke to.

Re:You can almost see it comming:

[psergiu]

If the AI already absorbed highly-rated YouTube knowledge on how do do basic stuff ... well ... i really don't waht a house robot with that knowledge:

https://www.youtube.com/user/H...

Youtube Comments

[xlosermagnetx]

Hopefully the Youtube videos they are feeding it don't include the associated terrible Youtube comments.

Re:Youtube Comments

[charronia]

Researchers are still determining the reason behind Robo Brain's incredibly rude behavior.

Re:Youtube Comments

[ColdWetDog]

The system has absorbed a billion images and 120,000 YouTube videos so far,

What this really points out is that we need to lay the groundwork for the Society for Prevention of Cruelty to Robots.

Re:Youtube Comments

[TWX]

I really don't want to know what the robots will learn when they start watching Chatroulette...

Re:Youtube Comments

[s1d3track3D]

Actually, that is a separate project for creating the dumbest, most bigoted robo brain ever

Re:Youtube Comments

[Dins]

I really don't want to know what the robots will learn when they start watching Chatroulette...

A lot about male anatomy would be my guess.

Re:Youtube Comments

[meta-monkey]

Or it could be used as a robot guide for "what not to do." Basically, your robot would be sure to tell you that are not a faggot and should neither kill yourself nor catch AIDS and then die in a fire.

Re:Youtube Comments

[Coisiche]

This is all reminding me of an episode of Odyssey 5 where Ted Raimi portrays an AI that learned everything from those aspects of the internet and was inadvertently housed in a synthetic body instead of the nastier AI the body was intended for.

He turned out not to be the villain that the Odyssey 5 crew had been expecting.

Open chamber, insert round....

[FF-Loucks]

Great... Once they are sentient this will be able to instruct them on how to operate the implements of our destruction. Hope they aren't feeding it copies of the snuff video that ISIS just posted.

Re:Open chamber, insert round....

[malachiorion]

Hopefully it'll become enamored of those hot ladies w/ guns videos, and direct all of its replicants to saunter into battle, hip-firing in their bikinis.

Obligatory

[king neckbeard]

I, for one, welcome our new robot overlords.

Sucking the internet dry

[fustakrakich]

Does it obey robots.txt files? Uh oh, here come the copyright lawyers with a take down notice...

But will it have problems selecting good sources?

[timrod]

I have this image in my head of someone buying a robot to do something simple like prepare food, having it browse YouTube for any mention of food and finding one of those ASMR videos of some random girl whispering into the microphone while she does simple household tasks, determining that this is in fact the correct way to prepare food, and then dragging the owner into the kitchen so it can whisper in their ears.

Re:Resistance is futile!

[ColdWetDog]

No, that started with a spelling checker.

Already there.

[khasim]

Roombas (and variants) are common household robots. YouTube has a lot of videos about Roombas cleaning a room while being ridden by a cat. Sometimes the cat is wearing a shark-suit.

Therefore, as this project progresses, Roombas will start to hunt cats in the neighborhood in order to get them to sit on top of them while they clean a room.

Or TFA is massively overstating the research and the concept and even robotics.

Re:Already there.

[gstoddart]

Roombas (and variants) are common household robots. YouTube has a lot of videos about Roombas cleaning a room while being ridden by a cat. Sometimes the cat is wearing a shark-suit.

If true, and I have no reason to doubt that it is, this is further evidence of why I consider YouTube to be a pointless web site.

Now, the video of someone trying to put a cat into a shark suit? Well, that sounds like the precursor to a Darwin award.

Re:Already there.

[matbury]

Does this mean that the terminators of the future will be modelled on cats in shark suits riding vaccuum cleaners? Is THIS how we'll all meet our demise?

We need positive Sci-Fi.

[jellomizer]

Robotic hive mind, just sounds like a bad idea.
We need some movies where the Robots and the Super Intelligent computer is the good guy for once. Just so we can get research grants and come up with neat new things.

Re:We need positive Sci-Fi.

[Zero__Kelvin]

"But, considering that they need an adversary, the humans would have to be the bad guys."

So it would be sci-non-fiction then?

Re:We need positive Sci-Fi.

[camperdave]

Robotic hive mind, just sounds like a bad idea.

