At Google X, a Top-Secret Lab Dreaming Up the Future | Hacker News Reader

At Google X, a Top-Secret Lab Dreaming Up the Future

nytimes.com

244 pointsssclafani95 comments

One nitpicky-type thing for me in the article is saying that Sebastian Thrun invented the first driverless car. Yes, he was the head of the Standford team who won in 2005, but he was on the CMU team the competition before where Red Whitaker was the team lead. The fact that these teams were quite huge also prevents someone from saying that one person invented the driverless car. Being team lead on a team who wins the second competition certainly doesn't mean that person invented the tech. I know I'm being anal about it, but there were a lot of people who competed in the DGC, and it is quite annoying to hear people mention his name in that regard. Having talked to Sebastian a few times during the competition, I would think he thinks the same way.

It's important to put some face to tech/science breakthroughs otherwise the general public doesn't relate. Also, in that vacuum you'll get politicians gaining credit. Kennedy gets a lot for credit the moon landings because their wasn't a clear face to put it to, while in the manhattan project at least oppenheimer gets credit.

Here's a collection of quotes from the triumvirate about artificial intelligence.

http://ignoranceisfutile.wordpress.com/2008/09/13/google-fou...

(Except, ignore the Moffett field thing. That has nothing to do with AI research as far as I know.)

Anyway, any Googler will confirm that Sergey Brin, in particular, keeps returning to this idea: that Google's ultimate destiny is to realize Strong AI.

I'm not saying they've achieved anything. But I would not be surprised if they were putting substantial resources into it. It's even directly related to their bottom line. It doesn't have to end in HAL 9000. They may come up with some technology that is just more insightful about search queries. But even that would be worth billions and billions.

You know, when you see all the quotes in one place like that, it does make their goals pretty obvious.

> I'm not saying they've achieved anything.

Clearly you're not aware of the CADIE incident a couple of years ago.

http://www.google.com/intl/en_us/landing/cadie/

Be sure to check the blog.

(By far my favourite piece of Google whimsy, for the detail and number of teams involved) (You're right, muro)

sorry - removed my previous comment in the meantime. It was obvious it's a joke, so I figured my comment is just noise.

And now I can replace it with even more words :)

It took me much longer than I'm comfortable admitting to realize that this was an April Fool's joke.

I've often wondered whether an AI system like OpenCyc could derive its knowledge base from web content, rather than relying on rules and assertions written by the AI's human caretakers. If anyone could do that, it would be Google.

"It’s a place where your refrigerator could be connected to the Internet, so it could order groceries when they ran low. Your dinner plate could post to a social network what you’re eating. [...] These are just a few of the dreams being chased at Google X, the clandestine lab where Google is tackling a list of 100 shoot-for-the-stars ideas. "

Those are shoot-for-the-stars ideas? I hope those are at the lowest end of the scale. They seem pretty boring to me.

Ironically enough, the refrigerator-ordering-ingredients thing was one of the talking points of Microsoft Home [1] when I worked there eight years ago...

[1] http://www.engadget.com/2011/05/03/microsofts-home-of-the-fu...

I suppose the space elevators would be on the higher end of the scale.

Sometimes high, sometimes low. It's up and down.

Haha, I walked right into that one.

Intelligent refrigerators based on computer vision (and not rfid) is actually a shoot for the stars kind of idea. It would also revolutionize retailers, distributors, and manufacturers.

I agree, I think we underestimate how awesome these technologies could be if brought together into a coherent service.

I would like my kitchen to tell me what I can cook based on what's on hand, and then give me the recipe for it. And say if I decide on a menu for the next week have it auto order what I need from amazon and give me a list of what else I need to buy, and where its on sale. Sign me up.

I've been hearing talk of self-restocking fridges for at least 10 years. It's the ultimate go-to cliche for the "connected devices" hype, that's now called the "internet of things."

It's the modern-day version of the Honeywell H316 Kitchen Computer: http://jdh.oxfordjournals.org/content/23/2/163/F9.large.jpg

apropos: http://valerieaurora.org/kitchen.html

I was thinking more of this: http://www.archive.org/details/Designfo1956

I think Google is perfect company to make those future products that have been 5 years away for the last 15 years. Intelligent fridge probably comes to mind first, but there are a lot. Some of those ideas just need clever people to work on it (check), resources for research (check) and ability to produce it in scalable way (check). Maybe we'll finally see an affordable house robot that helps more than disturbs.

have you considered that we don't have an internet fridge because not enough people want it?

