Google X: Waiting for a Moonshot | Hacker News Reader

Google X: Waiting for a Moonshot

nytimes.com

122 pointsehudla87 comments

Y'all misunderstand why Google X exists. It exists for three reasons:

1. To obscure the fact that Google is an advertising company, first and foremost, by creating dazzling PR. The more insane the project, the better. Self-flying balloon shooting lasers from the sky? Here's 10 million dollars, go ahead and do it. Over time people lose interest, and those projects disappear into the void.

2. To draw the "best and brightest" in with the promise that they'll work on self driving cars, only to have them repair some obscure dilapidated ad serving backend, where they'll spend years nurturing futile hope that they'll get to work on the cool stuff at some point, and wiping their tears with hundred dollar bills.

3. To keep very senior employees from jumping ship to competitors.

If those three goals are met, actually producing a moonshot every 5 years or so would be gravy, GoogleX will exist even without producing anything at all, indefinitely.

This is pretty much the reason the research wing of most organizations exist.

I don't disagree, but GoogleX is _not_ Google's research arm. Google Research actually produces very cool stuff that's beneficial to the company, and does so at a pretty rapid clip. GoogleX, on the other hand, takes bold, but not really research-worthy ideas and milks them for PR for years before they quietly die.

Do you know what was the lastest cool thing by google research?

It's mostly CS related. Most of their big data systems were initiated by the Research group.

That is not true.

The very latest cool thing by Google Research or Deepmind? That would be applying reinforcement learning to datacenters to cut cooling bills by 40%.

Where do I even begin. :-) State of the art achievements in object recognition, speech recognition, reinforcement learning, natural language parsing, natural language understanding, and so on and so forth, across the board. They employ basically all the prominent names in deep learning, except for Yann LeCun and a handful of others. Their research team is worth its weight in gold.

> 3. To keep very senior employees from jumping ship to competitors.

If I am not mistaken Bill Gates once mentioned in an interview that in the old days, Microsoft would rather pay someone to sit around than letting them work elsewhere.

And the second part of the joke is that these people usually are not real techies who produce, like Wozniak, Ritchie, Thompson, etc. but socially skilled, manipulative talking heads, who inevitable raise to topmost positions if an organization is old enough - the useless ballast of pointy haired managers. There is no better example of that kind than Balmer himself.

Excellent. Keep those characters around, then unleash them when a competitor is looking for a CEO or something. Sit back, get popcorn, watch a competitor crumble. Laugh maniacally?

It's just like that time when Marissa... nah, I was just kidding :)

Google, Yahoo, Verizon, Marissa Mayer, all spring to mind.

If you are referring to the whole Nokia episode, don't forget it was the Nokia board itself that placed a big sum of money on Elop's contract if he managed to get someone to buy the company.

I was not thinking Elop, wasn't he more of a Trojan Horse, a smart guy planted to do evil stuff?

And anyways, I was just joking... I even mentioned that :]

I was going to make a similar comment on another thread. Google X is mainly about the PR value around its image and recruiting/locking up talent. Google would rather have smart people locked up inside the company working on projects with a high probability of going nowhere rather than having them going to a competitor, or worse, creating the next major competitor. Facebook also engages in some of this, though not to the same extent as Google.

FB is just younger. They have plenty of horseshit projects already, the most recent and public one being Aquila. It's not like anyone is going to allow US-controlled drones stuffed with electronics of unknown provenance to fly in foreign nation airspace "for months at a time". CIA is salivating at the thought. Under the US laws, it's not like FB will be able to politely decline if they were asked to carry a little extra equipment.

A counter example would be Monsanto which pretty much saved (or at least fully transformed) the company from one of these small basic research groups.

It think this is more a phenomenon that these big companies become more and more like nations. Nations also fund basic research that 99% of the time has no economical benefit but that one time you invent coffeine free coffee and make a lot of money. But sure, nations are driven by the same or similar motivations you describe.

4. To keep Sergey Brin entertained.

If these points are true, then why has X been spun off into a separate entity (just "X", under the Alphabet holding company, with increasing separation from Google)?

So that if the public catches up to what it actually is PR damage on Google proper would be contained, and a new division ("Google Y"?) could be spun up from its tattered remains?

What always irritated me about Google X was exactly that rhetoric around "moonshots", as if Google had discovered this completely new technique for getting stuff done.

