news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

do androids dream of electric sheep? perhaps the answer lies in the world of modern robots.

hans peter brondmo, former vice president of google x and former ceo of everyday robots, said that at everyday robots, robots dream all night, learn in a simulated world, and can master new skills when they "wake up".

everyday robots has a futuristic lab where robotics expert peter pastor has configured seven industrial kuka robot arms in what he calls the “arm-farm.”

these robotic arms work around the clock, moving to random locations above the bins like claw machines, and then reaching out to grab items in the bins, such as sponges, legos, rubber ducks, or plastic bananas. cameras above the bins can capture every movement of the objects and the robotic arms, helping the system determine whether the grab was successful.

through repeated training, the robot's success rate increased from the initial 7% to more than 70%.

this result is encouraging, but it is still not enough. learning to pick up a yellow duck in a few months will not allow a robot to perform tasks in the real world in hundreds of years.

so everyday robots developed a cloud-based simulator that created more than 240 million robot instances, with a built-in real-world physics model that simulates the weight and friction of objects.

thousands of virtual robots perform tasks simultaneously through simulated cameras and simulated robotic arms, experiencing millions of failures, and finally the algorithms with excellent performance are applied to real physical robots for training in the real world.

everyday robots has made remarkable achievements in the field of robot learning. the debate between “end-to-end learning” and “hybrid methods” has never stopped, and everyday robots has made significant progress in both aspects.

in 2023, openai launched chatgpt, which made everyday robots realize the importance of data and set the next goal: to let millions of robots perform tasks in the real world, collect enough data to train the final model, and enable robots to perform complex tasks beyond the narrow training definition task range.

it took everyday robots seven years to transform robots from picking up objects to being able to roam around google buildings and do chores like cleaning tables and sorting garbage. the company is confident that it is leading a future era for robots.

hans peter brondmo talks to his mother on the phone every week, and she always asks him the same question: "when will the robot come?" he always answers: "it will take a while, mom."

unfortunately, hans peter brondmo's mother passed away in 2021 and did not live to see the arrival of the robot, which seemed to foreshadow the ending of everyday robots.

just two months after the launch of chatgpt, google suddenly announced the closure of everyday robots, and only a few robot projects and members were transferred to deepmind for further research.

hans peter brondmo also left everyday robots and turned to photography, seemingly feeling a little disillusioned.

ten years ago, google started a robot acquisition spree, acquiring companies such as boston dynamics one after another.

however, ten years later, boston dynamics has changed hands many times, and the glory of google robotics has fallen into a trough with the closure of everyday robots and the loss of many researchers.

the garden of eden for scientific research

in google's prospectus for its ipo, there is an open letter that the two founders insisted on including. the letter begins like this:

google is not a traditional company. we don't intend to become one.

the two founders were worried that going public would damage google's independence and objectivity, and that they might give up long-term interests for the short-term needs of the market. in order to ensure the independence of management decisions, google created a "double voting rights" structure: the voting rights of management class b shares are 10 times that of class a shares owned by the public.

at the end of the open letter, they emphasized:

don’t be evil. we firmly believe that we’ll be better off in the long run as a company that does good in the world — even if we forgo some short-term gains.

this open letter was later called the "don't be evil declaration."

google x was founded with this value in mind, with the goal of developing and launching technologies that can improve the lives of millions or even billions of people. all projects created by this lab are collectively referred to as "moonshots."it symbolizes working towards a seemingly impossible goal, just like the moon landing project.

unlike mainstream workplace rules, google x does not require employees to make detailed plans before making decisions or proposals. ceo astro teller welcomes employees to come up with all kinds of innovative and even crazy ideas.

astro teller once proposed the idea of ​​installing a speaker on a plant car (a technology that can roll in the field to collect plant data) to play motivational speeches to the plants. he doesn't care about other people's strange eyes. in his opinion, the value of an idea does not lie in how good the result is, but in whether it is innovative and can break conventional thinking.

he said, "if teams take the time to explore and test silly ideas like this, it can help 'break' their thinking and achieve results they might not otherwise have achieved."

of course, even google x has to take reality into account. there is a review meeting almost every week to discuss and reject the crazy ideas proposed by members. about 90% of the projects will be stopped at the proposal because they are too difficult or too expensive.

