scritic + google   79

Google Has Made a Mess of Robotics - Bloomberg
Sounds like a hit job to me.
The top roboticists netted by Google starting in 2013 have been given little direction and have little to show. They’re starting to strike out on their own.
google  research 
october 2017 by scritic
How Google Took Over the Classroom - The New York Times
Essentially first they captured the emails and apps market; then they brought in the Chromebook and now they have a Teacher Suite to do a lot of things like taking attendance. etc. a lesson in market capture.
google  moocs 
may 2017 by scritic
The secret lives of Google raters | Ars Technica
useful for both 100g and 100j on the AI section and on casual labor section
google  platformization  teaching 
april 2017 by scritic
searchqualityevaluatorguidelines.pdf
the rater guide google gives to its search engine raters, could be used for both 100g and 100j
google  platformization  teaching 
april 2017 by scritic
Fury Road: Did Uber Steal the Driverless Future From Google? - Bloomberg
As Google’s car project grew, a debate raged inside the company, reflecting a broader dispute about the direction of autonomous vehicles: Should the tech come gradually and be added to cars with drivers (through features like automatic parking and highway autopilot) or all at once (for instance, a fleet of fully autonomous cars operating in a city center)? Urmson, a former Carnegie Mellon professor, preferred the latter approach, arguing that incremental innovations might, paradoxically, make cars less safe. Levandowski believed otherwise and argued that Google should sell self-driving kits that could be retrofitted on cars, former colleagues say.

Urmson won out, and according to two former employees, Levandowski sulked openly. After one dispute between the two, Levandowski stopped coming to work for months, devoting his time to his side projects. This didn’t stop Page and Brin from discreetly acquiring 510 Systems and Anthony’s Robots for roughly $50 million in 2011.

As Google’s driverless car program matured, Levandowski seemed to become impatient. Creating a fully functioning driverless car means training a complicated hardware and software system to identify lane lines and red lights and to control the car’s movements. It also means writing software that anticipates thousands of unlikely “edge cases”—hairpin turns, drivers who use hand signals, covered bridges, recumbent bikes, and so on. That work seemed to bore Levandowski. He became increasingly frustrated at Google’s inability to operate its cars on city streets and decided to take matters into his own hands. “Engineers were like, ‘We are totally ready to go,’ and I’m like, ‘Let’s go then. Let’s see whether it’s real or a demo,’ ” Levandowski said in the summer interview.
google  uber  Silicon_valley 
april 2017 by scritic
A deep look at Google's biggest-ever search quality crisis
A really comprehensive look at some of the problems that google has been having with search and with the community in general; his main suggestion is that Google needs to say that the search is its best guess, otherwise, it'll keep offending people.
teaching  google 
april 2017 by scritic
Google's epic legal battle with Uber over self-driving technology, explained
Did uber take engineers from google for tacit knowledge or proprietary knowledge?
Uber  google  teaching  toblog  platformization  Silicon_valley 
april 2017 by scritic
Google Training Ad Placement Computers to Be Offended - The New York Times
The key line is this - both machines and humans are involved:

To train the computers, Google is applying machine-learning techniques — the underlying technology for many of its biggest breakthroughs, like the self-driving car. It has also brought in large human teams (it declined to say how big) to review the appropriateness of videos that computers flagged as questionable.

Essentially, they are training computers to recognize footage of a woman in a sports bra and leggings doing yoga poses in an exercise video safe for advertising and not sexually suggestive content. Similarly, they will mark video of a Hollywood action star waving a gun as acceptable to some advertisers, while flagging a similar image involving an Islamic State gunman as inappropriate.
platformization  youtube  google 
april 2017 by scritic
Forget AT&T. The Real Monopolies Are Google and Facebook. - The New York Times
In the past decade, an enormous reallocation of revenue of perhaps $50 billion a year has taken place, with economic value moving from creators of content to owners of monopoly platforms.

I reached this conclusion from the following statistics: Since 2000, recorded music revenues in the United States have fallen to $7.2 billion per year from $19.8 billion. Home entertainment video revenue fell to $18 billion in 2014 from $24.2 billion in 2006. United States newspaper ad revenue fell to $23.6 billion in 2013 from $65.8 billion in 2000.

And yet, by every available metric, people are consuming more music, video, news and books. During that same period, Google’s revenue grew to $74.5 billion from $400 million.
facebook  google  platformization 
december 2016 by scritic
Facebook and Google make lies as pretty as truth - The Verge
While feed formatting isn’t anything new, platforms like Google AMP, Facebook Instant Articles, and Apple News are also further breaking down the relationship between good design and credibility. In a platform world, all publishers end up looking more similar than different. That makes separating the real from the fake even harder.
google  facebook  polarization  trump 
december 2016 by scritic
Facebook Must Really Suck At Machine Learning | Elad Blog
An engineer says fAcebook can easily fix it through, yes, algorithms.
toblog  facebook  twitter  google  polarization  trump 
november 2016 by scritic
This is how Facebook’s fake-news writers make money - The Washington Post
“Google has more of an incentive to make information reliable,” Carroll noted, because Google’s business is based on providing accurate information to people who are looking for it. Facebook, though, “is about attention, not so much intention.” It’s generally good for Facebook’s business when something goes viral on the site, even if it’s not true.
google  facebook  polarization  platformization 
november 2016 by scritic
This is the real way big business peddles influence in Washington - Vox
"This represents several distinct channels of influence-peddling:

