297 comments

[ 3.5 ms ] story [ 236 ms ] thread
Aha, Boston Dynamics has such wild repertoire of robots. Well done. Forget Big Dog, I like this robot a lot: https://www.youtube.com/watch?v=6b4ZZQkcNEo

Maps, Self-Driving car, robots and a high quality analytics team. On top of that, robotic team comprises of various reputed groups with talented well regarded individuals. I look forward to the future. Also, how much disparity will arise in quality of living between North America and various Asian countries, or even Europe.

When I saw the headline, I was thinking to myself "I'm sure I've heard of these guys before" and then I remembered the Sand Flea. Such a cool little guy! Thanks for the video, it fills me with awe every time I see it.
So is google going to start supplying the DoD when the DARPA projects graduate?
Good thing they have that first law that prevents them from doing evil.
But not preventing them from introducing "bugs". I wonder when will we see first robot genocide? In 10 years? 20?
Whenever it happens I am sure it will be fully democratic and completely lawful.
It probably already happened in the 80s (Nicaragua, Honduras or Panama) with a partially automated AC130 gunship.
What part of an AC130 is automated?
The smaller more asymmetric military operations are often used to test prototype weapons. My response about automated genocide is that it most likely already happened. And I can't think of a better testbed than an AC130 gunship for an automatic aiming system, the operators only see thermal blobs anyway. Not like this, http://www.wired.com/dangerroom/2007/10/robot-cannon-ki/

This would be trivial to automate https://archive.org/details/AC-130_Gunship_Ops_in_Afghanista...

Wow, I hadn't heard of that story, truly scary stuff.

This video was linked to in that article, http://www.wired.com/dangerroom/2007/10/video-robo-weap/ hopefully every soldier, programmer, and engineer working on these systems has to watch similar videos so that they know the importance of getting this stuff right.

(comment deleted)
I don't think that's necessarily a negative thing, just because it's Google. Think about how many university professors receive grants from DoD. How many of them are actually purely for science? They are all practical.
> grants from DoD ... They are all practical

You'd be surprised. I've done a couple of grants for projects that ended up being completely open to public and DoD didn't really give a shit about, despite providing some sweet funding for them.

But Google... that's a different story.

I think you missed my implication to his response. The way he put it is that Google is evil so when it moves to DoD and it must be evil, and I say that's not necessarily true just because Google being Google.
The article states that they said that while they intend to honor Boston Dynamic's current government contracts, Google is not interested in becoming a government contractor.
They're just interested in becoming the government.
Reading the headline the first words out my mouth were "holy shit"

Not only does this bring some serious engineering talent to Google's robotics initiative, this acquisition should create some interesting new interactions between Google and the government moving forward.

My first thoughts were along: "Nope" and "This is how the robopocalypse starts".

Google & BD are both researching machine learning, it only makes sense for Google to buy them.

Both will sure gain from each other talents and data. Robotics for the self-driving car, and a brain for the humanoïds. Aweso... well not so much.
I'm so glad to see that, that I am not the only person with such dual reaction - This is sooo awsomme tech! & ok, were it is going might not be so cool at the end...
Agreed - My first reaction was "Google == Cyberdyne Systems". I hope a lot of good comes out of this, but I cannot help but think of worst case scenarios. Then again, maybe I consume too much sci-fi movies and books.
It doesn't help that the robots produced by BD look a bit like the ones from "robots versus humans" movies. This acquisition feels a bit scary, but also very exciting. In reality, self-awareness is probably far enough away that we don't have much to worry about for now.
Robots won't be that scary when they are self aware. It is at the stage just _before_ self awareness where they are easily controlled and very powerful.
You only need to read the news about what the government is doing with drones abroad, or with military equipment at home to think at worse case scenarios.
I know it's been said, and at this point it's kind of cliche, but I feel like I'm living in a sci-fi movie more SNF more with each passing day. This is all very exciting stuff.
Seriously. These were precisely the first words out of my mouth as well.

Worth pointing out that Ray Kurzweil[1] is the director of engineering at google, Peter Norvig[2] is the director of research, and they recently snatched up Geoffrey Hinton[3] and several of his grad students.

Not sure if excited or terrified. The future of google is going to be very interesting...

[1] - http://en.wikipedia.org/wiki/Ray_Kurzweil [2] - http://en.wikipedia.org/wiki/Peter_Norvig [3] - http://en.wikipedia.org/wiki/Geoffrey_Hinton

"The director of" is misleading, it is a director (there are at least two higher positions: VP and SVP).
Thanks for spotting this and pointing it out.

I'll take your word for it about Kurzweil. The wiki also says "a director" - good catch.