Not only have quite a few companies and organisations successfully completed huge, ambitious, forward-looking projects; but Google haven't, at least nothing remotely comparable to the actual moonshot they happily co-opt. Google has done lots of amazing stuff, but usually in a remorselessly incremental way.

Gmail and Google Maps (especially Street View) are the closest I can think of to "moonshots" in that they were big and bold improvements over what had come before. But neither one would be called a "moonshot" by Google X standards. In each case a back of the envelope estimate would be enough to show that it would work out (in technical terms, if not business terms).

So why does Google act like they're uniquely qualified to make moonshots work?

GMail and Google Maps were also both ruthlessly incremental. They seemed like big improvements over what had come before because they had incubated inside the Googleplex for a couple years before they were released, and because the competition hadn't improved in nearly a decade.

It's more that Search was a non-incremental leap forwards, and that Google was basically the only company of the dot-com era that had attempted an ambitious technological goal and succeeded. Google Search really was fundamentally different from previous technologies; its main rival Alta Vista, for example, ran on a single box and used very standard text-ranking techniques, while Google thought very much out-of-the-box in terms of ranking, scale, and architecture.

Google X is very much a Larry & Sergey project, and Larry (while Eric was CEO) was often very critical of how incremental employee projects were and how they weren't taking enough risks on fundamentally different approaches.

Google search used a couple of clever innovations -- the PageRank algorithm, and scaling via lots of small computers. But apart from that, isn't it almost the definition of incremental? You don't need a massive investment to make it work, just a couple of grad students with 10 PCs to make a proof of concept, then 100 PCs to make it better, then 500...

Thinking about it, Google's key business trick has never been moonshots, it's been "throwing things against the wall to see what sticks". It's a wide approach (invest speculatively in lots of different ideas) rather than a deep one (bet heavily on a really long shot).

They practically operate their own business incubator.

Amazon is a much better example of a moonshot than Google search. The key idea is easy to state: an easy-to-use online store for almost anything, with fast and cheap delivery. But how on earth do you make that work in practice? You need a ton of business deals with various manufacturers and brands, you need warehouses everywhere, you need a massive delivery network... the investment required is just staggering. And indeed, for many years they were criticised for not making a profit, but the investment is clearly paying off.

At the beginning, Jeff have only thought of an online book company, instead becoming the 'Everything Store'. This legacy still exists as ASIN (Amazon Stock Indentification Number) is extended on ISBN to work on legacy systems.

"In each case a back of the envelope estimate would be enough to show that it would work out (in technical terms, if not business terms)."

I would think that applies to the Apollo project, too. An envelope can hold a quick calculation of how large your rocket needs to be, but it isn't large enough to hold any of the minor details that may cause problems. One also could argue it was done "in a remorselessly incremental way".

>Not only have quite a few companies and organisations successfully completed huge, ambitious, forward-looking projects

Hmm, and what would those be?

The first example that came to mind (and made me add "organisations") is the creation of the National Health Service in the UK. A massive undertaking, especially as the country was still recovering from WWII, but massively successful.

Apollo, of course. The Manhattan project. Lots of NASA space probes.

Hmm, to be honest I'm less sure about companies... The phone network? Or laying railtrack right across the US?

Oh, and a topical one: the ARM chip! It was totally ridiculous for a tiny company like Acorn to jump straight from the 6502 to their own custom, cutting-edge microprocessor.

Okay, to me some of your examples seem high risk, but still incremental. There was plenty of prior research/innovation/development in those areas that they built upon. But to be fair, I also think that its almost impossible for someone to independently do something 'earth shattering' without building upon other people's work.

You're right, I suppose it's a continuum -- everything is incremental to some degree or other.

The Manhattan project is probably the best example, because there was nothing to build on except "science strongly suggests we could make a bomb out of this; we have no idea yet how to solve the practical issues, or even what all the practical issues are." It was partly incremental in that a much smaller nuclear bomb would still have been useful, but there was definitely a big risk that it just wouldn't work at all.

This bugs me especially when Silicon Valley companies get involved in life sciences, closer to my field. So much of the buzz is around "innovation" and daring to "think outside of the box," while so much of the actual work comes in implementing and tinkering with basic ideas. It's very weird to see a company say "We're not bound by the slow pace of the establishment, we're going to dream big, and look for a cure for cancer itself!" People have been looking for a cure for cancer for thousands of years--the problem in healthcare has rarely been too-small dreams.