Google's views on policy issues are simply well-known and well-understood by relevant people in Washington thanks to the fact that they are able to spend a lot of money on making them well-known and well-understood.
Google's civically minded work helps make it well-regarded among the general public, so that policy initiatives that have an upside for Google (like unlocking television set-top boxes) play as smart politics in a way that's not the case for widely hated cable companies.
Google is well-regarded in Washington policy circles both inside and outside the government, so influential people are predisposed to hear them out fairly on contentious issues.
Google is genuinely useful to people in the government who are genuinely trying to do good things, which cultivates the mentality that a strong and globally competitive Google is good for the United States of America. It may even be true!"
google  research  platformization  politics 
april 2016 by scritic
The Sadness and Beauty of Watching Google’s AI Play Go | WIRED
Like little else, this path to victory highlights the power and the mystery of the machine learning technologies that underpin Google’s creation—technologies that are already reinventing so many online services inside companies like Google and Facebook, and are poised to remake everything from scientific research to robotics. With these technologies, AlphaGo could learn the game by examining thousands of human Go moves, and then it could master the game by playing itself over and over and over again. The result is a system of unprecedented beauty.
google  artificial_intelligence  machinelearning  public_discourse 
march 2016 by scritic
Google Puts Boston Dynamics Up for Sale in Robotics Retreat - Bloomberg Business
Google acquired Boston Dynamics in late 2013 as part of a spree of acquisitions in the field of robotics. The deals were spearheaded by Andy Rubin, former chief of the Android division, and brought about 300 robotics engineers into Google. Rubin left the company in October 2014. Over the following year, the robot initiative, dubbed Replicant, was plagued by leadership changes, failures to collaborate between companies and an unsuccessful effort to recruit a new leader.
At the heart of Replicant’s trouble, said a person familiar with the group, was a reluctance by Boston Dynamics executives to work with Google’s other robot engineers in California and Tokyo and the unit’s failure to come up with products that could be released in the near term.
Tensions between Boston Dynamics and the rest of the Replicant group spilled into open view within Google, when written minutes of a Nov. 11 meeting and several subsequent e-mails were inadvertently published to an online forum that was accessible to other Google workers. These documents were made available to Bloomberg News by a Google employee who spotted them.
google  robotics  artificial_intelligence 
march 2016 by scritic
How Larry Page’s Obsessions Became Google’s Business - The New York Times
It is perfectly possible in Silicon Valley to be for-profit and hope to make the world better. Sometimes it is called social entrepreneurship; other times, something else.

"Lately, he has talked more about his belief that for-profit companies can be a force for social good and change. During a 2014 interview with Charlie Rose, Mr. Page said that instead of a nonprofit or philanthropic organization, he would rather leave his money to an entrepreneur like Mr. Musk."
siliconvalley  google  public_discourse  platformization 
january 2016 by scritic
Google's plan with Alphabet: build the Bell Labs of the 21st century - Vox
Bell Labs, Google, Alphabet - how the funding of science has changed and how that relates to Google's decision to become Alphabet
platformization  google 
august 2015 by scritic
The Deep Mind of Demis Hassabis — Backchannel — Medium
How big of a boost is it to use Google’s infrastructure?

It’s huge. That’s another big reason we teamed up with Google. We had tons of venture money and amazing backers, but to build the computer infrastructure and engineering infrastructure that Google had would have taken a decade. Now we can do our research much quickly because we can run a million experiments in parallel.

The big leap you are making is not only to dig into things like structured databases but to analyze unstructured information — such as documents or images on the Internet — and be able to make use of them as well, right?

Exactly. That’s where the big gains are going to be in the next few years. I also think the only path to developing really powerful AI would be to use this unstructured information. It’s also called unsupervised learning— you just give it data and it learns by itself what to do with it, what the structure is, what the insights are. We are only interested in that kind of AI.

One of the people you work with at Google is Geoff Hinton, a pioneer of neural networks. Has his work been crucial to yours?

Sure. He had this big paper in 2006 that rejuvenated this whole area. And he introduced this idea of deep neural networks—Deep Learning. The other big thing that we have here is reinforcement learning, which we think is equally important. A lot of what Deep Mind has done so far is combining those two promising areas of research together in a really fundamental way. And that’s resulted in the Atari game player, which really is the first demonstration of an agent that goes from pixels to action, as we call it.
google  artificial_intelligence  public_discourse  platformization 
july 2015 by scritic
Facebook's study of news revealed its plans to be the next top search engine - Vox
Interesting but premature. First of all, building an SVM is cheap so just building one shows nothing. Secondly the Techcrunch article shows that Facebook is creating an index of its own articles, the ones that get posted -- that's not an index that's competing with Google, I don't think. Still, worth reading

"Facebook is building a search engine to rival Google. The bot is a preview of how that search will rank stories. This week, John Constine and Kyle Russell at TechCrunch shared screenshots of a newly discovered way to use Facebook search. The "Add a Link" function lets you post search results to your page that originally come from outside of Facebook."
platformization  facebook  public_discourse  publishing  google 
may 2015 by scritic
Human-level control through deep reinforcement learning : Nature : Nature Publishing Group
The theory of reinforcement learning provides a normative account1, deeply rooted in psychological2 and neuroscientific3 perspectives on animal behaviour, of how agents may optimize their control of an environment. To use reinforcement learning successfully in situations approaching real-world complexity, however, agents are confronted with a difficult task: they must derive efficient representations of the environment from high-dimensional sensory inputs, and use these to generalize past experience to new situations. Remarkably, humans and other animals seem to solve this problem through a harmonious combination of reinforcement learning and hierarchical sensory processing systems4, 5, the former evidenced by a wealth of neural data revealing notable parallels between the phasic signals emitted by dopaminergic neurons and temporal difference reinforcement learning algorithms3. While reinforcement learning agents have achieved some successes in a variety of domains6, 7, 8, their applicability has previously been limited to domains in which useful features can be handcrafted, or to domains with fully observed, low-dimensional state spaces. Here we use recent advances in training deep neural networks9, 10, 11 to develop a novel artificial agent, termed a deep Q-network, that can learn successful policies directly from high-dimensional sensory inputs using end-to-end reinforcement learning. We tested this agent on the challenging domain of classic Atari 2600 games12. We demonstrate that the deep Q-network agent, receiving only the pixels and the game score as inputs, was able to surpass the performance of all previous algorithms and achieve a level comparable to that of a professional human games tester across a set of 49 games, using the same algorithm, network architecture and hyperparameters. This work bridges the divide between high-dimensional sensory inputs and actions, resulting in the first artificial agent that is capable of learning to excel at a diverse array of challenging tasks.
machinelearning  google  artificial_intelligence 
february 2015 by scritic
Hoping Google’s Lab Is a Rainmaker - NYTimes.com
Google’s research arm, Google X, is called the company’s Moonshot Factory. One reason the company picked the word “Moonshot” was to remind people to tackle big problems that may well blow up in their faces.