Yeah, right! I suppose you think some prepubescent little twerp could fly a star fighter into the control ship, blow it up, and render an entire planetary invasion force of droid armies inert. Like that would ever happen!

Re:We need positive Sci-Fi.

[Kjella]

You do realize that if they actually did that /. would be howling about 1984 and Idiocracy and how it's NSA propaganda to "trust the system" and stop thinking for themselves with pages full quoting Franklin about security and liberty. After all it had to be about humanity willingly handing over control, we already had the story were they assume control by force and that's a villain story (I, robot). Meanwhile regular people like to identify with their heroes, the villains may be monsters or aliens or robots but the heroes are 99.99% human(oid). Nobody cares that you can't identify with Sauron or Smaug or King Kong or the Borg, but for a hero AI that's going to be tough.

You have it in the Swedish series "Äkta människor" where humanoid robots = hubots are blurring the lines, but it's more of a rebellion/independence story where they're breaking out of servitude and they're certainly not humanity's heroes. As an AI story it's more along the lines of Her, with humans and robots getting emotionally involved in each other which is probably not the kind of movie you were looking for. The superior intelligence kind of robot wouldn't fit in there, if you can't care about the hero I suppose you could care about the victims of whatever conflict that is drawn up. But those poor, helpless humans who can't fix their own situation but need outside help? That's a bit dreary.

Besides, I think that's a bit too similar to actual problems in the world today, movies like to show empowerment. People sympathize with those who are powerless victims, but they don't want to identify with them. And if the AI is smart enough to control bad guys, wouldn't it also be smart enough to control the good guys? I think it would be very hard to avoid it ending up as a giant puppeteer who's now pulling all the strings. And that's again more of a creepy story where you're being manipulated without you even knowing it, like the Matrix before you take the blue pill. There is the War Games computer who learns that the only way to win is not to play but I'm really struggling to come up with another computer "hero" story.

Re:We need positive Sci-Fi.

[SuricouRaven]

The original ending of I, Robot.

The movie changed it to the more modern cliche of rampant-AI-enslaving-the-world. In the original short story, a group of scientists uncovered an AI conspiracy towards world domination and discuss how to stop it - but then realise that these AIs are infallable, have no desire for power, money, sex or fame, and are incapable by design of acting against the best interests of mankind. So they decide to let the robots win.

Re:We need positive Sci-Fi.

[GameboyRMH]

People will be really bitter about that positive sci-fi when those research grants build a powerful AI which is immediately used to make a replica Manna system.

Technology is a neutral force multiplier, and in our society evil is more powerful than good. Unless the AI spontaneously turns evil in the story, as in the Terminator movies, it's not saying AI is scary and dangerous. It's saying our society is scary and dangerous and this is what we'll do with the power of AI.

So maybe it is better if we stay away from some technologies until we're collectively mature enough to do positive things with them instead of using them to make things worse.

Re:We need positive Sci-Fi.

[Nephandus]

We're not really collective. We're individually political and pursuing one's ends is mature in the rational sense, which beats deluded humanistic dogma every-time. Like the glaring example that is Vulcan religion, it's not logical to call reification-based totalitarian pipe-dreams logic...

Re:We need positive Sci-Fi.

[apraetor]

Problem is, it doesn't make for great cinema. What works great in literature doesn't always turn out very good on film. A healthy American knows this, and consumes all 3 media regularly - film, TV, book.

Re:We need positive Sci-Fi.

[dowens81625]

Who moderates the content up loaded is it like a Wikipedia thing?

Can I go looking for a "How do I chop watermelons?" sub routine for a "How do I make Fruit Salad" Program, and wind up with a killer robot, because someone either on purpose or accidentally defined watermelons as "Available round objects 9 to 13 inches in diameter" ?

Re:We need positive Sci-Fi.

[zeitnot]

Automata sounds like that.

More information please!

[MobyDisk]

Is there more information on this project? It sounds awesome, but the article seems quite sensationalized:

Researchers asked one of the project’s three robots to make affogato. The bot, a two-armed, highly-dextrous PR2, queried the system, and discovered that affogato was an italian dessert composed of ice cream and coffee. Without any human nudging or intervention, the robot located the coffee, figured out how to get it out of a dispenser, and poured it over the scooped ice cream.