This:

> “I’m an optimist: I believe that there is an endless supply of huge problems affecting humanity.”

combined with this:

> “I think we could meet it, but it would be more like 15 to 20 years.” That was too far away, which was why she recommended killing it.

...seem to be fundamentally at odds. Or, at the least, leave an extremely narrow sliver of projects: big problems that can be solved in under 10 years. And how many projects started out in the engineer's mind as a short task but turned into a decade-long struggle?

Also, I can't look at a picture of that guy in rollerblades without thinking of this:

http://www.businessinsider.com/google-x-astro-teller-hbo-sil...

We all stand on the shoulders of giants.

Huge problems are still solved. Small problems are solved and accumulate until the large problems are suddenly not so large. Then in the blink of an eye they become commodities.

Drones, virtual reality, autonomous vehicles.

And drones are still not yet commercially viable, VR is not yet amazing and in the hands of consumers / widely applicable outside of gaming, and autonomous vehicles are currently hyped-up versions of adaptive cruise control. We're still probably 10 years away for drones and autonomous driving, and 2-5 for VR in terms of large mainstream impact (possibly much longer for truly autonomous cars, depending on AI advancements, laws, etc...I wouldn't be surprised if it's more like 25-30 years until it's the norm).

> And drones are still not yet commercially viable

I don't really understand here, are you talking about a specific area that they're not commercially viable? Because people make and sell them for a profit.

I also see a hell of a lot of drone shots on YouTube. What was once a cinematographic technique that costs millions of dollars, is now within reach of everyone.

I think that's what you SV folks call "disruptive".

> Because people make and sell them for a profit.

I'd say using it for a profit counts even more ;-)

As far as drones go, I guess it depends on what you mean by commercially viable:

https://en.m.wikipedia.org/wiki/Northrop_Grumman_RQ-4_Global...

Drones: Aside from the military applications. DJI is solving a significant problem in agriculture with this: http://www.dji.com/product/mg-1

VR: You say 'outside of gaming', are you aware of how large a market gaming is? (roughly a $25-$30B / year in terms of software alone, theres 50 million units of ps4 & xbone shipped)

Autonomous vehicles: Are you aware that John Deere is a leader in this field? https://www.washingtonpost.com/news/the-switch/wp/2015/06/22...

I know of people using drones to sell shots of homes to real-estate agents.

A 3000 euro drone paid for itself within weeks.

I agree. Think about the Manhattan Project: six years from concrete idea to working prototype. For three of those years, they had a significant chunk of Earth's smartest people living and working together with few distractions or outside responsibilities, an almost unlimited budget, supported by satellite sites all over the U.S. (plus Canada and Britain), logistic support supplied by the military with high priority, and, oh yeah, the motivation that if they didn't win the race Germany could rule the world.

For today's most important "moonshots," like fusion and large scale water purification and pulling lots and lots of carbon out of the atmosphere and colonizing the solar system, we might not even have the kind of clear starting point they had, where the fundamental science had been done and the rest was mostly engineering. It doesn't seem like solving these problems is compatible with any kind of near-term profit motive.

It feels like the corporate version of a lottery: if enough people buy lottery tickets, someone will win. It just probably won't be you. If enough organizations with gigantic piles of cash spend them on long-term research, life on this planet could get way better for all of us. But each organization would have to individually decide to do something whose most likely effect is just throwing money away.

So how do you convince a corporation to start buying lottery tickets?

You're better off funding research and startups directly. We need more DARPA and similar research investment funds.

how could i do that?

>Mr. Teller appears to be always on Rollerblades.

The article seems to confirm the rollerblades anecdote from the New Yorker Silicon Valley story [0]

>Teller ended the meeting by standing up in a huff, but his attempt at a dramatic exit was marred by the fact that he was wearing Rollerblades. He wobbled to the door in silence. “Then there was this awkward moment of him fumbling with his I.D. badge, trying to get the door to open,” Kemper said. “It felt like it lasted an hour. We were all trying not to laugh. Even while it was happening, I knew we were all thinking the same thing: Can we use this?” In the end, the joke was deemed “too hacky to use on the show.”

[0] http://www.newyorker.com/culture/culture-desk/how-silicon-va...

Research divisions exist for recruitment purposes. Sometimes it's just a bait and switch, being promised "R&D" and then getting interviewed for Google Play Store once someone realizes you're not qualified enough. Other times, you work on research for six months, and then there's this team at the company that needs work from you for two years.