Last month, after years of promotion, Google ended a test trial of its Internet-connected glasses, called Glass. While the device seemed to have promising commercial applications in hospitals or on factory floors, its first pass at the consumer world was unsuccessful.

The very public failure of Glass points to a bigger question. After patiently abiding a steep increase in research and development spending on efforts that range from biology to space exploration, Wall Street is starting to wonder when — and if — Google’s science projects will pay off.
google  research 
february 2015 by scritic
How long can Google keep investing in outlandish science projects? - Vox
America's largest companies are investing way less in science than they used to. Decades ago, firms like AT&T and IBM ran massive labs where scientists could dream big and pursue the sort of research that won Nobel Prizes — even if it didn’t translate immediately into new products. (AT&T’s Bell Labs famously helped invent the laser and the transistor.) But that’s increasingly a rare sight.

And to see why, just look at Google.
google  research 
february 2015 by scritic
Google Maps Tenth Anniversary | Re/code
A history of google maps -- and how it made its way into google.
google  history  technology  platformization  public_discourse 
february 2015 by scritic
A Prickly Partnership for Uber and Google - NYTimes.com
When Google’s venture capital arm poured more than $250 million into Uber in 2013, it looked like a match made in tech heaven.

Google, with billions of dollars in the bank and house-by-house maps of most of the planet, seemed like the perfect partner for Uber, the hugely popular ride-hailing service.

Bliss between the Internet titan and the hot young company, however, has proved fleeting.

Uber recently announced plans to develop self-driving cars, a longtime pet project at Google. It is also adding engineers who are experts on mapping technology. And the company, based in San Francisco, has been in talks with Google’s advertising archrival, Facebook, to find ways to work together.

Not to be outdone, Google has been experimenting with a ride-sharing app similar to Uber’s, according to Bloomberg. And both companies have long toyed with the idea of offering same-day delivery of items like groceries and other staples.

Continue reading the main story
RELATED COVERAGE

Uber to Open Center for Research on Self-Driving CarsFEB. 2, 2015
Hard-Charging Uber Tries Olive Branch FEB. 1, 2015
Add it all up and the business partners, just two years after that major investment, are “going to war,” as the Bloomberg report put it.
google  uber  platformization  public_discourse 
february 2015 by scritic
Why Google Glass Broke - NYTimes.com
At the time, unknown to anyone outside X, an impassioned split was forming between X engineers about the most basic functions of Google Glass. One faction argued that it should be worn all day, like a “fashionable device,” while others thought it should be worn only for specific utilitarian functions. Still, nearly everyone at X was in agreement that the current prototype was just that: a prototype, with major kinks to be worked out.

Continue reading the main story
There was one notable dissenter. Mr. Brin knew Google Glass wasn’t a finished product and that it needed work, but he wanted that to take place in public, not in a top-secret lab. Mr. Brin argued that X should release Glass to consumers and use their feedback to iterate and improve the design.
agile  google  software 
february 2015 by scritic
Demis Hassabis, Founder of DeepMind Technologies and Artificial-Intelligence Wunderkind at Google, Wants Machines to Think Like Us | MIT Technology Review
Profile of a deeplearning startup acquired by Google

Google snapped up DeepMind shortly after it demonstrated software capable of teaching itself to play classic video games to a super-human level (see “Is Google Cornering the Market on Deep Learning?”). At the TED conference in Vancouver this year, Google CEO Larry Page gushed about Hassabis and called his company’s technology “one of the most exciting things I’ve seen in a long time.”
google  data_science  artificial_intelligence 
december 2014 by scritic
Google Lifts the Turing Award Into Nobel Territory - NYTimes.com
The A.M.Turing Award is often called the Nobel prize of computer science. Now, thanks to Google‘s largesse, it will be a Nobel-level prize financially: $1 million.

The quadrupling of the prize money, announced on Thursday by the Association for Computing Machinery, the professional organization that administers the award, is intended to elevate the prominence and recognition of computer science. The move can be seen as another sign of the boom times in technology.

Computing is increasingly an ingredient in every field, from biology to business. College students are rushing to take computer science courses, encouraged by their parents. It’s not just a skill but a mind-set. Computational thinking is the future, where the excitement and money is. Quants rule.

But the Turing Award celebrates the slower and deeper side of computing. It is given, said Alexander L. Wolf, president of A.C.M. and a professor of computer science at Imperial College London, to the “true pioneers” who are “fundamental contributors to the science and technology of computing.”

Previous recipients have not been household names, except in very geeky households. They did not make fortunes, but they created the underlying insights in mathematics, and in software and hardware design, that helped make personal computers, the Internet, online commerce, social networks and smartphones a reality.

Yet computer science is also a practical, problem-solving discipline. At the announcement event on Thursday morning in New York, Stuart Feldman, vice president of engineering at Google, said he had been impressed by the blend of the theoretical and practical sides of computing in reading over the Turing Award citations since 1966. The prize, he said, recognizes both “the finest of thought and the broadest of impact.”
computer_science  public_discourse  google  data_science  platformization 
november 2014 by scritic
Google Classroom | Hacker News
Long thread on Hacker News about Google classrooms
google  platformization  public_discourse 
october 2014 by scritic
Google and expertise: what to generalize? - Bror's Blog
[Note here that Google's approach to expertise ties in with CS's outlook towards expertise in general]

Where things get stickier, as Willingham points out, is on the matter of expertise. Friedman reports that at Google, "[t]he least important attribute they look for is 'expertise.' . . . Most of the time the nonexpert will come up with the same answer, added Bock, 'because most of the time it’s not that hard.'"

Yikes!

