Profile cover photo
Profile photo
Пётр Минин
Пётр's posts

Post has attachment
Обнаруживаешь себя в случайной точке планеты на Google Street View, и надо понять, где ты. Весьма познавательно и интересно.

+Gmail, what's the point in displaying "approximate" number of search results? Like "1-20 of about 117" when the real number is 50 or "1-20 of about 63" when the real number is 396. It's not informative at all and may even be misleading. Quickness of response is good and well but not when it's achieved at such a cost.
Also, please tell whoever maintains the Russian localization that "почти 63" is not a correct translation of "about 63", it says "almost 63".

+Gmail, there's a bug in your Android app. After I open it and then navigate back from it I always end up on the home screen, no matter what I was doing when opening Gmail. It happens when I open it from a notification or the recent app list, starting from launcher is OK. Futhermore, it shows the default home screen although I have a custom one set as default. Isn't that a bug? My Android is 4.0.4, Xperia P.

Post has shared content
Круто! Если такие интерактивные штуки ещё не применяются на уроках астрономии в школах, то обязательно должны. Только так можно почувствовать масштаб. И благодаря отметке "Один световой год" это словосочетание может для кого-то перестать быть пустым звуком или синонимом "дофига далеко". Ну и просто можно немного почувствовать себя в контексте галактики.
What do 100,000 stars look like? 

The new Chrome Experiment “100,000 Stars” tries to answer that by visualizing Earth’s immediate stellar neighborhood using WebGL, CSS3D, and Web Audio. Take a tour through the neighborhood, or zoom around on your own, and marvel at the vast scale & beauty of our galaxy. If you’d like to read more about this experiment, head over to the Chrome Blog:

Check out "100,000 Stars" here:

Post has attachment
Это должен знать каждый, lol :)

Post has shared content
Гугл построил нейронную сеть на 16000 ядрах ЦП и более чем с милиардом синапсов, чтобы распознавать объекты (произвольных типов, без label'ов) на видео с Ютюба.
The +Research at Google team has recently been working on new approaches to large-scale machine learning (the technology that enables things like speech recognition). Today's machine learning technology is difficult to adapt to new uses, because it relies on "labeled data"—for example, to build a system that can distinguish between pictures of cars and motorcycles, we'd have to have tens of thousands of pictures that have already been labeled as “car” or “motorcycle” to train the system. 

We've been exploring an alternative approach that uses unlabeled data. This works by building artificial neural networks, which loosely simulate the brain's learning processes. We developed a distributed computing infrastructure for training large-scale neural networks. Then, we took an artificial neural network and spread the computation across 16,000 CPU cores and trained models with more than 1 billion connections (a far greater number than has been done in the past). We showed our network 10 million stills from YouTube videos to see what it might learn. 

Funnily enough, one of our artificial neurons learned to respond strongly to pictures of cats. This network had never been told what a cat was, nor was it given a single image labeled as a cat. Instead, it “discovered” what a cat looked like by itself from only unlabeled YouTube stills. 

This research is exciting in part because it's a step toward developing machine learning systems that scale well. We're also working on applying this approach to other areas besides image recognition, which could eventually mean that the tools you use every day work better, faster and smarter.

We’re reporting on these experiments at ICML this week. You can find the full paper on our research site: (PDF) or read our blog post for more details. 

+Gmail and +Google+, one more idea. I communicate with people in different languages but my name always displays the same for them. So I have to choose whether I want it to be in English which can be read by most but looks strange when I write to my compatriots or if I want it in Russian which is natural but unsuitable for foreigners. So I wish there were an option to save my name in several languages and to have an appropriate variant displayed based on my message language in Gmail or based on the reader's language settings in G+. In G+ there is an option to enter "other names" but I guess they are used for search purposes only, please correct me if I'm wrong. Thank you.

+Gmail, you really should add an ability to create custom views where I can include, for example, Inbox minus a specific tag plus Sent Mail. Currently I have to use All Mail all the time in order to see sent mail in the list and I have a tag for automatic notification emails that I would like to filter out. Thank you.

Post has shared content
Для тех, кто знаком с Доктром Кто :D

Post has attachment
A solution for some problems with #YouTube.
Wait while more posts are being loaded