Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
AI Communications Google The Internet

At X, Failure Is Not an Option: It's a Feature (Astro Teller's 2016 TED Talk) (backchannel.com) 70

New submitter Evan Hansen writes: Everyone likes to pays lip service to "fail fast," but when was the last time your boss gave you a bonus when your project was killed? In his 2016 TED Talk, concluded just moments ago, Astro Teller, the head of Alphabet's X R&D lab shares some never-before revealed stories of his team's failures and iterations, and explains how "fail fast" can be more than a trite cliche. The first X project was the self-driving car, and subsequent ones include Google Glass, Project Loon's Internet service via balloon, Makani energy kites, and a drone delivery service dubbed Project Wing.
This discussion has been archived. No new comments can be posted.

At X, Failure Is Not an Option: It's a Feature (Astro Teller's 2016 TED Talk)

Comments Filter:
  • by DNS-and-BIND ( 461968 ) on Monday February 15, 2016 @09:30PM (#51516433) Homepage

    Man, it has got to be great being at the top. Where you can fail and nobody will fire you for it. In the rest of the world, when you are associated with a failed project, that puts you first in line the next time layoffs come around. That almost happened to me once, but luckily I smelled it coming and managed to distance myself from my friend who was in charge of the failing project. It broke up our friendship, but I still had a job afterwards and he didn't. So, you could say I "failed fast" in that I quickly determined that my friend had a failure face and I quickly failed to stick with him when times got tough.

    But for some reason I was never invited to give a TED talk about it. I wonder why? Oh yeah, I'm a filthy prole, I forgot. Why did Astro Teller get to give a TED talk? Because he's the grandson of Edward Teller, the father of the hydrogen bomb. His father was a professor. Stanford couldn't admit him fast enough. He got into Google because Googlers distrust people who are not like themselves. He even gets away with a stupid-ass name like "Astro" where in the rest of the world a name like that will get you swirlies.

    • Re: (Score:3, Insightful)

      by JoeMerchant ( 803320 )

      I'm fairly sure your friend's project wasn't a fail-fast investigation, but rather a sure thing development.

      I think the best line in the talk was "shifting perspective is more powerful than being smart, if you're coming at an established problem from an established approach, you're competing with all the other smart people who came before you, and that's a terrible place to be competitively." So, what do they do? They try unusual approaches, they fail, they try again, they fail better next time, and most

      • What I was saying flew right over your head, didn't it? All that education, and you never learned a thing.

        • What I was saying flew right over your head, didn't it? All that education, and you never learned a thing.

          Don't worry about it, the stuff they talk about at TED obviously doesn't apply in your life.

          • The stuff they talk about at TED doesn't apply to anyone's life. TED is hot air, fluff, and pie-in-the-sky bullshit. It's a fucking joke!

            • by Ol Olsoc ( 1175323 ) on Tuesday February 16, 2016 @12:11AM (#51517077)

              The stuff they talk about at TED doesn't apply to anyone's life. TED is hot air, fluff, and pie-in-the-sky bullshit. It's a fucking joke!

              My favorite TED talk

              : https://www.youtube.com/watch?... [youtube.com]

              The talk itself is some self indulgent stuff about how she came to believe that there is a God. But irony abounds!

              http://gawker.com/professional... [gawker.com]

              This new reformed woman had a bit of a road rage incident, followed a 73 year old woman home, ran over her with her car, then started to back up to run over her again but fortunately was stopped.

              I haven't looked at TED talks the same way ever since .

              • I knew before I even clicked the link, but you're referring to TEDx talks and not TED talks. Yes, there is a difference and it's huge. It's pretty much a guarantee that any woowoo shit will be a TEDx (the 'x' is probably for 'x'-tra crazay).

              • She believes God put her on this earth to rid us of annoying drivers?

      • by Anonymous Coward

        "I'm fairly sure your friend's project wasn't a fail-fast investigation, but rather a sure thing development."

        Nice Marketing-Speak there.

        "I think the best line in the talk was "shifting perspective is more powerful than being smart, if you're coming at an established problem from an established approach, you're competing with all the other smart people who came before you, and that's a terrible place to be competitively." So, what do they do? They try unusual approaches, they fail, they try again, they fail

        • So being able to shift perspective is something smart people don't do?

          It's not about smart, it's about appetite for risk. Plenty of smart people work for big, successful, conservative companies. For the most part, they don't stay employed there by suggesting wild approaches to established problems. There are lots of creative people in the world, companies that embrace that creativity in an attempt to create new, disruptive and potentially highly valuable technologies are not as common.

      • They try unusual approaches, they fail, they try again, they fail better next time, and most importantly: they identify when they fail to free up resources to try something else.