For most organizations at scale, and for most roles, cognitive science is very clear that this is not how expertise works. What actually happens, as Doug Lemov relates in a WSJ editorial about the importance of practice (and Rick Hess and I relate in our book), is that lots of deliberate practice with feedback, often slow and error filled at the start, leads to more and more parts of a complex task becoming automated, embedded in fast, fluent, non-verbal long-term memory.

This makes expert solutions to standard problems increasingly fast and error-free, which is great, where efficiency matters. (Google is "special," though, around efficiency - remember the free food?) Perhaps more important when an area is rapidly changing, this internal automation also frees up our working memory. Working memory is the slower, narrower, flexible, verbal part of our minds that tackles new problems and the hardest problems (including learning new things). By automating more parts of tasks, these no longer tie up working memory, allowing working memory to act at a higher level, on bigger "chunks" of problems, and to try out more types of patterns and solutions because of what's immediately available from long term memory.

So for most organizations, the idea that you can ignore what expertise a person already has on-board for any task of any complexity, and just go for a pure novice who knows how to learn, is a recipe for inefficiency and challenges. Do you really want a "fast learner" trying to diagnose your chest pains? ("Give me a second - working Google now. . . Heart trouble - heart trouble - country western songs - wait for it, almost there. . .") As Lemov says in his editorial, even for new problems, "[r]ote learning and conceptual thinking often feed synergistically on each other, freeing our brain capacity for those tasks that require the maximum amount of attention and creativity."

Obviously, Google is a unique place, with an amazing pool of talent to draw from, and has done so successfully for many years: it would be churlish to suggest "think what they could be if only. . ." What's helpful, though, is to filter some of the comments of Mr. Bock by what we know applies to most minds out there, as we seek to understand how to make things work better elsewhere.

Paying a lot more attention to getting the right repeated levels of practice and feedback, the right motivation, and the right evidence-based design of instruction, can systematically build out the kind of expertise that is both efficient and creative, whether at work or at school.
- See more at: http://brorsblog.typepad.com/brors-blog/2014/04/nature-of-expertise-and-predictions-of-success.html#sthash.q3va118Z.dpuf
google  platformization  expertisevsaggregation 
september 2014 by scritic
The 10 Algorithms That Dominate Our World
Nice list of algorithms. Compare this to the Medium post which refutest this by compiling a diff. list of algorithms that are essentially content-less.
algorithms  facebook  google  platformization  toblog 
july 2014 by scritic
Larry Page's Plan For People Whose Jobs Are Replaced By Tech: Work Less
Machine learning will replace more and more human jobs, leading to vast unemployment in the next few decades, just like 19th-century farmers.

But today’s opiate of the masses is work. It makes people feel needed and occupied. People like vacation, but also like feeling busy. The trick, Page said, will be to convince people to simply work less so we can all have a little work.

Anthropologists have shown that people only need housing, security and opportunity for their children to be happy, Page said — and that doesn’t take much to produce for everyone. We’re overproducing on things not central to happiness, but people need work to be happy. The answer? Redistribute jobs so everyone works a little — just less than before.

“So the idea that everyone needs to work frantically to meet people’s needs is just not true,” Page said. “I do think there’s a problem that we don’t recognize that. I think there’s also a social problem that a lot of people aren’t happy if they don’t have anything to do. So we need to give people things to do. We need to feel like you’re needed, wanted and have something productive to do.”
google  machinelearning  platformization  workers  futurism 
july 2014 by scritic
Volatile and Decentralized: Why I'm leaving Harvard
There is one simple reason that I'm leaving academia: I simply love work I'm doing at Google. I get to hack all day, working on problems that are orders of magnitude larger and more interesting than I can work on at any university. That is really hard to beat, and is worth more to me than having "Prof." in front of my name, or a big office, or even permanent employment. In many ways, working at Google is realizing the dream I've had of building big systems my entire career.

As I've blogged about before, being a professor is not the job I thought it would be. There's a lot of overhead involved, and (at least for me) getting funding is a lot harder than it should be. Also, it's increasingly hard to do "big systems" work in an academic setting. Arguably the problems in industry are so much larger than what most academics can tackle. It would be nice if that would change, but you know the saying -- if you can't beat 'em, join 'em.

The cynical view is that as an academic systems researcher, the very best possible outcome for your research is that someone at Google or Microsoft or Facebook reads one of your papers, gets inspired by it, and implements something like it internally. Chances are they will have to change your idea drastically to get it to actually work, and you'll never hear about it. And of course the amount of overhead and red tape (grant proposals, teaching, committee work, etc.) you have to do apart from the interesting technical work severely limits your ability to actually get to that point. At Google, I have a much more direct route from idea to execution to impact. I can just sit down and write the code and deploy the system, on more machines than I will ever have access to at a university. I personally find this far more satisfying than the elaborate academic process.

Of course, academic research is incredibly important, and forms the basis for much of what happens in industry. The question for me is simply which side of the innovation pipeline I want to work on. Academics have a lot of freedom, but this comes at the cost of high overhead and a longer path from idea to application. I really admire the academics who have had major impact outside of the ivory tower, like David Patterson at Berkeley. I also admire the professors who flourish in an academic setting, writing books, giving talks, mentoring students, sitting on government advisory boards, all that. I never found most of those things very satisfying, and all of that extra work only takes away from time spent building systems, which is what I really want to be doing.
google  academy  corporation  platformization 
july 2014 by scritic
How to Save the News - James Fallows - The Atlantic
The three pillars of the new online business model, as I heard them invariably described, are distribution, engagement, and monetization. That is: getting news to more people, and more people to news-oriented sites; making the presentation of news more interesting, varied, and involving; and converting these larger and more strongly committed audiences into revenue, through both subscription fees and ads. Conveniently, each calls on areas of Google’s expertise. “Not knowing as much about the news business as the newspapers do, it is unlikely that we can solve the problems better than they can,” Nikesh Arora told me. “But we are willing to support any formal and informal effort that newspapers or journalists more generally want to make” to come up with new sources of money.