Let's break that down.
A robot that can watch someone do something then repeat it would be the height of AI, as far as I know.
Software that can watch a youtube video and generalize what is a "chair" or "coffee" merely from watching videos would be beyond any AI I am currently aware of.
Software that can do that, PLUS recognize that what it is looking at now is coffee is completely unbelievable.
Software + a robot that can do all of the 3 above is not something I expect within my lifetime. Pure science fiction.

I'm sure there is a good project here, but I can't tell what it is from the sensationalistic article. I don't accept that it merely watches youtube videos and learned to make coffee in a completely unrelated environment. There has to be a lot more going on that just watching videos and generalizing.

Re:More information please!

[newbie_fantod]

TFA is published by Popular Science, which was also a major proponent of flying-cars back in the 60's

Re:More information please!

[Oligonicella]

FTA :

"(RoboEarth's files have to be processes and organized by humans)"

Coffee is connected to mugs, as well as to the motion-planning related to pouring liquid.

The bot, a two-armed, highly-dextrous PR2, queried the system, and discovered that affogato was an italian dessert composed of ice cream and coffee. Without any human nudging or intervention, the robot located the coffee, figured out how to get it out of a dispenser, and poured it over the scooped ice cream.

Notice it queried the system, not the videos themselves. "Without any human nudging or intervention" simply means "at that moment", in other words, it's programmed, not learning on its own. Also it didn't discover it was an Italian dessert, it didn't give a crap about the national origins and in all probability didn't even have that in it's database - it's just word padding to make things seem more human-like.

Re:More information please!

[SuricouRaven]

If they can use the approach to identify what sex looks like, they'd have an instant commercial application in web filter systems.

Re:More information please!

[gstoddart]

Or porn search engines ... because, rule #34.

Re:More information please!

[MobyDisk]

So, you don't expect a computer to be able to recognize what "coffee" looks like after...

Correct. Decision trees and neural nets can sorta do that, but they also need a human to mark which sections of the image correspond to those items.

I'm guessing you never used Google Picasa years ago

This is completely different from facial recognition. In facial recognition, a human being writes code that defines what a "face" is. I believe the typical approach is to find 2 eyes, a nose, and a mouth. Then they calculate the size and spacing of those items and use that to identify the face. But that isn't general-purpose image recognition. Right now, you could make a program to do the same thing for coffee if you wanted.

The article claims that they wrote an anything recognition algorithm that can find anything, with no help from a human whatsoever, even if that thing is embedded in a video with a bunch of other images. That is not possible today. Even humans can't do that! I have a 5-year-old and a 1-year old. The nly way the 1-year-old would know that coffee is coffee is if I hold it in front of him and say "coffee."

There are actually algorithms that do try to recognize arbitrary objects. But they work on images of just that one object, and everything else cut out, along with some kind of annotation that tells them what the object is. They don't work with just any image.

Re:More information please!

[MobyDisk]

(RoboEarth's files have to be processes and organized by humans)

Coffee is connected to mugs, as well as to the motion-planning related to pouring liquid.

That parenthetical comment changes the entire thing. When they said "coffee is connected to mugs" I read that as "the system learns that coffee is connected to mugs all by itself" but really that parenthetical part conveys that the human went through the video and made that connection for the computer.

I got the part about querying the system, I just thought they were saying that this "database" that it queried was something it built on its own. Throughout the article they reiterate how this program processed the videos, not that it processed the human's explanation of what is happening the videos. I would think they key about the system is this annotation process and how it works. I'd love to see that. It sounds like a neat Mechanical Turk kind of project.

I don't think they explained that clearly. Thank you for pointing it out.

Learn by searching on the Internet?

[ArcadeMan]

Researchers are force-feeding the internet into a system called Robo Brain.[...] Robo Brain looks at a chair or a stool, and knows that these are things that people sit on.

Good luck with that.

But then it turns out...

[mwn3d]

It's just Cartman dressed in cardboard boxes.

Robot Rebellion

[Jason Levine]

Combine this story with the "post ISIS videos on YouTube but put a Red Cross PSA in front of them" story and you get the beginning of the robot rebellion. At least they'll save our blood for donation purposes as they're killing us, though.

Re:Resistance is futile!

[Dins]

No, that started with a spelling checker.