Ask the former MSR executive, who fixed the accounting of their losses by pointing out how people at MSR were working on regular MS commercial products too.

Eh, you've got it all backwards. You cannot get a job at X by going into the ordinary Google hiring pipeline, and in any case the Google hiring system does not hire people for specific products like Google Play Store. Furthermore, one does not freely transfer between X and Google, in either direction.

How does one get hired at X? Their jobs page just leads you to google.com/about/careers

Perhaps if you have to go to the jobs page, you are not qualified?

(It's like, if you have to ask for the price, you can't afford it?)

referrals and a PhD (usually)

I don't believe this is true. I know someone was initially picked to interview for YouTube but then got interviewed for X after declining YouTube.

YouTube has retained their own quasi-independent recruiting regime, along with X, Verily, Fiber, etc.

Actually, they are part of the normal Google pipeline (or at least they were before the Alphabet split). A few years ago, I was contacted by a recruiter from Google and after talking to her I ended up being interviewed by someone for a job at Google X.

A few years ago was before Google reorganized into Alphabet.

Here's the thing -- well, two things, really:

1) The article mentions Obi Felton's Foundry team. This was explicitly added to solve one of the deficiencies at X: X is staffed by hardcore engineers & research scientists, not businesspeople. The Foundry team was created to add a commercialization path & process for X research, which never previously existed (except in special cases like self-driving car). That should, in theory, help bring more things to market.

2) Google is averse to investing in anything that isn't a moonshot (where I liberally define moonshot as "at least $1b or 1b users"). So many great ideas and products get shot down because they just aren't "Google scale" even though they'd be perfectly viable, and sometimes wildly profitable, businesses on their own. Because the bar for business viability is so high, lots of things aren't even tried, which is a shame.

I'm impressed that Google/X has been willing to give the self-driving car group enough time and money to do it right. They have about 60 self-driving cars, being driven by paid employees. They have their own proving ground. All that can't be cheap.

I'm disappointed that Google bought all those robotics companies and ran them into the ground.

Well, when the person who probably oversaw those acquisitions leaves, what do you expect to happen? The people who took over probably didn't have the same vision.

That just shows there are at least two ways the problem could have been avoided — don't do the acquisition, or don't lose that key person.

...Or make sure the person taking over has the right vision. Three ways! :)

> bought all those robotics companies and ran them into the ground

Did this actually happen? I know there have been cultural issues with Boston Dynamics, but the other acquisitions seem to have gone fine. Folks from Redwood, Meka, Bot & Dolly, and Holomni are still working at Google and many of the founders are director-level in the robotics groups.

Bot and Dolly had customers, yet they seem to have dropped out of the market they were serving.

Reading this I'm starting to wonder if they are failing too early.

I've often reached a point in fixing some object, or some code, or whatever, when I'm certain I can't do it. I'm ready to just shut it down.

But I have no choice - I have to make it work. So I keep plugging away.

And so far I've managed almost every time - despite being certain I couldn't. Now I'm smarter and no longer so certain :)

But if I got a bonus for failing - well. In each case I could easily, and convincingly, and truthfully! demonstrate it couldn't be done.

I think one of the biggest problems with Google X is that they give their employees a false sense of what I consider "full freedom" which is a big component in making moonshots possible. All of the biggest tech companies in the last 20 years were not conceived within a larger corporation - they evolved out of a garage or a university. What makes Alphabet think that by fueling employees under a new division of their company is going to spin off more cash cows for them?

Also, what happens if an employee manages to produce something that does become a huge success? Will they be able to enjoy the same success as if they had spun up that idea on their own?

> Waiting for a Moonshot

It sounds like they're doing interesting work.

But you don't "wait" for a moonshot, hoping something of that magnitude comes out of all but random motivation from individual team members.

You produce a moonshot by deciding to do that exact thing. Which is exactly how the actual moonshot happened.

If you have random inputs, you'll have random outputs of unknown magnitude.

> Xerox pioneered the graphical interface for computers — the idea that people could navigate with a mouse rather than typing obscure commands on a screen. But it was a young company named Apple that turned that idea into a giant business.

It's weird how the article fails to mentions past successes for other R&D labs. Xerox Parc also invented the laser printer which certainly made a lot of money for the company.

As a standalone device, powered by Adobe Postscript and HP PCL/JCL.

Where did the embedded version in the Star end up?