In practice this involves projects like the ones I’m about to describe, which share two other traits beyond the “distribution, engagement, monetization” strategy that officially unites them. One is the Google concept of “permanent beta” and continuous experimentation—learning what does work by seeing all the things that don’t. “We believe that teams must be nimble and able to fail quickly,” Josh Cohen told me. (I resisted making the obvious joke about the contrast with the journalism world, which believes in slow and statesmanlike failure.) “The three most important things any newspaper can do now are experiment, experiment, and experiment,” Hal Varian said.
google  journalism  platformization 
july 2014 by scritic
Google Will Finance Carnegie Mellon’s MOOC Research – Wired Campus - Blogs - The Chronicle of Higher Education
The grant also figures to strengthen Google’s already-substantial ties to Carnegie Mellon. When the company opened a Pittsburgh outpost, in 2006, its first facilities were on the university’s campus. And a Carnegie Mellon professor, Andrew Moore, was hired as Google Pittsburgh’s director. He spent eight years there, until this April, when he accepted an offer to become dean of Carnegie Mellon’s school of computer science.

Even a list of past recipients of Google Focused Research Awards is telling. Of the 62 projects funded by the grants, six involved researchers at Carnegie Mellon. No other institution had more than four.
google  cmu  moocs  funding 
june 2014 by scritic
(7) How do I make the most of working at Google? - Quora
Learn as much about machine learning as you can, Google is a leader in the field and you won't have the opportunity anywhere else.
google  machinelearning  public_discourse 
june 2014 by scritic
(7) Machine Learning: What are best places to work as a machine learning researcher? - Quora
One of the main reasons why many professors left academia to work at Google Research is because machine learning research is all about the data. And Google has the data, the computers, and the expertise necessary to apply machine learning to those problems. The days of NIPS-y point clouds, MNIST digit recogition, and K-means on simulated data are over. Welcome to the era of Machine Learning being used to looks for the cure for cancer, to search for signs of life in astronomical data, and data-centers being used to learn visual concepts. In other words, many ML researchers would agree is that the exciting part of ML is applying it to real-world problems, and this means scaling must be addressed from day one. Google is the champion of scaling.
machinelearning  siliconvalley  google  microsoft 
june 2014 by scritic
Google’s Push Past Search | TechCrunch
As a recent analysis indicated, Google’s traditional search is not working on mobile as well as it did on the desktop web. Sifting through organic search results on a mobile device is a sub-optimal experience, especially when compared to the push notifications and personalized streams of cards that have made mobile apps from Facebook, Twitter and Tinder so habit-forming and successful.

Google is getting well ahead of its mobile organic search problem, especially on Android where it has full control of the end-to-end mobile experience. Google has strung together push notifications, a stream of predictive answers and an answer box in an attempt to answer a search query three times before showing organic search results.
google  platformization 
june 2014 by scritic
(7) Could two smart computer science Ph.D students create a search engine that unseats Google? How vulnerable is Google to this possibility? - Quora
There are also some scale effects, like doing a good job at query suggestions as you type relies on having access to data from a lot of users' previous queries. This is also true of things like spelling correction. Data from Google toolbar (the complete browsing history of most toolbar users) feeds back into ranking as well.

One other thing that's different now is that the web is much bigger and so the fixed cost of crawling it is outside the scope of a university research budget (or most startups).

Another issue with results ranking is that Google has played an arms race with spammers and SEO people for the last 10 years, and in each round, Google's algorithm advances and spammers' techniques advance. Someone starting from scratch right now has to go up against all of these battle-hardened spammers without any experience on the other side.

A final reason is that Google has pulled a lot of the best computer science students out of academia and they work at Google instead of doing PhDs on information retrieval, so the research that would lead to this is happening inside Google if it's happening anywhere.
google  websearch 
june 2014 by scritic
(7) What parts of Google software engineering culture do you use and propagate after you left Google? - Quora
Automate testing to scale your code. Google has an extremely strong culture of unit testing, for which "Testing on the Toilet" is but one illustrative example. Nearly every code change I worked on was accompanied by a unit test, and code reviewers would rigorously check for them. It made developing a given change slower, but it also meant that hundreds or thousands of engineers could scalably make changes to the same parts of the codebase without sacrificing too much quality or reliability. In the same way that Google invested in shared tools, it would also heavily in shared testing frameworks and educating people in best testing practices to make writing tests easier
google  siliconvalley  agile 
june 2014 by scritic
Google’s Next Phase in Driverless Cars: No Steering Wheel or Brake Pedals - NYTimes.com
Mr. Brin said the change in Google’s car strategy did not mean that the company was giving up on its ultimate goal of transforming modern transportation.

“Obviously it will take time, a long time, but I think it has a lot of potential,” he said. “Self-driving cars have the potential to drive in trains much closer together and, in theory, in the future at much higher speeds.

“There is nothing to say that once you demonstrate the safety, why can’t you go 100 miles per hour?”
technology  google  cars 
may 2014 by scritic
Why Facebook is Becoming Like Google+ | MIT Technology Review
At least, that’s an implication of what Facebook CEO Mark Zuckerberg has begun saying about his strategy. In an interview with the New York Times last month he explained that his company is now working on “unbundling the big blue app”. That means taking what Facebook has been up to now – a website or app you visit to see a feed of updates from friends and post your own – and splintering it into many separate and more specialized services.

This strategy isn’t entirely new. It’s how the highly successful Instagram and Facebook Messenger apps work today. But Zuckerberg has now made it clear that he’s committing to moving away from what Facebook used to be. What was the core of Facebook is becoming a kind of login service, which provides your digital identity and some social connections to a suite of single-purpose apps and services.

That’s what Google+ is today. Although you can use the service like the original Facebook, it seems relatively few people do (we lack objective data) and Google looks to be retreating from that idea. In practice, Google+ is a kind of unified identity service that links up your actions inside Google’s various products, from your YouTube activity to your Hangouts video chats to app, movie and music purchases made on Android mobile devices.

That’s also what Zuckerberg seems to be working towards with his “unbundling” strategy. The Facebook cofounder told the New York Times that this approach was needed because mobile devices make people prefer “single-purpose, first class experiences” rather than multi-functional ones like the one Facebook has traditionally offered.