once approved, the support provided by google x is incredible.

astro teller once asked google to provide $30 million for his project, but the other party directly gave him $150 million. eric schmidt, then ceo of google, said, "if i give you $30 million, you will come back next month to ask for $30 million."

at google x, the frequency of project failure is much higher than that of other labs. however, unlike traditional companies, google x does not regard failed projects as a shame and shut down the entire project and lay off the relevant team once a failure occurs.

astro teller established a project failure bonus to encourage everyone to take some risky attempts and to comfort those employees who tried their best but failed.

google x also held sharing sessions at the suggestion of employees, where employees could share their experiences of failed projects and changes in their lives.

pursuing innovation, encouraging risk-taking, and embracing failure, google x has created a pure scientific research paradise in this atmosphere.

google has been exploring the combination of ai and robotics, with the goal of creating robots that can perform complex tasks. google x's general learning robot project, everyday robots, is the product of this vision.

everyday robots’ research is progressing very smoothly. the team has developed an ai model that can be trained in a simulated environment and transfer the excellent performance results to actual robot hardware for further testing and optimization.

although the complexity and unpredictability of the real world leads to a significant gap between robot performance in simulation and reality, this has not hindered everyday robots' research pace.

these robots gradually acquired the ability to complete simple tasks such as tidying up desks in real environments, which also means that robots can effectively recognize humans and objects around them with the help of ai. soon, these robots were given more responsibilities and were busy in google buildings, handling various daily chores, becoming true "special employees."

not only chores, everyday robots has expanded the application of robots to art.

catie cuan, a stanford phd student and former professional dancer, worked with engineers to develop an algorithm that can train robots based on choreographers’ preferences. in her free time, she dances with these robots to explore new fusions of dance and technology.

tom engbersen from the netherlands was very interested in this, and together with catie cuan, he designed different sound effects for each joint of the robot. after turning on the music mode, each movement of the robot will create a unique orchestral piece.

hans peter brondmo and his mother always believed that these robots would soon enter the daily lives of ordinary people. but his mother did not live to see this day, and the cruelty of reality did not end there.

the conflict between ideal and reality

business black hole?

according to consulting firm evercore, google invested $650 million in research and development budget on google x, but incurred a loss of $900 million.

silicon valley entrepreneurs and academics often criticize google x for "burning money". how could it consume so many resources but still fail to produce a truly "phenomenal" product?

ultimately, google x is more like an innovation playground for idealists than a qualified business laboratory.

for example, the well-known google glass achieved a milestone in technology, but ultimately failed due to high costs and a cold market response and was unable to be commercialized.

google glass was officially released in april 2013. the initial version, explorer edition, was priced at $1,500, or nearly rmb 10,000. needless to say, the technological value of google glass is great, but unfortunately the mainstream market is not buying into it.

on the one hand, google glass's convenient shooting function involves privacy issues and has caused great privacy controversy.

on the other hand, social acceptance of this new high-tech glasses is surprisingly low. according to a survey by mobile application company bite interactive, 38% of respondents said they would not wear google glass even if the price was within their budget. another 45% of people are worried that wearing google glass will cause social embarrassment or unpleasantness.

kevin werbach, a professor of legal studies and business ethics at wharton, also noted that “it may take several years for google glass to be affordable enough for the general public, and it may take even longer for society to accept such products.”

google glass is just the tip of the iceberg of failed projects at google x. other projects include the internet balloon loon, the modular phone project ara, and the energy kite makani, all of which were unable to go to market due to factors such as being too complex or too expensive.

innovation is sometimes like a gamble. google's bets are big, but it doesn't always win.

in 2015, google established alphabet as the parent company of google. at the same time, it separated the google x, deepmind, waymo and other departments that were originally operated independently and put them under the name of alphabet as subsidiaries.

alphabet's purpose is to help google focus on its profitable core business without being held back by "money-burning" innovation projects.

january 20, 2023alphabet failed to withstand the economic downturn and announced that it would lay off about 12,000 employees, accounting for 6% of its total workforce.everyday robots is one of the few projects that was disbanded due to budget cuts., becoming a victim in this game of innovation and interests.