        The skill that smells failure early is the same one that smells a distant success. That is you have the vision/intuition/awareness to see things which are not in the visible range of others.
        A less aware person will keep working on xyz even after it is exhibiting signs of failure...his read-resolution of failure is low; Thus as you train yourself to spot failure early; you also gain to see that unusual smell of success from a vast distance -- like how a shark knows the presence of an injured prey from a

        • by Gr8Apes ( 679165 )

          The skill that smells failure early is the same one that smells a distant success.

          Hate to break it to you, but the ability to smell shit no way relates to being able to smell roses. Some people just can't smell the roses.

          • depends if the shit is 10 km away or 1 feet away. It doesn't need keen awareness to do the latter. Basically you can't go to the left too much without swinging the other side - ying n yang of existence.
    • Man, it has got to be great being at the top. (snip)

      Kudos to you DNS-and-BIND. Your comment is more insightful than the parent article.

    • by guruevi ( 827432 ) <<eb.ebucgnikoms> <ta> <ive>> on Monday February 15, 2016 @10:42PM (#51516763) Homepage

      The problem with firing everyone that makes a mistake is that you create a culture that just goes with IBM/Microsoft/contractor because you can just shift the blame. You spend insane amounts of money without actually innovating or doing anything that progresses your company. In the end your company goes bust because it can't compete with entities that are more agile and are innovators in their market.

      • And like most things, the sane answer is somewhere in the grey area.

        • by JoeMerchant ( 803320 ) on Monday February 15, 2016 @11:08PM (#51516867)

          And like most things, the sane answer is somewhere in the grey area.

          The sane answer is the boring answer - no risk no reward...

          Of course, serious R&D, whether it's insane spoiled children in California today, or gentleman scientists at Bell Labs in the 1950s and 60s, basically requires a lot of input capital... many ideas will fail to prove their worth, and once in awhile you'll get something like the transistor.

    • by Shawn Willden ( 2914343 ) on Monday February 15, 2016 @11:24PM (#51516927)

      Man, it has got to be great being at the top. Where you can fail and nobody will fire you for it. In the rest of the world, when you are associated with a failed project, that puts you first in line the next time layoffs come around.

      This depends on the nature of the project. If you're doing yet another ERP system integration, yes, you should succeed, and there should be negative consequences if you don't, because there is nothing new involved, just lots and lots of detail-oriented grunt work. It's hard, but good planning and careful attention to detail will get you to the end of the job, and if it doesn't, it's because you did a poor job, not because the job was not doable.

      But when you're breaking new ground, trying to do things that have never been done before, if your management expects success, they're idiots. I grant that lots of companies are led by idiots, but that doesn't make them any less wrong. Google is where it is precisely because its leadership understands that failure is always possible, and the more audacious the goals the more likely it is that you'll fail. When you're pushing the envelope, true failure isn't when the project stumbles and falls due to some major technological obstacle. It's when you fail because of something you really could have foreseen... or when you fail to learn from your failure.

      Google has a culture of post mortem analyses to ensure that that last sort of failure doesn't happen. Whenever anything goes seriously wrong, a detailed analysis of what went wrong, and when, and why, is conducted. Not with the goal of identifying people to blame, but to determine what lessons can be learned and what can be done to ensure that failure doesn't happen again. For someone who grew up outside of the Google culture, it's pretty terrifying to go through your first post mortem, for something you built. It's hard not to be defensive. After a few times, though, you begin to internalize the fact that unless the cause was you just not doing your job, you have nothing to worry about and may well come out with kudos or even a bonus.

      He got into Google because Googlers distrust people who are not like themselves.

      I got into Google as a 42-year old graduate of a podunk little four-year state school that no one has heard of, with an unaccredited CS program. I can't comment on whether I'm like the typical Googler, because I'm not sure what that is. Aside from the well-publicized issues around gender and race, which exist across nearly all tech companies, it's a very diverse place (note that race and gender diversity are important, but they're far from the only forms of diversity). The "Stanford MSCS" is well-represented, of course, but I work with one engineer who has only an associate degree and another who got his GED at age 29 and never attended college at all.

      He even gets away with a stupid-ass name like "Astro" where in the rest of the world a name like that will get you swirlies.

      Very true. All sorts of quirkiness is accepted and even embraced at Google. That's one of the things I most enjoy about working here.

      • , true failure isn't when the project stumbles and falls due to some major technological obstacle. It's when you fail because of something you really could have foreseen... or when you fail to learn from your failure.

        While it's true to see it from failure perspective, the real benefit of the new approach is you may find a hidden new efficient path which no one before you saw. It's like how columbus went west to reach India; You never know what treasure you may arrive at because you are moving in a totally unknown path. The rewards may be staggering. I see moving into the unknown is the key here -- trusting your gut.