Another reason could be that the Facebook model of social networking has run its course. Some people that have used the service for years find the newsfeed a turn off these days. The clutter of many different friends, relatives and acquaintances doesn’t produce compelling content, and makes a difficult audience to post content for.

People don’t seem interested in pruning or categorizing their friend lists to address the problem. Google+ has failed to catch on as a social destination despite putting its “circles” feature to do that at the core of the service. Further evidence that people aren’t so interested in News feed any more comes from the lack of success of two recent Facebook experiments, Paper and Home, both of which essentially just repackage News feed content.

There are benefits to end users in the unbundled strategy that Facebook and Google have seemingly agreed on. Google’s disparate services now work better together. Facebook’s adoption of the strategy means that popular services that it acquires don’t necessarily get shut down or forcibly integrated into its existing products. Instagram and WhatsApp, for example, continue much as they did before.

Of course, an unbundled future also helps Facebook and Google’s efforts to make money off of data on human behavior.
facebook  google  Web 
may 2014 by scritic
How Google money is helping turn the political right against strong copyrights - Vox
That has changed. Today, Silicon Valley companies — especially Google — give generously to right-leaning think tanks that publish writing skeptical of copyright protection. Google has donated to almost every right-leaning think tank in Washington, including the R Street Institute, the Cato Institute, the Mercatus Center, the American Enterprise Institute, and the Heritage Foundation. So if you're a right-leaning copyright skeptic, it's easy to find organizations to publish your work.
google  copyright  ip 
may 2014 by scritic
Google Maps Has Forsaken Us | TechCrunch
Or, as Matt Haughey puts it, on Medium: “That’s Google trying to be helpful, but not being actually helpful: and in reality, confusing me … Google indexes all of the world’s information … Why doesn’t Google bring any of that additional information into searches and results within the Maps app?”

Good question, Matt.

I suppose one possible culprit is Google’s increasing use of large-scale distributed deep networks. (PDF) Such networks are, by their nature, black boxes whose outcomes cannot be traced to any particular algorithm or line of code. Maybe I’m just unlucky enough to be an outlier who keeps running into anomalous outcomes, which have actually improved for the vast majority of users.

Maybe. But I doubt it. Instead I strongly suspect the answer is strongly related to some kind of byzantine system of feudal silos and fiefdoms within the Google empire. Because when I’m not tooling around the East Bay in search of tasty In-N-Out Burger, I’m busy writing web and smartphone apps for HappyFunCorp, and in that capacity I keep running into kind of astonishing Google-product bug reports like:
google  public_discourse  machinelearning 
may 2014 by scritic
Google Stops Mining Education Gmail And Google Apps Accounts For Ad Targeting | TechCrunch
Google will no longer scan student and teacher Gmail messages or use data from Apps for Education for advertising purposes, the company told the WSJ today. The move comes after Google’s use of data from its education products came under fire by students and others during a court case last year that claimed the scanning violated user privacy rights.
google  privacy  public_discourse  higher_ed 
may 2014 by scritic
(5) Google: How does the Google "20 percent time" really work? - Quora
Interesting replies...

While I was there (2010), the group I was in (AdWords Campaign Management) discouraged any 20% project that wasn't a feature for the group's product. Actually, they discouraged it period. But they never said this explicitly, they would always do it in a roundabout way.

If you ever decided to take on a 20% project, they wouldn't say no or block you explicitly, you would just start feeling the grumbling and the side effects of going against their wishes everywhere you'd go.
google  management  siliconvalley 
april 2014 by scritic
(2) Google: Is Google overreacting to the rise of Facebook? - Quora
An interesting thread: read more.

Google cannot execute on its mission statement if an ever-growing portion of the world’s information is inaccessible to its systems and algorithms.
facebook  google  siliconvalley 
april 2014 by scritic
The A/B Test: Inside the Technology That's Changing the Rules of Business | Wired Business | Wired.com
A/B testing was a new insight in the realm of politics, but its use on the web dates back at least to the turn of the millennium. At Google—whose rise as a Silicon Valley powerhouse has done more than anything else to spread the A/B gospel over the past decade—engineers ran their first A/B test on February 27, 2000. They had often wondered whether the number of results the search engine displayed per page, which then (as now) defaulted to 10, was optimal for users. So they ran an experiment. To 0.1 percent of the search engine’s traffic, they presented 20 results per page; another 0.1 percent saw 25 results, and another, 30.

Due to a technical glitch, the experiment was a disaster. The pages viewed by the experimental groups loaded significantly slower than the control did, causing the relevant metrics to tank. But that in itself yielded a critical insight—tenths of a second could make or break user satisfaction in a precisely quantifiable way. Soon Google tweaked its response times and allowed real A/B testing to blossom. In 2011 the company ran more than 7,000 A/B tests on its search algorithm. Amazon.com, Netflix, and eBay are also A/B addicts, constantly testing potential site changes on live (and unsuspecting) users.

Today, A/B is ubiquitous, and one of the strange consequences of that ubiquity is that the way we think about the web has become increasingly outdated. We talk about the Google homepage or the Amazon checkout screen, but it’s now more accurate to say that you visited a Google homepage, an Amazon checkout screen. What percentage of Google users are getting some kind of “experimental” page or results when they initiate a search? Google employees I spoke with wouldn’t give a precise answer—”decent,” chuckles Scott Huffman, who oversees testing on Google Search. Use of a technique called multivariate testing, in which myriad A/B tests essentially run simultaneously in as many combinations as possible, means that the percentage of users getting some kind of tweak may well approach 100 percent, making “the Google search experience” a sort of Platonic ideal: never encountered directly but glimpsed only through imperfect derivations and variations.