Here's what google says about Norvig though: http://research.google.com/pubs/author205.html

FWIW, I don't think reporting to higher ups (e.g. VP, SVP) precludes someone from being "the director" if there are no other people in the department with this title / position.

Asimov added a 4th law:

  A robot may not harm humanity, or, by inaction, allow humanity to come to harm.
Is it just me, or is, "don't be evil" a little too close for comfort?
It's all right. If adopted, the second part has provision for protecting us from terrorists. I can relax now.
That would be the "Zeroth Law", zeroth because it supersedes all the others. A version of it first showed up in Asimov's short story "The Evitable Conflict":

https://en.wikipedia.org/wiki/The_Evitable_Conflict

https://en.wikipedia.org/wiki/Zeroth_Law_of_Robotics#Zeroth_...

If you give robots autonomy, they inevitably end up having to make moral decisions. For example, "Should I, an autonomous car, run over the elderly man or the girl with terminal cancer, those being the only two options?"

Asimov's laws (initially suggested by an editor, John W. Campbell) were a first pass at some principles for decision-making. Others have since devised more elaborate ones.

It's all still just words. No two humans could ever fully agree on what constitutes "harm", racist ones would even debate who is "human".

It's like everybody has the public methods what_is_harm() and what_is_a_human(), but how they're implemented matters, not the function signature. The laws of robots are just placeholders for something that never got written.

Words are tricky. What if robots decide that the best way to avoid harm coming to humans is to prevent any being born? So that means you have to rephrase it, into something like maximizing human happiness. Oh no, now they're injecting everyone with drugs and cloning humans by the planetfull. What now? Actual source or it didn't happen.. just creating an empty file called solution_to_the_problem.txt doesn't work, I tried it.

> Words are tricky.

Words are relatively easy. Ethics are not reducible to lexicography.

That's why trying to reduce ethics to a few words is.. tricky.

> Ethics are not reducible to lexicography.

Nor is anything else, which was kind of my whole point.

There are robot stories by Asimov that explore a lot of these questions. The Three Laws were a basis for storytelling more than anything.
They are placeholders to something written several billion times over in various ways in every human brain.

We don't learn ethics by reading it as a dry description; why do we expect to build other minds that do?

Because that would be consistent with how those minds "learn" everything else they "know".
Why? Dogs are autonomous as well, but live happily without such philosophical conflict.
The intelligence of the dog is limited. I guess it depends on whether we'll let such robots have limited intelligence, too, or we'll give them access to everything, and make them as smart as they can be (much smarter than us).
Or admit that there's no single axis we can use to measure intelligence. The computers will have virtually unlimited memory space that they can access perfectly and huge processing power but will have no idea how to perform basic actions that dogs understand instinctively.
Dogs are also stupid enough to jump into someone and break the person's leg while catching a frisbee, as happened to a friend of mine who had just gone off health insurance.
(comment deleted)
I feel like robotics in a few years is what big data is today. So what is it that one should be doing to get ready to jump into robotics in a few years?
You mean everyone and their grandmother will be a robotics expert?

Most of the people I run into today working on "big data" could still effectively fit the workloads in a single powerful MySQL box.

May be thats true, but it is irrelevant to my question.
arduino, lego mindstorms, linear algebra

Actually, a lot of the math for 3d game engines transfers directly to robotics. Matrix coordinate transformation to/from each joint to figure out where/how to place it.

Pragmatic question: I'm a mathematician by training, and currently a machine learning engineer at a large Silicon Valley company. How does one in their late twenties jump into really working robotics? Machine learning was something I could watch lectures on, read papers about, work the math out on paper, then spend years in Python/C++ getting a feel for across various problem domains. Physical robots don't seem as accessible.
Computer Vision is a good gateway.

Also, play with Arduino some.

Sensors. All robots need sensors and your math background will serve you well.
It sounds like we have the same background - I'd recommend getting into microcontroller programming with a platform like Arduino, and maybe look into some computer vision libraries (OpenCV is great).
This seems like a stretch. I am a big fan of robotics, and I get the Hard Problem appeal but it doesn't seem like a good use of Google's funds. It does make for some interesting speculation though.
Wouldn't you say they're looking for their "next business"?

Why not robots? Higher barrier to entry, untapped market. It is a place where a ton of capital can really shine.

So why wasn't Apple in there bidding against them?

I could actually see Apple doing a rather good job with robots. Making sleek, stylish efficient (servant|killing) machines...

On a more sober note. I see Apple as being likely to go after the home automation market. In fact I wouldn't be too surprised if they partnered up with an RV manufacturer to try out technology for the "iHome".