Not unless it could order food for the dog, too, while it was at it. Oh, that didn't work out?

My wife would like something that cleans up after the kids (safely, etc). I would like something that turns things off when people leave, but I suppose smart switches would do the job cheaper than a robot. Not as cool, though, especially if it were like an inexpensive little "monkey bot" about 3 feet tall :-)

http://fuckyeahinternetfridge.tumblr.com

You seem to have cut off the quote right before the shoot-for-the-stars ideas begin:

"... Your robot could go to the office while you stay home in your pajamas. And you could, perhaps, take an elevator to outer space."

a telepresence robot you say? looks like they've been scooped by bell labs: http://www.itnews.com.au/News/278794,bell-labs-builds-telepr... (among MANY others)

Perhaps these latter ideas are simply large/hard problems, not "ideas so strange no one has even ever considered them before".

However, if you were a hush-hush R&D lab investigating some ideas that were so far out there no one has ever heard of them, would you really tell them to the NYT so they can include them in their profile of your lab? I doubt it.

I'd expect anything interesting to have been covered by an NDA.

Those ideas are also pure speculation by the author. The closest she comes to backing any of them up was when she said that the founders like space elevators.

The "Internet connected refrigerator that manages food for you" has been an example of this sort of thinking for at least 10 years, and it's no closer to being something that anyone actually wants.

Except advertisers. Remember, it's still Google.

Seeing Google AdWords on my fridge trying to sell me stuff or alternates to things I am almost out of is going to be an experience.

I was thinking of Google watching the advertising of food products and tracking how much of that results in purchases that end up in your fridge.

Advertisers would give their left arm to be able to close the loop from ad to store to home. I know the "club cards" can kind of do that now, but the stores have no idea what ads you saw before walking in (other than the coupons in your hand).

John Ringo has a book set called the Troy Rising series where multiple characters get implants connected to the network. The description of bill collecting and advertising is hilarious, great for advertisers, and a strong reason not to get said implants.

If Google is willing to subsidize a free refrigerator for me so they can serve ads on it, I will consider it. :-)

There have been times in my life where I've thought that I've had a "genius" idea and every cell in my body has said "Don't tell anyone. Hoard it. Cultivate it. And you'll be a revolutionary!"

And then I end up sharing. And there never ceases to be one or two or four perspectives shared that I never thought of before. And in the end, I think the idea flourishes because I was open and shared and iterated and shared those iterations.

Do these "top secret" labs ever make the big cognitive leaps that we hope for?

PARC gave the world Graphical User Interfaces, object oriented programming, ethernet, laser printers, precursors to PostScript, and a few other major breakthroughs.

It wasn't necessarily a "secret lab", but its distance from XEROX management on the east coast made it effectively an under the radar operation (to their detriment, of course)

Yeah not really. If you look at Sketchpad and the Mother Of All Demos you see the foundations and inventions of modern user interfaces, and C++ style objects were invented first in Simula.

So PARC was mostly just the industrialization of already known ideas and computer interactions.

Most of these only really took off once PARC shared them with other people and companies.

I'd assume that it depends on the "top secret" -ness of the lab. In a large enough top secret lab, there is a diverse enough set of perspectives to still provide the sharing culture that ideas need to flourish.

In a right-sized enough top secret lab, you can keep the "that's just too crazy, it'll never work" attitude out, while affording the isolation from real-world concerns necessary for "too crazy" to become "just crazy enough".

> Do these "top secret" labs ever make the big cognitive leaps that we hope for?

Clifford Cocks discovered public key cryptography years before RSA, while he was working for Government Communications Head Quarter (GCHQ) which is an environment with (especially at the time he did it) a lot of secrecy, both from workers to the outside world but also amongst workers internally.

Who knows what else they've discovered that has not yet been released?

Big cognitive leaps? Judging by the first paragraph of the article: "the future is being imagined...It’s a place where your refrigerator could be connected to the Internet, so it could order groceries when they ran low."

the answer is "no". Of course that could just reflect a lack of imagination in NYT journalism, but really? I remember writing about that idea in early 1995 - about the same time that there was a fashion for universities to wire their Coke machines to the net, and about 5 years after Cambridge University frst wired its coffee pot to a network: http://www.cl.cam.ac.uk/coffee/coffee.html

> Of course that could just reflect a lack of imagination in NYT journalism

Don't underestimate that factor. Journalists, even really good ones, have no clue about technology at all.

It’s a place where your refrigerator could be connected to the Internet, so it could order groceries when they ran low. Your dinner plate could post to a social network what you’re eating.