Still, despite its widening prevalence, the technique is not simple. It takes some fancy technological footwork to divert user traffic and rearrange a site on the fly; segmenting users and making sense of the results requires deep knowledge of statistics. This is a barrier for any firm that lacks the resources to create and adjudicate its own tests. In 2006 Google released its Website Optimizer, which provided a free tool for anyone who wanted to run A/B tests. But the tool required site designers to create full sets of code for both A and B—meaning that nonprogrammers (marketing, editorial, or product people) couldn’t run tests without first taxing their engineers to write multiple versions of everything. Consequently there was a huge delay in getting results as companies waited for the code to be written and go live.

In 2009 this remained a problem in need of a solution. After the Obama campaign ended, Siroker was left amazed at the efficacy of A/B testing but also at the paucity of tools that would make it easily accessible. “The thought of using the tools we used then made me grimace,” he says. By the end of the year, Siroker joined forces with another ex-Googler, named Pete Koomen, and they launched a startup with the goal of bringing A/B tools to the corporate masses, dubbing it Optimizely. They signed up their first customer by accident. “Before we even spent much time working on the product,” Siroker explains, “I called up one of the guys from the Obama campaign, who had started up a digital marketing firm. I told him what I was up to, and about 20 minutes in, he suddenly said, ‘Well, that sounds great. Send me an invoice.’ He thought it was a sales call.”

The pair had made a sale, but they still didn’t have a product. So Siroker and Koomen started coding. Unlike the earlier A/B tools, they designed Optimizely to be usable by nonprogrammers, with a powerful graphical interface that lets clients drag, resize, retype, replace, insert, and delete on the fly. Then it tracks user behavior and delivers results. It’s an intuitive platform that offers the A/B experience, previously the sole province of web giants like Google and Amazon, to small and midsize companies—even ones without a hardcore engineering or testing team.
google  ABTesting  research 
march 2014 by scritic
The business and politics of search engines: A comparative study of Baidu and Google’s search results of Internet events in China
Despite growing interest in search engines in China, relatively few empirical studies have examined their sociopolitical implications. This study fills several research gaps by comparing query results (N = 6320) from China’s two leading search engines, Baidu and Google, focusing on accessibility, overlap, ranking, and bias patterns. Analysis of query results of 316 popular Chinese Internet events reveals the following: (1) after Google moved its servers from Mainland China to Hong Kong, its results are equally if not more likely to be inaccessible than Baidu’s, and Baidu’s filtering is much subtler than the Great Firewall’s wholesale blocking of Google’s results; (2) there is low overlap (6.8%) and little ranking similarity between Baidu’s and Google’s results, implying different search engines, different results and different social realities; and (3) Baidu rarely links to its competitors Hudong Baike or Chinese Wikipedia, while their presence in Google’s results is much more prominent, raising search bias concerns. These results suggest search engines can be architecturally altered to serve political regimes, arbitrary in rendering social realities and biased toward self-interest.
research  internet  google 
february 2014 by scritic
Are the robots about to rise? Google's new director of engineering thinks so… | Technology | The Observer
Google has bought almost every machine-learning and robotics company it can find, or at least, rates. It made headlines two months ago, when it bought Boston Dynamics, the firm that produces spectacular, terrifyingly life-like military robots, for an "undisclosed" but undoubtedly massive sum. It spent $3.2bn (£1.9bn) on smart thermostat maker Nest Labs. And this month, it bought the secretive and cutting-edge British artificial intelligence startup DeepMind for £242m.

And those are just the big deals. It also bought Bot & Dolly, Meka Robotics, Holomni, Redwood Robotics and Schaft, and another AI startup, DNNresearch. It hired Geoff Hinton, a British computer scientist who's probably the world's leading expert on neural networks. And it has embarked upon what one DeepMind investor told the technology publication Re/code two weeks ago was "a Manhattan project of AI". If artificial intelligence was really possible, and if anybody could do it, he said, "this will be the team". The future, in ways we can't even begin to imagine, will be Google's.
google  research  artificial_intelligence  machinelearning  from twitter_favs
february 2014 by scritic
Google's Hybrid Approach to Research
The challenge in organizing R&D is great because CS is an increasingly broad and diverse field. It combines aspects of mathematical reasoning, engineering methodology, and the empirical approaches of the scientific method. The empirical components are clearly on the upswing, in part because the computer systems we construct have become so large that analytic techniques cannot properly describe their properties, because the systems now dynamically adjust to the hard-to-predict needs of a diverse user community, and because the systems can learn from vast data sets and large numbers of interactive sessions that provide continuous feedback.

We have also noted that CS is an expanding sphere, where the core of the field (Theory, Operating Systems, etc.) continues to grow in depth, while the field keeps expanding into neighboring application areas. Research results come not only from universities, but also from companies, large and small. The way that research results are disseminated is also evolving and the peer-reviewed paper is under threat as the dominant dissemination method. Open source releases, standards specifications, data releases, and novel commercial systems that set new standards upon which others then build, are increasingly important.

Because of the time-frame and effort involved, Google's approach to research is iterative and usually involves writing production, or near-production, code from day one. Elaborate research prototypes are rarely created, since their development delays the launch of improved end-user services. Typically, a single team iteratively explores fundamental research ideas, develops and maintains the software, and helps operate the resulting Google services - all driven by real-world experience and concrete data. This long-term engagement serves to eliminate most risk to technology transfer from research to engineering. This approach also helps ensure that research efforts produce results that benefit Google's users, by allowing research ideas and implementations to be honed on empirical data and real-world constraints, and by utilizing even failed efforts to gather valuable data and statistics for further attempts.
research  google  computer_science  platformization 
february 2014 by scritic
Why Did Google Pay $400 Million for DeepMind? | MIT Technology Review
In an interview last month, before the DeepMind acquisition, Peter Norvig, a director of research at Google, estimated that his company already employed “less than 50 percent but certainly more than 5 percent” of the world’s leading experts in machine learning, the wider discipline of which deep learning is the cutting edge.

Companies like Google expect deep learning to help them create new types of products that can understand and learn from the images, text, and video clogging the Web. And to a significant degree, leading academic scientists have embraced Silicon Valley, where they can command teams of engineers instead of students and have access to the largest, most interesting data sets. “It’s a combination of the computing resources we have and the headcounts we can offer,” Norvig said. “At Google, if you want a copy of the Web, well, we just happen to have one sitting around.”