      • by swb ( 14022 ) on Tuesday February 16, 2016 @06:34AM (#51518133)

        This depends on the nature of the project. If you're doing yet another ERP system integration, yes, you should succeed, and there should be negative consequences if you don't, because there is nothing new involved, just lots and lots of detail-oriented grunt work. It's hard, but good planning and careful attention to detail will get you to the end of the job, and if it doesn't, it's because you did a poor job, not because the job was not doable.

        Who's the "you" in this? Any computer system project that is sufficiently embedded in business process has failure modes so far outside the reach of anyone on the technical side that blaming technical people for it is incredibly myopic. They often end up failing even when the technology works right.

        And almost all of this is beyond the reach any single individual who's not a senior manager and then there are still conceptual problems with where blame ought to be assigned. And once you get into "you" as a team/group, often with differentiated roles and responsibilities, where failure modes can vary widely.

        I work for a small IT consultancy and I work on projects where all kinds of problems crop up. Most don't cause the project to "fail" (even when they go badly, I think there are various organizational and individual reasons for resisting declaring failure) but even when I'm the sole implementer of the project, there are failure reasons that go beyond me. Poor scoping. Client interference. Unrealistic timelines. Inadequate resources. The list is long and full of things I had no control over.

        • by bigpat ( 158134 )

          I'd say that failure of business systems projects are pretty common and largely the result of either:

          1) business people (project managers, analysts, middle management) that don't think about time or money and are dead set to make people adhere to arbitrary "requirements" that should have been cut as soon as someone said it would take more than 30 seconds to do or they never even asked and assumed it would take 30 seconds to do because they lack experience to know.

          and/or

          2) IT people that will give the busine

          • by swb ( 14022 )

            IT people that will give the business as much rope as they want to hang themselves with without pushing back because they are either working by billable hours or are used to working billable hours and are more than willing to let project costs and timelines spiral out of control because that is what puts the meals on the table and they can just blame the project management and requirements when management suddenly decides they would have rather have a project succeed than get every requirement implemented.

            This may be a practice that "big consulting" can get away with, but almost never is it something you can get away with in the SMB sphere. Any project that accrues change orders resulting in cost increases of more than about 10% gets turned into a big pissing match and often resulting in a compromises that make nobody really happy.

            I've seen squabbles over $300 worth of travel to get someone on site to fix a problem or 2-3 hours of time.

            Nobody wins when projects spiral like this; I've actually seen more "dam

            • by bigpat ( 158134 )

              Oh I've seen consultants milk contracts and rolling requirements specification for tens of thousands of dollars and months of work in a midsize company.

              Yes, small businesses are usually much much tighter with money and I've seen them have the opposite problem where if you have to justify every expense with the real cost that things simply don't get done that should get done.

              • by swb ( 14022 )

                Maybe I'm just lucky (or unlucky, although I'd make no more money personally) that I've never seen this happen, and certainly never deliberately.

                The only thing that seems to come close are projects that get strung out over time due to customer cancellations of cutover dates or other scheduling complexities. This just kills project efficiency.

                Scoping problems happen, too, where some kind of details are missed or problems that don't turn up until you start an upgrade but usually these get dealt with pretty o

                • by bigpat ( 158134 )

                  Yes, software projects are notorious for this scoping issue. Think healthcare.gov. Without ascribing even any milking motive... it is simply very hard to conceptualize what it might take to build out some software systems. If you conceive a simple button on a screen, it may take a few seconds to place the button there but then it has to do something when you press it.

                  People often think in terms of UI for scoping purposes. But think about the data, and back end processing that needs to happen. It is

                  • by swb ( 14022 )

                    The parent article I originally responded to was describing ERP rollouts or something "well known", so those might also fit the endless spiral definition because any project with sufficient business process penetration is likely to have a broad set of stakeholders (bordering on infinite if you take customers into account) and nearly impossible to scope second-order and beyond consequences.

                    Now that we've exchanged these messages and I've thought about it, one reason (besides project type) why I can see some

        • This depends on the nature of the project. If you're doing yet another ERP system integration, yes, you should succeed, and there should be negative consequences if you don't, because there is nothing new involved, just lots and lots of detail-oriented grunt work. It's hard, but good planning and careful attention to detail will get you to the end of the job, and if it doesn't, it's because you did a poor job, not because the job was not doable.

          Who's the "you" in this? Any computer system project that is sufficiently embedded in business process has failure modes so far outside the reach of anyone on the technical side that blaming technical people for it is incredibly myopic. They often end up failing even when the technology works right.