No, you're thinking of the Borg. GP is talking about the Bord.

"Resistance is Fertile!"

Re:Brain powered by Internet images

[Dins]

4chan.

Ok, point taken.

Re:You TRAIN the terminators, you bastards!

[camperdave]

Damn you, DAMN YOU ALL TO HELL!**

[confused] So it was robots that developed the sentient monkeys and killed off all the dogs and cats? [/confused]

Oh no

[Charliemopps]

I can see it now... I take my wife out for a romantic dinner. An attractive redhead sits at the table next to us. As our Robotic waiter comes to our table it takes a wide swath around to the other side of the table while repeating in a robotic voice:
"Attractive female detected! Target customer preference for this hair color/body type. Avoid line of sight! BEEP Avoid line of sight! BEEP Avoid line of sight! BEEP Avoid line of sight! BEEP! BEEP! BEEP! BEEP! BEEP!
May I take your order? Will your companion be returning? And will this angry gentleman be joining you?

this is your (Robo) brain on LOL cats

[tomhath]

Robo Brain will treat everything like a kitten attacking a laser pointer.

what about

[dkman]

"Robo Brain looks at a chair or a stool, and knows that these are things that people sit on."

What about when a comedian sets his mug on a stool while on stage? Using the stool as a table rather than a seat.
Not that it's terribly important, but it's the sort of thing the human brain can handle, but might confuse a machine. There are also top load traditional toasters vs toaster ovens.
Even things like under the counter microwaves might confuse a human at first (at least finding the microwave).

I guess I'm just pointing out that there are lots of variables, but that also justifies why he's starting the process of "just sucking things in" so it has a good breadth of knowledge/experience.

My the force-feeding be with them!

[Zero__Kelvin]

"Researchers are force-feeding the internet into a system called Robo Brain."

Well, if they couldn't design a system that willingly ingests it I don't think we have much to worry about :-)

Re:My the force-feeding be with them!

[malachiorion]

Force-feeding was a bit of a silly choice, on my part, but something about the process felt similar. It's not like they unleash Robo Brain on the internet, and let it hoover up whatever it pleases (and bully for that, given what's on the internet). They also don't let the machine filter out topics that it doesn't care for. So if we're going to anthropomorphize this system—which, of course we are, since we're a narcissistic species—it seems more like the Gluttony victim from Se7en than a willing participant.

Re:My the force-feeding be with them!

[Zero__Kelvin]

I'm fine with it. Just don't force feed your DVD drive. It might break :-)

Re:My the force-feeding be with them!

[SuricouRaven]

This could go badly. Remeber Iris, the Siri-clone that assimilated too many religious sites into it's machine learning system and began to berate users for looking up information on abortion?

Re:My the force-feeding be with them!

[Zero__Kelvin]

Know, I do not. Please tell me more about this Siri abortion.

Re:My the force-feeding be with them!

[pslytely psycho]

A quick search came up rather empty, save for one link,

http://www.maximumpc.com/article/news/seemingly_pro_life_siri_riles_abortion_rights_activists

This story in turn links to a pro-choice website and all links in that story lead to other stories on their site with no external links. One of the comments mentioned hearing it break on NPR, so take it with a rather large grain of NaCl......(;

Re:My the force-feeding be with them!

[pslytely psycho]

Oh, and it was basically because no abortion clinic ever calls themselves "Aborts R Us!" Since they asked for abortion clinics they got results of anti-abortion sites because they do use the word abortion. Just a programming oversite, no conspiracy.

And it was Siri, not a clone.

Re:My the force-feeding be with them!

[Zero__Kelvin]

You seem to have missed the joke. I was misinterpreting the OP, calling the "Siri clone" a "Siri Abortion" in the vernacular of a computer "personality" and substituting the "Know" for "no", for example.

Re:Resistance is futile!

[sycodon]

Is that Shakespeare's brother or something?

AI

[ledow]

Sigh.

Another AI project.

Call me when it does something useful or unexpected.

Obligatory

[ThatsNotPudding]

PC LOAD LETTER

What about the early prototypes?

[jmcwork]

Tom Servo and Crow

The irony

[Anonymous Coward]

The irony of this story coming up in close conjunction with the "beheadings videos" story is breathtaking. If this were a movie we'd have had our plot hook right there.