In truth, I don't know why no one else is bidding in. Perhaps it seems too risky or too outside the core business.

As for home automation, sure. I cannot see why these companies aren't trying to stake out the Internet of Things in the home more.

Well to some extent they are always looking for the next business. And discarding as 'useless' $100M/yr businesses because they didn't show up in the numbers when your annual revenue is $60+B a year.

Robotics isn't a 'smart gap' market (where if only smart people would focus on it they could dominate it) there are lots of brilliant people already toiling away there. It isn't an advertising market.

In Asimov's vision of the future the robotics company that developed the positronic brain is the dominant corporation in the galaxy, and perhaps Larry is working on a positronic brain equivalent. But who knows. Maybe they are afraid of Bezo's drone army, or Musk's rockets. But why Boston Dynamics and not iRobot? Why a fragment of Willow Garage and not Neato? Why a self driving car but not Tesla? Lots of questions. Its a fun puzzle to try to solve.

I remember people saying the same thing about the YouTube acquisition. At the time, it was very difficult to see how YouTube could possibly be valuable to Google. Everyone knew that videos were a solved problem: use Ares to download them, and DivX to watch them. And the purchase price was something like 800 million. 800 million?! What a sucker Google was!

What a bargain 800 million was in hindsight. And what a monumental effect the acquisition had.

It's hard to predict whether this acquisition could be on the same scale. But considering the space, it may very well be.

I believe it was 800 million per founder.
It was double that, actually: $1.65 billion. But the point remains valid, that was some incredible foresight on Google's part, even if everyone at the time thought it was a ridiculous amount.
No you can very obviously sell advertizing on you tube which fits with Googles business
Why do you say that? To me it seems like a completely logical extension of the research work Google has been doing with autonomous cars and so on.
Robotics is multifaceted and difficult in all of them. Boston Dynamics is probably the best place in the world for motion dynamics, gaits, and walking mechanics.

I'm amazed, but I suppose I shouldn't be - Google made a great choice.

HN gets a lot of post about acquisitions and whatnot, but I feel this is probably one of the most important in recent memory.

The implications of taking this talent into consumer robotics (or augmenting their presence in the defense community) will ripple through the tech world.

Yup. Absolutely agreed. I really didn't expect this.
Couldn't agree more - I would rate this as the second most important news item this year, after the NSA revelations.
It's amazing because it's so clearly, wildly complementary: Google becomes the prefrontal cortex to Boston Dynamics' robotic central nervous systems. I can't see this pairing as anything other than the beginning of the world of sci-fi robotics.
Which often doesn't end well.
It "often doesn't end well"? It hasn't ever happened!

Stories about it often end in a way that makes for a good story. That's not saying the same thing at all.

There's a basis in reality for those. many powerful technologies have been used in the fast for "evil".
I would go further and say "all" rather than "many". However there is the real question of: on the balance - does this help more than it hurts.

Additionally just the very act of being concerned about misuses helps mitigate them.

And many have been used for good. The technology itself doesn't have any will of it's own. People just use it. And people do both good and evil.
Until it does have a will of its own.
Not likely. We're software developers, not soulware developers.
(comment deleted)
> Google becomes the prefrontal cortex to Boston Dynamics' robotic central nervous systems.

Or to put it another way, Google's computers plan the strategy, and Boston Dynamics' controllers take care of the tactics.

Google is building one hell of a robotics patent portfolio.
I 2nd that. Given this is the eighth company to be acquired by Google in the last 6 months this will be interesting to see where Google takes this.
Google acquired something else? Shocking!
So does this mean Google will possibly move out of the search and advertising business within 5-10-15 years due to low margins and what have you?
As funny as that is, their core business has extremely high margins.

If all Google did is run their search engine and ad business, they'd have a 50%+ net income margin.

For fiscal 2010, they printed $8.5b in net income on $29.5b in sales (28.8%). That's with all the other things they do that don't have good margins (or lose money) eating up their core margin.

Very few things have better margins than software (Microsoft once hit a 41% net margin, for fiscal 2000; nothing better than software monopoly margins).

>Very few things have better margins than software

Ad business and search engine are not selling software.

I didn't say selling software.

I said very few things have higher margins than software. Google's search engine and ad platforms are software systems, and Google is a software company first and foremost; they always have been, and always will be.

The extrapolated principle is obvious here. Google writes it once, and scales it toward infinity. The same principle worked for Microsoft with Windows; the same principle is at work for Salesforce.com or Facebook or Oracle, regardless of what we call it ("cloud"). This principle is what makes most software so profitable, per user there's usually nothing to uniquely manufacture or create. With scale margins properly skyrocket.