The stuff that could have been invented 5-10 years ago but was not interesting enough to warrant the effort

I've been hearing about fridges-that-order-groceries for at least ten years. It's one of those dumb ideas that just keeps coming up, despite the fact that there's no demand for it. Basically, I don't trust my fridge with my credit card... and besides, I don't want to order groceries by delivery anyway, particularly not the kind of perishable goods that I'd be keeping in my refrigerator.

I suppose I wouldn't mind having a fridge which knew what was in it, so I could check what I needed to pick up on my way home before I left the office. But I doubt the system could be made smart enough that it actually saved me more mental effort than it consumed. The low-tech version is to simply put cameras in your fridge so that you can check its contents remotely... but even if I had this set up I doubt I'd use it all that often.

Maybe I'm being overly harsh on the smart-fridge idea. But I think it's representative of a huge class of dumb ideas: let's use fancy technology to save some minor effort for the consumer, at the cost of the consumer now having to go to the effort of acquiring, learning and maintaining yet another fancy piece of new technology. Screw it, I want my microwave with dials back.

What about commercial fridges like restaurants? Or work place refrigerators? I can see the set-it-and-forget-it feature working well in these scenarios.

Although I share your sentiment, there is contrary evidence.

Consider this essay:

http://www.taquitos.net/chipworld/articles/snaxpo-learned.sh...

> Convenience in foods will continue. Anything that takes the steps out of a process will become successful. Witness the "chip/dip" combo packs that Frito Lay is now selling. Open the package, you have chips and dip right there. No need to go to the fridge! At the testing area, Jays demonstrated a combo pack that had mini pretzels and honey mustard that was extremely tasty. Other examples included a pre-cooked baked potato that includes the sour cream, butter and salt and pepper inside the package. All you have to do is nuke the potato for a few minutes and you've got instant lunch.

I'm really not sure how fast food product ideas contradict my point about adding unnecessary complications to home electronics, but... damn, a one-step microwaveable baked potato sounds great.

Or, more realistically: not economically viable enough to warrant dropping millions of dollars into R&D that may never be profitable. Google's piles of cash change this.

OR profitable but not as profitable as the things the companies with those resources were working on during those years.

While I might enjoy the convenience, the way I stock my fridge just isn't a constant datapoint. If there are no eggs in my fridge, it might be because I don't want any eggs for a while, not because I didn't think of buying any.

To solve this, you'd need human interaction etc. etc. and you soon arrive at a point where you just stock the frigging thing yourself anyhow. For a one-household effort, this really doesn't warrant the effort.

A different story could be approaching it as some sort of communal effort, say the combined amount of fridges in one multi-story house or student dorm. If you could plan your meals /assisted/ by your fridge and it could tell you that the guy next door has eggs that are free, you wouldn't need to buy a whole carton just because you want to bake a cake, thus possibly reducing waste.

In the end, this puts us back to the old privacy implications versus efficiency problem that remains largely unsolved.

I'd prefer ones that warn you when they get too warm via text message. Might not do anything if the power went out, but if the door was slightly ajar, it'd be a lot better to get an alert than to have all your ice cream melt and your meat thaw.

Oh, and they should start putting windows in the doors so that people don't stand there with the door open, thinking about what to eat.

I'm pretty sure all of that is trivial for a good engineer to wire up these days. Though if we ever had a 'smart' enough fridge, I could also see people integrating things with the various social dieting apps or whatever.

I am hacker and a geek, but nowadays I am more and more interested in marketing. I am wondering in what part the existence of such a lab and the emphasis in their communication on A.I. is a conscious marketing effort. I can imagine that even if nothing from that lab will make direct profit for them, and even if their search engine will never be even close to strong A.I. they can communicate an excellent marketing message: when you use Google you use strong A.I. (or at least almost). If we think about it this way the secretiveness can also make the story more exciting from a marketing perspective. Also if this is their main motivation then they will create more 'spectacular' prototypes rather than just honestly trying to solve deep problems.

Being more and more interested in marketing in the way you've described means to me that you are cultivating a more and more robust, healthy skepticism of the messaging that comes from our media/corporate centers.

"Because Google X is a breeding ground for big bets that could turn into colossal failures or Google’s next big business — and it could take years to figure out which — just the idea of these experiments terrifies some shareholders and analysts.

“These moon-shot projects are a very Google-y thing for them to do,” said Colin W. Gillis, an analyst at BGC Partners. “People don’t love it but they tolerate it because their core search business is firing away.” "

Is it just me or does this represent the worst of Wall Street short-term thinking? I mean, the first sentence could really be read as an indictment of venture capital as a whole.