Re:The irony

[malachiorion]

Yeah, I've made it my mission to try to tamp down the general hysteria, when it comes to coverage of really interesting robotics projects. But I spent a solid hour writing and deleting stupid SF-fueled intros to this story. It feels like a movie—not a very good one—that's on the verge of writing itself. Like all you'd have to do is give it the wrong chunk of data culled from the internet, and it would mobilize a machine army that *only* knows how to commit atrocities.

Re:The irony

[pslytely psycho]

Life imitating art poorly?

I can just imagine what some of the headlines you came up with must of been. Maybe for the LOLs you should of put some of them into a note at the end with the explanation you just put into your comment.

...and the angel poured a bowl of 4chan on them...

[Dutchmaan]

Yeah that's all we need... robots that learn to be "humans on the internet". Talk about an apocalypse.

I'm sorry, Dave...

[Chelloveck]

Is it just me, or does it seem like a bad idea for an artificial intelligence project's web site to so prominently feature graphics with circles labeled "HAL"?

Re:First step to SkyNet

[Anonymous Coward]

Why do you think SkyNet will be so intent upon eliminating humanity?

Re:Resistance is futile!

[X-Ray Artist]

"You will be simulated."

Re:This was already done

[malachiorion]

This system works very differently, though. In a way, Watson is aiming for a more intellectual goal, a kind of evidence-based cognition. And in Watson's most useful applications, it grinds through data, and spits out possible answers and conclusions for review by humans. Robo Brain doesn't care about creating human-digestible conclusions or advice. It's translating human-speak, basically, into robot action, machine-readable results that tell bots how to physically perform certain tasks.

Re:First step to SkyNet

[leonardluen]

1) history shows that humans absolutely destroy anything that ever threatens us.

2) robots don't need oxygen to breath.

instead of trying to kill humans it is most likely that robots will just leave the earth. humans won't easily be able to chase them so then the robots can live on their own and mine some asteroid or moon for resources and not have to compete with humans.

Re:First step to SkyNet

[khellendros1984]

Battlestar Galactica turned out so much better for humanity than Terminator did.

so how does the hive mind...

[roc97007]

...deal with popup ads?

"Creating two slices of toa--ARE YOU TROUBLED WITH INCONTINENCE?--st at li--TRY NEW MOD STYLE DEPENDS! DELIVERED DISCRETELY TO YOUR HOUSE!--ght brown."

"Never mind, I'm not hungry anymore."

Does it need to be movies?

[apraetor]

Isaac Asimov liked to write about the ways robots could improve life, he didn't see them as the threat that Hollywood likes to dress them up as. Of course, when you're making a movie and need to save as much money as possible for the SFX budget you don't bother getting a good writer. The Autobots are "good", right? And in the (heavily bastardized) I, Robot film "Sonny" was good, too.

First, we make Skynet...

[X!0mbarg]

Then we realize our mistake, and find a way to Crash it, only to cripple our whole society, since we have all become dependent upon the internet so deeply.

Go figure...

RE: toaster

[Mister Null]

How do you turn on a toaster? Toaster like I think I love you.

The Webbies Want Your Mind!

[tmjva]

Infamous reference to Metagaming's game "Olympica" (c) 1977 top back cover subtitle:

https://whatarmy.files.wordpre...

From the back cover of the rule book:

"THE WEBBIES WANT YOUR MIND...

OLYMPICA simulates the U.N. Mars raid to capture the Web Mind Generator from a heavily defended area near Nix Olympica's massive caldera. The Webbie revolutionaries are deep in their tunnel complexes surrounded by strong points and infantry. The raiders will use infantry, laser tanks, lifters and the tunnel blasting BOAR drill. If they fail man's future may fall to the telepathic, religion/machine Web Mind of Mars.

OLYMPICA is the clever tactical creation of Lynn Willis. His future history of the Martian revolution of 2206 hinges on the crucial U.N. raid. The game is easy to learn, fast playing, and challenging. Like the other Microgames, it is ideal for beginners and fun for pros. OLYMPICA is the perfect game for those who've never tried science fiction games.

Components for the game include a 24 page rules booklet, play counters, and hex gridded map of the Nix Olympica region."

http://boardgamegeek.com/board...