common problem of large companies

according to alphabet's 2023 financial report, advertising business accounts for about 80% of google's total revenue, cloud computing services contribute 10% of the total revenue, and the remaining 10% comes from hardware and other businesses.

google's dominance in the advertising industry guarantees a steady stream of huge revenue, and also reflects its heavy reliance on advertising. even its new profit growth point, cloud computing services, have remained loss-making until 2023 when it finally achieved its first profit.

there are indeed many failed cases in the market where profitable projects are single and highly dependent on income from a specific business.

the most famous case is nokia, once the world's largest mobile phone manufacturer. it had no rivals in the feature phone market, but because it failed to transform into smartphones in a timely manner, it quickly lost market share and was eventually acquired by microsoft in 2013, withdrawing from the mainstream mobile phone market.

another classic example is yahoo, which once relied on portal websites and search engines to become an internet giant, and advertising business was also its main source of income. however, it was defeated by google's offensive and was acquired by verizon in 2017 and reorganized into oath.

we can still vividly remember how yahoo declined in the past. now it may be on the verge of following yahoo’s old path. google is also very anxious, otherwise it would not have stopped its research when its own robots were gaining momentum and turned to follow openai to start building large language models.

a strange phenomenon has emerged at google. on the one hand, various innovative projects emerge in an endless stream, and there is nothing that the industry cannot think of that google cannot do; on the other hand, google does not have any trump card projects that can stand out in areas other than advertising, and always lags behind.

the difficulty of profitable open source has been a tight chokehold on google, but misfortunes never come singly, and google has begun to lose employees.

sheryl sandberg, google's vice president of advertising and commercialization, left for meta, thomas kurian, vice president of cloud business, left for oracle, and youtube ceo susan wojcicki also ended her 25-year career at google.

many of the employees who participated in google's robot moon landing project have gone to other companies: yan mengyuan went to openai, eric jang chose the startup 1x, and yao lu, one of the initiators of deepmind rt-1, rt-2, rt-x and other projects, is now a senior researcher at nvidia...

a senior operations engineer at google wrote a resignation report on his blog, reviewing his nine years at google:

initially, what attracted her to google was its technology. however, as she continued to be promoted, she felt increasingly stressed by the company's bloated organization, tedious and meaningless meetings, repetitive work, marginalization, and frustrated innovation, and eventually chose to leave google.

this also reveals a bigger problem: the expansion of a company will inevitably lead to a complex hierarchical structure and low decision-making efficiency. google, which wanted to rely on double voting rights to get rid of this problem, eventually suffered from the common problem of large companies being bloated and rigid.

a former everyday robots employee said in an interview,the company had a hard time defining its mission. the team couldn’t decide whether their goal was an advanced research project or a commercially viable product.

google, which once had clear development goals, is now lost at the crossroads of innovation with a vague mission.

google's lost way of innovation

what's the problem?

back in 2010, google founders larry page and sergey brin decided to set up a new department to study the "moon landing plan":

technologies that sound far-fetched and sci-fi may one day make the world a better place.

google x is this bold experiment, and it has indeed spawned a lot of "black technologies" that sound very sci-fi.

for example, the famous google glass, when worn, can directly place a virtual screen in front of your eyes, which can be controlled through gestures, voice, etc., and is also equipped with functions such as bone conduction.

another example is the 280 earth project that targets the problem of carbon emissions. if the carbon dioxide content in the atmosphere is too high, the carbon dioxide will be directly absorbed. the absorbed carbon dioxide can also be used to produce fresh water, and the waste heat in the process can be used to power the next stage.

google established google x to bring all the innovation projects with the highest risks and the lowest short-term returns into one place and operate it as an "external brain".

as early as the beginning of its listing, the two founders of google were worried that the short-term interests demanded by future shareholders would conflict violently with the long-term investment required for scientific research, causing google to lose its original intention and abandon long-term projects, no longer innovate but only make quick money.

in order to avoid this situation, the special google x was created, which is both an "external brain" and the original intention.

google once had a "20% innovation time" policy, which means that employees can spend one day a week on projects other than their job duties. this innovative management rule has spawned excellent products such as gmail, adsense, google news, and gtalk.

but the policy has been abolished and was considered "dead in name only" by internal employees as early as 2013.