So whether we're talking about software search, or software retail, or an ad business built on software (eg Google's hyper automated ad system) or software services, the difference doesn't matter much.

My childhood dream was to buid robots. I left that dream when I realized doing robotics mostly meant industrial arms and pick-and-place machines. Seeing developments over the last few years is inspiring, to say the least. Maybe it's time to consider coming back in.
No, it's not. If you want to build robots at home you will have to accept they will be vacuum cleaner level robots. For "real" robotics we need better software for controlling robot body. And that requires artificial brain. And we don't know yet how to build it, because we don't know how our brain works in enough detail yet.
http://www.youtube.com/user/BostonDynamics

Here is a link to Boston Dynamics' YouTube channel in case any of you want to see some of the amazing robots they've built over the years.

Boston Dynamics' robot knowledge (and patents) combined with Google ability to create awesome software, such as their self-driving car, will be a great combination.

I'm not sure where going is going with all of these recent acquisitions. Home and office robots to make our lives easier? What in the heck are they up to at Google X?

The US DoD probably doesn't like this too much, considering that Google stated that they do not want to step into the government contracting sector. Boston Dynamics seemed like one of the most promising contractors for DARPA in the robotics industry.

IMO it's only a matter of time before Google steps into the cyborg industry. It's almost scary to think about, but it's inevitable and at least Google has some experience in dealing with products and services that are controversial, such as data privacy.

I was about to rip on Apple for not doing anything innovative or even remotely interesting in years, but it's sad to see all of that money be wasted on rounded corners, perfecting fonts, and scrolling animations.

Apple had the same issue re government contracting when the bought PA-semi. The honored existing contracts and made no new ones afaik.
When Google starts investing/buying into companies doing bio-compatible materials, this is when we'll see they are moving to cybernetics. It's the basic research that is started before big advances in the field appear.
> it's sad to see all of that money be wasted on rounded corners, perfecting fonts, and scrolling animations.

It may seem like a waste from a technocratic point of view, but it's probably worth it just for popularizing and normalizing technology. If we didn't have that bump in style and usability in smartphones/tablets, people would be less comfortable with carrying around the internet in their pocket, there'd be less demand for tech like 4G, etc. Now people's minds are a bit more open to pervasive tech, and it may help adoption of whatever tech BD+Google develops.

I think you've hit on something important that I didn't realize at first; patents. Those BD patents have got to be worth a lot of time & money to Google. When I first read that Google had acquired BD I was shocked and a little scared, but now it makes a lot of sense.
Why does that alone mean it makes more sense?
The video for the robot running is just Wow. Really amazing.

I thought I had seen all this before. Then it clicked. Only in the movies.

To see it real like this, really amazing. All you need is a nice dog custome over it.

We should all be very concerned about this. Ray Kurzweil, who is a Google executive, hopes that one day machines will achieve 'singularity'. If this ever happens, and it just might, humanity will be replaced by machines. The Matrix would literally become reality.
Are you just stringing words together by rolling dice here?

If the planets align with Saggittarius, and they just might, crystal chakra energy will release thetans and quantum bio vibrations will slaughter thousands. Sylvia Browne's dreams will literally become reality.

1. In the Matrix, humans were not replaced by machines.

2. The Matrix was not a post-singularity world

3. Ray Kurzweil hopes humans will achieve {some kind of singularity}( it's not clearly defined) for the benefit of humans.

4. This news is about physical moving robots, it's the less relevant part. If you fear self improving AI then Google's usual software work should be what you panic about.

Are you Ray Kurzweil's personal assistant? If not, then I'm amazed you can read his mind.

You're naive to think that the combination of Google and the work of a military robotics contractor is benign.

I cite Ray Kurzweil's website, a questions and answers session written by Ray Kurzweil:

Intelligent nanorobots will be deeply integrated in our bodies, our brains, and our environment, overcoming pollution and poverty, providing vastly extended longevity, full-immersion virtual reality incorporating all of the senses (like The Matrix), “experience beaming” (like “Being John Malkovich”), and vastly enhanced human intelligence. - http://www.kurzweilai.net/singularity-q-a

This is not "humans being replaced by machines". All these claimed things are beneficial for people. You claim he actually wants something different to what he says, why do you think that?

You're naive to think that the combination of Google and the work of a military robotics contractor is benign

This is just throwing an insult at me. You aren't addressing any of the points I raised about your disconnected reasoning and flawed comparisons.

Does "don't be evil" mean "don't build robots that the government will use to kill or harm people"?