I think what Google's best at is not so much innovation as taking existing innovations typically reserved for Dept. of Defense/B2B uses and bringing it to consumers for free.

We already know the DoD already has tech that is years ahead. Google is one of the few companies who can not only afford to bring crazy breakthroughs like Google Street but to do so in a financially profitable or feasible manner.

I disagree that dod has the best tech already in "top secret". The best tech is often dod sponsored, but is made public by the university researchers who usually create it. And it's not costly, you can go read it and use it, often the stuff is released as open source software on the researchers' website.

I think we'd be surprised to see some of the things that might be in the bowels of, say, Sandia.

I used to think they were all special genius places, until i see that it's not their people winning publicly open challenges like netflix, darpa grand challenge, x prize, or goldfarb.

People at those "special genius places" can't participate in any of the things you mentioned. They get paid for other, much more important jobs. Darpa grand challenge wasn't due to the government not being able to do something, it was due to it being much cheaper to just pay someone to do it for you. In the end, lots of teams competed and only a few got any money out of it, the rest of us ended up just volunteering for the government for a while in the hopes of making a lot of money.

I've worked with some GOTS software that came from Sandia and some that came from Lawrence Livermore. It's honestly kind of crappy.

here's a good test: the ROSE framework is developed by LLNL. it's basically an LLVM also-ran. compare its code quality, its use experience, and its overall worth (the code for both is open source!)

I don't believe that's true. The military's been trying to develop self driving vehicles for decades and one of their teams did compete in the darpa challenge and lost. And I don't see self-driving vehicles being unimportant now that the wars are fought against civilian combatants. The military is (or should be) desperate for them, they would avoid almost all military deaths that occur now (ied's blowing up humvees transporting things from the greenzone in baghdad for example).

What are the more important stuff that government scientists deal with that academic/industrial scientists are not up to? Don't believe the hype the military-industrial complex sells about itself.

No one from the government competed in the Darpa Grand Challenge. It was all private companies or universities. I was there, DARPA sponsored it and oversaw it, that's all. Check the wiki link below and tell me which team listed was from the DoD?

They wanted self-driving cars for military transport into dangerous locations. The first DGC there were only a few competitors and none of us did very well. In 2005 there were quite a few (listed here http://en.wikipedia.org/wiki/DARPA_Grand_Challenge_(2005)) with lots of teams making it through to the desert final stage.

I was part of the military-industrial complex, and yes there is a lot of wasteful B.S. only there to make more things to make more money, but there is stuff there that is not being done at all (or at least not well) outside of the government. Those things are the things that are labelled Secret or Top Secret in most cases. Like all governments, if the US government has an advantage in any little way, it likes to keep it that way.

Wouldn't you consider the mitre entry basically a government entry (it's a non-profit funded by the government).

No, because they are all civilians who don't even carry GS rankings. Mitre is just a government contractor like everyone else in that space, but they are different because they are a not-for-profit corporation. They serve contracts for the government, but so do people like Lockheed and Northrop. Like those other two, the government couldn't just shut down Mitre.

"But is made public by the university researchers who usually create it" <-- yeah, the stuff you know about is made public, and the stuff you don't.... isn't. That's what secret means. I think I heard that 50% of the research in Neuroscience labs today is DoD sponsored and never sees the light of day. Mind-hacking anyone?

Tech that could be applied to civilians and made money on doesn't stay hidden for long - this is one place where the "greedy" mba types are the good guys.

I take it you think there are MBA types that wander around NSA headquarters, grab the dragonfly-size spy robots and bring them to market?

yeah, exactly.

there has to be a management structure even at nsa, and management types are "networkers" who'll talk to people with money to invest.

NSA gets their money whether you like it or not. They certainly have people in charge who talk to the legislative branch, but they aren't MBA/marketing people. The are usually ex-generals. Other agencies might seek outside contractors, but the NSA is one of the few who do most of their work themselves. I didn't work there, but that was my understanding at the time I worked in other parts of the government.

Doesn't matter about their backgrounds, mba or ex-general, people who rise to executive levels rub shoulders with investors at cocktail parties and deals happen.

Seriously, what are you talking about? What investors? You have proven you don't know what you are talking about in this thread, but I am curious what you think is true? Who are the heads of the NSA meeting with who would be considered investors? The NSA meets with the Armed Services board and the Senate intelligence board, and that's about it. I get it, you think the government is b.s., but groups like the NSA don't seek outside investment.