The self-driving car project that came from Google X sparked the huge investment in autonomous vehicles, and gave Google a huge head start. In addition, they positioned themselves well with a large investment in Uber. The moonshot has succeeded/will succeed from a technological sense. Now they just need the right go-to-market approach.

While I don't disagree with you, I think the real credit goes to Dr. Tony Tether who was the director of DARPA for the duration of the self-driving car competitions. He set aside money for the races and helped design them. After the last race in 2007 Google acquired most of their talent from various teams.

Definitely, but to me it seems like Google was responsible for jumping on it and taking it from university research projects to a viable "moonshot." So it's much less waiting for a moonshot - they already have it - more just waiting for the launch

I wonder if they make available the research they've done on dead projects, like the vertical farming, failed because they couldn't get staple crops to grow well enough.

If you have a great idea and are actually talented, why would you want to go to Google X, rather than start a company yourself and have far more control and freedom in the open market? You could always get Google to invest later, and if you're right and successful you probably have a lot more options for investment than just limiting yourself to Google in the first place. That is, if you're really serious about your idea. Or are you just there to play with Google's money?

You don't see Elon Musk or Richard Branson clamoring to join Google X, or just about any other startup founder.

The entire premise behind Google X (save other than a hype machine) is flawed on basic principle. Seriously, what actual real thing has come out of Google X in all these years?

fallback. I assume they like doing what they do and they prefer having job security. I'd rather try moonshot projects thinking about how to develop it rather than constantly wondering about the business model and if it will be successful (read: stress). They have other people to think about that (the foundry). And they will have much more resources. While I agree that VC funding might be an analogue in a startup situation, it's just not the same. You'll still have the steady paycheck at the end of the month, no matter what

Of course, if someone is confident of their idea and wants to start their own company, they will. Just google 'ex-googler startup'

Astro Teller. Grandson of Edward Teller.

https://en.m.wikipedia.org/wiki/Edward_Teller

This is only tangentially-esoterically related, but due to your comment I ran across some online book called Unsong about a Kabbalistic sweatshop in Silicon Valley. [0] It sounded like a novel humoristic idea. Apparently by coincidence the lead character is named Aaron Smith-Teller, and worked on AI at some point. And Astro Teller wrote a book about AI and also sold a movie pitch to Paramount called Golem (a Kabbalistic reference).

And Gwern posted an interesting thing on Edward Teller's Atomic Alphabet. [1]

[0] http://www.goodreads.com/book/show/28589297-unsong

[1] https://www.reddit.com/r/unsong/comments/4pzyvq/edward_telle...

Isn't Unsong by the same author as Slate Star Codex?

Yes, it's authored by Scott Alexander.

It's not a coincidence; it's discussed in chapter 6 that Aaron Smith-Teller's grandfather was Edward Teller.

http://unsongbook.com/chapter-6-till-we-have-built-jerusalem...

Sorry I wasn't clear. This is the part I was referring to:

>I recently learned that Astro Teller, the real-life grandson of Edward Teller, is the head of Google X Labs. And has a degree from Stanford in “symbolic computation”. And has written a fiction book about a man who creates a self-aware AI. I knew none of this when I wrote the first few chapters of this book. Needless to say, nothing is ever a coincidence.

http://unsongbook.com/authors-note-2-podcast-llull-meetups/

Shouldn't we say Marsshot now? :)

Anybody want to take a shot at a less baity, accurate, neutral title? Preferably using representative language from the article.

The Business Model of Moonshots at Google X (not grounded by language in the article, however)

Corporate Research Labs Chasing Moonshot Projects; Most of Them Don't Pan Out Financially

Google X research lab: Waiting for a Moonshot

Waiting for a Moonshot at the Google X research lab

("Waiting for a Moonshot" is a prominent sub-heading in the article.)

Incorporating another subheading, perhaps:

Google X research lab: Waiting for a Moonshot after 'Failing to Fail'

Waiting for a Moonshot after 'Failing to Fail' at the Google X research lab

Ok, we'll use that suggestion; incorporating a subheading is a good idea. Thanks!

Inside Google X: Moonshots and Rewarding Failure.

Not sure how this is baity. The jetpack portion refers to an actual example of research done by X and the profit promise portion refers to the hit or miss nature of the research as it applies to Alphabet's main goal of making money. Am I missing something?

It guess it isn't super baity, yet it felt that way to me. I'm not sure exactly why, other than that it invokes the "they promised us flying cars" meme.

Reply on news.ycombinator.com