It would be nice to see some clarification on this from Google management. There's only so far you can go in high end robotics research before the pentagon comes a-knocking with wheelbarrows of freshly printed money.

in b4 "skynet"

Google's "Don't be evil" carries no more weight than your local mechanic's "The best car shop in town!". It's cheap branding.
You and I know this, but I imagine there are many Googlers who take it to heart, and the cognitive dissonance involved in their building war machines may be useful to our ends of reducing American-perpetrated violence.
How about we wait until Google actually starts building such weapons before we start accusing them of actually wanting to do so.
What makes you think we'd ever be told?
How about we wait until someone actually accuses Google of that before we accuse them of accusing Google of that.
It is flatly impossible to build such devices and prevent them from being weaponized.
But what is more evil - that your coworker works on encryption technology that is used to pilot drones, or that your coworker didn't work on that technology, and thus in order to foil a terrorist plot, ground missions were required that ended up costing the lives of a strict superset of people? Don't get me wrong - I don't condone the lengths to which our government uses war technology, but as an engineer, the probability distribution over {my technology will be evil by some metric, my technology will promote good in this world by that metric} is by no means certain.
The former is more evil. The potential disasters associated with the latter are a significant deterrent to doing it unless it really matters.
(Shrug) I don't care what you're working on, it can be used for evil. It is never rational to fixate on tools and objects. Worry more about actions.
It seems your "foil a terrorist plot" is some sort of preemptive policing. Why are we doing that, again? And who are these "terrorists"?
Puzzling that you would only be concerned with violence from one source.
It used to mean something... I wonder for how long
What if the robots are only used to kill evil people?
> What if the robots are only used to kill evil people?

Evil by whose definition is usually the question.

"Calamitous intent" evil or just "meh" evil?
David Bowie, Klaus Nomi, and Iggy Pop insist meh evil is plenty.
Who cares? Prevention is more efficient than remediation anyways.
Then we will reclassify everyone we want to kill as 'evil', just like the administration reclassified any adult male in a combat zone as an 'enemy combatant'.
FTA:

"Dr. Raibert has said in the past that he does not consider his company to be a military contractor — it is merely trying to advance robotics technology. Google executives said the company would honor existing military contracts, but that it did not plan to move toward becoming a military contractor on its own."

If one is working on military contracts, then one is a military contractor.

If they are owned by Google and Google honors those contracts, then this is the day that Google publicly became a military contractor.

Or they will direct their talents not to the military any more (after honoring existing contracts) to other more non-violent applications.
One can only hope.

That was my fantasy upon reading the headline: "Google acquires and renders harmless army of kill-bots"

Sadly, I'm afraid it was a dream.

If they didn't buy the company, they would have to honor the contracts anyways. If they break the contract presumably they will have to face consequences, possibly paying money to DARPA so some other company can do it.

Besides not everyone believes DARPA is "evil".

There are two sides to this story. The one I consider almost insignificant is that by honouring existing contracts, they have become a defence contractor in the literal sense.

The more important one is that by not doing new work for the military, they have just remove one of the hottest companies from the defence market.

I wonder whether the US government has ways of withholding regulatory approval of this deal because of this.

If the robots are used for military logistics (moving stuff around, but not directly engaging in combat) would you consider them to be "robots that the government will use to kill or harm people"?
It would be better, but it's still enabling the US military. Even from the perspective of US allies and beneficiaries it can be distasteful to deal with companies that directly enable the DoD.

My own politics aside, there's a pragmatic concern: the NSA is part of the DoD. If Google is fulfilling military contracts, then it's going to be hard to do that and to push back hard against NSA.

Unless they're only doing it so as not to be sued for breach of contract. If they're not chasing future grants from DARPA then there's no incentive at all to play nice on other fronts.
Dito for AI and ML - I know some who moved out of AI research for this very reason.
There is no doubt in my mind: this is an acquisition that will change the world
Holy shit.. this is huge.. This would give google huge advantage in robotics. What scare me that google is creeping into critical areas of modern life. And they usually dominate every field they enter.
Like social networking? Smartphone manufacturing? Ebooks?

There are plenty of fields that Google has entered that they haven't dominated.

You'll be surprised, but Google dominates social networking. Don't forget about YT.
YouTube isn't really a social network.
Why not? It has all properties of a social network. Is Instagram a social network? Pinterest? Same principles, different look. It doesn't have to have "friends" to be a social network, you know.
Most of YouTube doesn't work as a social network, and most YouTube users don't use it for social networking.

I don't find the argument from features compelling because so many different services have the same features. On a broad definition, there are millions of social networking sites, but on a broad definition, the term becomes so close to meaningless that it doesn't tell you anything useful.

anyone know what the price that Google paid? I know they haven't disclosed it, but any guesstimates?