          Yes, the "you" in my statement encompasses the relevant business organizations as well as technology. The entirety of the project, not just the technical people. This applies throughout the remainder of my post as well.

    • by moorley ( 69393 )

      Woohoo!

      Well said!

      Nice to know I am not the only insightful cynic...

      We live in interesting times...

  • so we can move on and leave you behind.
  • Google Glass? (Score:4, Insightful)

    by 93 Escort Wagon ( 326346 ) on Monday February 15, 2016 @09:38PM (#51516477)

    The inventor of Google Glass either left Google or was asked to leave. But, in either case, I don't see how that fits this narrative.

    • Re: (Score:3, Funny)

      by Anonymous Coward

      Probably for being a Glasshole.

    • The inventor of Google Glass either left Google or was asked to leave. But, in either case, I don't see how that fits this narrative.

      Sounds like he might not have been a good fit for the fail and try again culture.

      • Google brought in a marketer who worked at "Calvin Klein, Swatch, Coach, The Gap, Old Navy," etc. And now I think Tony Fadell is now at the top of the project's leadership.
        • Google brought in a marketer who worked at "Calvin Klein, Swatch, Coach, The Gap, Old Navy," etc. And now I think Tony Fadell is now at the top of the project's leadership.

          That fits with the biggest criticism of Google Glass $1500 edition.

    • The inventor of Google Glass is a Georgia Tech professor [gatech.edu] who I assume only showed up at Google because they wanted to make a product out of his research. Why should he keep hanging out there after the project is done when he could be in the Caribbean figuring out how to talk with dolphins instead? (Yes, that is actually what he's researching now.)

  • by creimer ( 824291 ) on Monday February 15, 2016 @10:49PM (#51516785) Homepage

    Fail fast works better for sales calls. Talk to as many people as possible to overcome the fear, doubts and frustrations about selling. The faster you fail at selling, the more likely to get a sale.

    • Fail fast works better for sales calls. Talk to as many people as possible to overcome the fear, doubts and frustrations about selling. The faster you fail at selling, the more likely to get a sale.

      Absolutely true, and it's the same kind of personality test: you've got to be resilient. There are different ways of getting that resilience, a common one in sales is magical or delusional thinking - which would be a not so great fit for product development engineering, but in sales the key is never giving up - believing in your product, and transferring that enthusiasm into your mark (potential customer's) mind. That and incessantly reminding them about the "opportunity to purchase."

      I worked for a much l

  • For my work, failure is not an option, it comes standard! I wish I was compensated for all the failure and destruction I have left in my wake. That said, the most dramatic failure that followed me I cannot take credit for (I worked for CompUSA a few years before they went belly-up) but plenty of less spectacular failures can be tied to me and I never got bonuses for them.
    • I made a mistake that got 17 people killed, and I got a promotion!

      -Future Google Autonomous Car "Engineer"

  • by penguinoid ( 724646 ) on Monday February 15, 2016 @11:20PM (#51516913) Homepage Journal

    The fastest failure, is when you don't even start. And no, you're not getting a bonus for that either.

    • by bigpat ( 158134 )

      The fastest failure, is when you don't even start. And no, you're not getting a bonus for that either.

      Well, assuming you don't just say no to every risky idea and are adding value to the decision chain by approving projects that are more likely to succeed then yes I think you would eventually be getting a bonus based on success of some of those projects. Or if you are not a decision maker then at least you can inform your management about a dead end and then hopefully get reassigned to something you can succeed at doing.

      Overall, I think the real test is whether you can document the unknowns and document t

  • Fail fast is another way of saying "spot a dumb idea before you've spent a lot of time and treasure, and kill it before it spreads." Unfortunately the MBA types thought it was a business strategy.

  • My ATX mobos fail regularly, I dunno what they're talking about.

  • by colinwb ( 827584 ) on Tuesday February 16, 2016 @01:51AM (#51517355)
    The British economics commentator Hamish McRae once wrote about a bright young executive put in charge of a project which subsequently failed: he tried to resign but was told by a superior that he had just been through an expensive training course in what not to do, and that if he thought that he could leave and take that knowledge to a competitor he had better think again.
    • That is dumb advice. He didn't learn "what not to do" because there are infinite ways for a project to fail. You can fail at everything and it won't make your next project any more likely to be successful.
  • Even the smartest researchers will have some trouble imagining the end results of their ideas far in the future. You have to experiment. You can try to guess which ideas are the best, but then you have to try them out. Many will fail. The whole point in “fail fast” is to do feasibility experiments up front to eliminate the bad ideas as fast as possible so you can move on to the next one. If you’re any good, you’ll have some success rate — 10% would be good.

    Even in academia

Serving coffee on aircraft causes turbulence.

Working...