No need to get rude, I'm not some teenage troll guy, I'm here for intelligent discussion. I don't seen where I've "proven" myself wrong.

This thread is not about investment into nsa, but commercialization of nsa (or other military/government org) technology. My point was that regardless of nsa'a official stance of secrecy, tech gets "leaked" out into the commercial sphere. You can't stop a good thing from spreading, no matter how hard you try.

Are nuclear bombs good? Do the heads of our national security agencies take technologies that could potentially kill the whole world's population and, because they, like every other executive in the world, are an MBA-types, sell it to whoever they happen to meet at a cocktail party? You may not be a teenager, but I can't say that you aren't a troll.

Nuclear tech, in the form of reactors has been commercialized for quite some time if you haven't noticed. And no, I'm not a troll, just because someone takes a contrary view they are not automatically a troll.

There aren't profit-oriented 'managers', there are 'administrators'.

Having recently ended four years working in a neuroscience department at Harvard Medical School...

No.

We're not aware of what they have that's considered classified. Back in the 70s, a former prof of mine was doing his phd. He was trying to figure stuff out and couldn't. Through one turn or another, he said the DoD told him, "OK, kid, come over here, we'll help yo out." He got classified access to see stuff that blew his mind and helped him get going.

The point is that back then, at least one guy's experience was that the military helped the university researchers, not the other way around. When I look at some of the capabilities in military hardware today, I wouldn't doubt that the DoD is still way up there. Certainly, a lot of medical innovations (as one example) are still made on the battlefield first, even today.

How would you know if it is top secret? Most top secret tech never gets seen outside of the top secret circles. Unless you have a TS/SCI clearance, the best you can do is speculate on what exists now. Obviously not all fields are covered in the government, but things that are very important to them (signal intelligence, networking, logistical management, satellite communication) are all very well ahead of the consumer/public space in my experiences.

I've got doubts about their logistical and abilities, now that they increasingly use private companies to get their stuff around.

> We already know the DoD already has tech that is years ahead.

You'd be amazed at the age of a lot of the tech involved in military technology.

Once a design gets approved there's very little opportunity to change it, and so years later people are having to try and source obsolete ICs or transistors. And if the designers are careless things get obsolete really quickly; I know a design that used a Bourns resistor network with a 1% tolerance. This package was once offered by Bourns, but the standard was 2% tolerance, and the 1% part was rare and thus very expensive. See the number of brokers offering new-old stock or obsolete components at expensive prices. (Obligatory "YOUR TAX DOLLARS AT WORK" quote here.)

I can't help but feel the article starts by missing the point. Connecting random devices to the internet is only a good thing when you get something neat out. Sticking a 3G modem in random devices & a bit of hardware is low-hanging fruit.

Awesome to see both Andrew Ng (ml-class.com) and Sebastian Thrun (ai-class.com) involved in these projects.

I wonder if they will ever work on a flawless mobile OS that has tons of great apps and great offline support for Google Docs.

Nothing is ever flawless. If you focus on refining one technology too long, the world will change around you and it'll become obsolete. Imagine Microsoft patching Windows 95 for 15 years.

X for Xerox? When can I expect futuristic visualizations from Google Design?

Maybe it's somehow an important proof-of-concept AI challenge, but restocking my fridge does not make the list of things I desperately need better technology for.

The driverless cars are old news and space elevators are a niche at best. Whatever makes Google X interesting must truly be top secret.

> space elevators are a niche at best

What does that even mean?

Only Willy Wonka and Richard Branson would buy one?

internet search is just a niche.

this is where Google Labs went when it was "cut"

not sure if you're being serious, but no, google labs was a bunch of experimental projects run by normal google software engineers. this is an entirely separate thing.

partially a joke and partially just hopeful, yes. to explain myself, Google seems to have attempted to appease investors by focusing on their core money making products, while still maintaining their commitment to innovation with spinoffs like this. so it's the best of both worlds for Google and for us in this case. now if we could only find Google Code Search, Google Labs, and the old design of Google Reader somewhere in there too....

The spaghetti cannon continues to boom.

Too bad I can't watch the spaghetti cannon at work during lunch. I know it's silly, but silly is fun.

Stuff like YouTube (as well as Netflix) is helping to kill broadcast TV (alright, reality TV and other low value cruft is also helping to kill broadcast, as well as fixed show schedules). Long term, internet video is more democratic, as well as flexible and fun. Perhaps stuff like YouTube will have more effect on society that internet-fridges and AI cars?

Reply on news.ycombinator.com