at that time, if employees wanted to enjoy this benefit, they had to apply, and it became increasingly difficult to pass the approval process. there was also an internal employee efficiency testing system to determine whether employees were 100% committed to their work.

in fact, the cancellation of the "20% innovation time" is also the inevitable result of google's pursuit of efficiency and standardization after its scale has expanded, but google has not given up the pursuit of innovation. in a sense, google x is the substitute of this policy and has undertaken the mission of innovation.

but judging from its performance in recent years, google has not only failed to maintain its previous innovation atmosphere, but has also fallen into a new innovation trap.

eric schmidt, former ceo of google, let loose in a speech at stanford not long ago, saying that "google lost to startups such as openai in the ai ​​competition because google's remote work policy caused employees to not work hard enough."

this is a microcosm of google's innovation trap. the top management is unwilling to relax working hours and wants to force employees to focus on their work by tightening management policies, hoping that this will improve work efficiency and allow employees to produce a steady stream of innovative products.

google x's design kitchen

but history tells us that the start-up phase is when a company is most innovative, and only at this time is the company truly innovation-oriented; once the company expands in size, it will have to face the difficult trade-off between innovation and profitable core business.

history also tells us that history is always surprisingly similar and human beings always repeat the same mistakes.

nokia, yahoo, and blackberry all missed the opportunity to transform due to being too conservative and not innovative enough, and were abandoned by the cruel market. however, google still had the fluke mentality that "change will not come so soon", coupled with short-term financial pressure, gradually fell behind in innovation.

the embodied intelligence craze is here. does google still have a chance?

according to cb insights, the total investment in the global robotics industry reached us$8.9 billion in 2022, an increase of 44% over 2021. among them, embodied intelligence startups such as agility robotics and boston dynamics have received large-scale financing, and it is expected that the global robotics market will exceed us$40 billion by 2025.

in the past two years, the development of embodied intelligence has made great strides, and major companies and scientific research teams have made considerable breakthroughs.

nvidia is one of the leaders. for example, the gr00t project launched at the gtc developer conference this year, and eureka, which was previously named one of the "top ten nvidia projects in 2023". in addition to the full-stack robot platform, nvidia also announced an acceleration plan for humanoid robot developers, aiming to further promote the rapid development of embodied intelligence.

not only nvidia, but also 1x, a company invested by openai, is not willing to lag behind. it has launched a new humanoid robot neo and released their own world model.

fei-fei li's team at stanford university also made major updates to the household robot and proposed the new rekep technology. at the same time, world labs, founded by fei-fei li, announced plans to build a "big world model" that allows ai to perceive, generate and interact in the 3d world.

chelsea finn's team, another pillar in the field of embodied intelligence at stanford university, collaborated with deepmind on the aloha project and launched the mobile aloha cooking robot at the beginning of this year. recently, they launched the upgraded aloha 2 robot, which can perform more delicate movements. in addition, chelsea finn's team also proposed a new algorithm for bidirectional decoding bid.

however, not all companies will survive the wave of embodied intelligence.

an everyday robots employee once said in an interview, "the company made premature promises about practical and reasonably priced robots but failed to deliver. i hope everyday robots will not make the same mistake."

everyday robots, which once adhered to the "moon landing plan" principle, still fell before the craze of embodied intelligence, but at google, its mission did not end there.

the robotic arm that once “caught dolls” in arm-farm is now showing its prowess in deepmind’s new project. the robotic arm played back and forth with human players, becoming the first learning robot agent to reach the level of amateur human players in table tennis.

google still has a chance of winning in this wave of embodied intelligence, but the prerequisite is that it needs to create products that are amazing enough to differentiate itself from other competitors.

when you win, whatever you say is right.

there is a famous saying in the school of ruguan, "after entering the pass, there will be great scholars to debate for me." as long as you are strong enough, there will be people to defend you and stand up for you. in fact, google did not lose in management policies, but lost in the courage to innovate.

what innovation requires is wild growth rather than a defined framework. dancing in shackles will inevitably fail to achieve a qualitative breakthrough.

the young man who first tinkered with the search engine in the garage was able to challenge the internet giant with just his passion. if google does not want to miss this wave of embodied intelligence, it should reflect on its current timid situation, regain the courage to subvert itself, and meet the challenges from up-and-coming companies.