Last year, Google also grabbed renowned University of Toronto deep-learning researcher Geoff Hinton and a passel of his students when it acquired Hinton’s company, DNNresearch. Hinton now works part-time at Google. “We said to Geoff, ‘We like your stuff. Would you like to run models that are 100 times bigger than anyone else’s?’ That was attractive to him,” Norvig said.
research  machinelearning  google 
february 2014 by scritic
(1) Google: Why is machine learning used heavily for Google's ad ranking and less for their search ranking? - Quora
410

Edmond Lau, Ex-Google Search Quality Engineer
Votes by Alon Amit, Former Google Group Product Manager, Jeremy Hoffman, Google software engineer (search quality), Jackie Bavaro, Google PM for 3 years, Saikat Bhadra, Former Googler. Worked in Commerce team and hel..., Piaw Na, Worked at Google, and 404 more.
From what I gathered while I was there, Amit Singhal, who heads Google's core ranking team, has a philosophical bias against using machine learning in search ranking. My understanding for the two main reasons behind this philosophy is:

In a machine learning system, it's hard to explain and ascertain why a particular search result ranks more highly than another result for a given query. The explainability of a certain decision can be fairly elusive; most machine learning algorithms tend to be black boxes that at best expose weights and models that can only paint a coarse picture of why a certain decision was made.
Even in situations where someone succeeds in identifying the signals that factored into why one result was ranked more highly than other, it's difficult to directly tweak a machine learning-based system to boost the importance of certain signals over others in isolated contexts. The signals and features that feed into a machine learning system tend to only indirectly affect the output through layers of weights, and this lack of direct control means that even if a human can explain why one web page is better than another for a given query, it can be difficult to embed that human intuition into a system based on machine learning.


Rule-based scoring metrics, while still complex, provide a greater opportunity for engineers to directly tweak weights in specific situations. From Google's dominance in web search, it's fairly clear that the decision to optimize for explainability and control over search result rankings has been successful at allowing the team to iterate and improve rapidly on search ranking quality. The team launched 450 improvements in 2008 [1], and the number is likely only growing with time.

Ads ranking, on the other hand, tends to be much more of an optimization problem where the quality of two ads are much harder to compare and intuit than two web page results. Whereas web pages are fairly distinctive and can be compared and rated by human evaluators on their relevance and quality for a given query [2], the short three- or four-line ads that appear in web search all look fairly similar to humans. It might be easy for a human to identify an obviously terrible ad, but it's difficult to compare two reasonable ones:


Branding differences, subtle textual cues, and behavioral traits of the user, which are hard for humans to intuit but easy for machines to identify, become much more important. Moreover, different advertisers have different budgets and different bids, making ad ranking more of a revenue optimization problem than merely a quality optimization problem. Because humans are less able to understand the decision behind an ads ranking decision that may work well empirically, explainability and control -- both of which are important for search ranking -- become comparatively less useful in ads ranking, and machine learning becomes a much more viable option.

--------
[1] http://googleblog.blogspot.com/2...
[2] http://googleblog.blogspot.com/2...

Suggest Edits
machinelearning  research  google 
february 2014 by scritic
Google Expands Role In Digital Education, Teams Up With edX To Build A YouTube For Free Online Courses | TechCrunch
With today’s partnership, edX is expanding that mission, as its partnership with Google will enable any institution or business to post courses on MOOC.org and presumably open the doors to public access of edX’s content. It will also offer a more diverse range of content from non-profit institutions to higher education institutions and businesses, edX President Anant Agarwal said.

However, while edX has been building out its own open-source platform, Google said in its blog post today that it will be hosting MOOC.org in its own cloud and — not only that — it will also be lending a team of developers to the partnership to help develop Open edX, the organization’s aforementioned open-source MOOC platform.

In turn, for Google, its partnership with edX represents yet another step in its own foray into digital education — one that includes Google Play for Education, Google Apps for Education, and the launch of its own open-source course building platform, Course Builder, in September of last year.

In its blog post today, Google explained that, over the last year, Course Builder has served as as an experiment in how to build and optimize an open, digital course platform at scale. Since launching, Google said that the platform had become host to a variety of courses on “everything from game theory to philanthropy” and is now being used by both universities and non-profits alike to experiment with MOOC content and “make education more accessible … enabling educators to easily teach at scale on top of its cloud services.”

As to Google’s role in the MOOC movement and digital education, the company says that it will continue to “make contributions to the online education space, the findings of which will be shared directly to the online education community and the Open edX platform.” Of course, this partnership, like the Open Education Alliance, is still a work in progress. Details on what type of content, how it will be structured and just what businesses and teachers will be able to get out of the platform remains unclear.

Google did, however, choose to comment on the state of MOOCs and their role in the future of online education, saying, “our industry is in the early stages of MOOCs, and lots of experimentation is still needed to find the best way to meet the educational needs of the world. An open ecosystem with multiple players encourages rapid experimentation and innovation, and we applaud the work going on in this space today.”

It’s still far too early to say what the ramifications are for Google’s partnership with edX and its participation in the Open Education Alliance. Certainly, if both can work towards promoting a “new meritocracy in higher education,” as my colleague Greg surmised in his recent post, they could collectively bring enormous changes to higher education and the scope and breadth of education as it exists today.

Certainly, Udacity, edX and Google seem hellbent on recalibrating the focus of higher education and learning content, focusing on content that will help students learn how to become part of a modern, and increasingly more technical, global workforce. Whether the increasing role of Google and other tech companies in the educational landscape will be welcomed by academia is one thing, however, at the very least, these two experiments could serve to boost the profile of MOOC-style education, particularly of edX itself. It also seems to indicate the increasing likelihood that, whoever should win the battle to become the world’s largest open course platform, Google will be there to lend a hand — and share a piece of the pie.
MOOCS  google  edx  public_discourse 
february 2014 by scritic

Copy this bookmark:



description:


tags: