Top all-time · #183
The latest research from Google
Company blog · googleresearch.blogspot.comaidata
The latest research from Google is the company blog at googleresearch.blogspot.com, writing about ai and data. Since 2007, 249 different Hacker News readers have linked it 368 times, most often “Nearly all binary search and merge sort implementations are broken (2006)”. It is still writing: the last post was on 2026-10-27, and you can follow it by feed.
Rarely recommended · 2 of 40 sampled comments recommend it?
Why HN links it
HN readers link the Google AI Blog as the primary source for Google research announcements such as AlphaGo, TensorFlow, HDR+ and SyntaxNet, and cite its posts on binary search bugs and hiring.
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- Nearly all binary search and merge sort implementations are broken (2006)27 people · 27 mentions · first linked 2009-05-22
- Hiring: The Lake Wobegon Strategy (2006)11 people · 12 mentions · first linked 2007-06-06
- All Our N-gram are Belong to You8 people · 8 mentions · first linked 2007-04-17
- A picture is worth a thousand words: building a natural description of images6 people · 6 mentions · first linked 2014-11-18
- Large-scale graph computing at Google6 people · 6 mentions · first linked 2009-06-15
- Inceptionism: Going Deeper into Neural Networks6 people · 6 mentions · first linked 2015-06-18
- AlphaGo: Mastering the Ancient Game of Go with Machine Learning6 people · 6 mentions · first linked 2016-01-27
- CE+T Power wins Google's $1M high density inverter challenge6 people · 6 mentions · first linked 2016-03-01
- Speed Matters – Google Research Blog5 people · 5 mentions · first linked 2009-06-24
- Teach Yourself Deep Learning with TensorFlow and Udacity5 people · 5 mentions · first linked 2016-01-21
What Hacker News says?
Somehow, someone always forgets that engineering is about trade-offs, and so every few years we can an indignant series of articles about how stupid and ignorant today's programmers are and how we should all go back to the same processes that they called stupid and ignorant 4-5 years ago. Relying on reimplementation, copy-paste, npm shrinkwrap, or various other ways of importing third-party code into your repository…
Comment linking /2006/06/extra-extra-read-all-about-it-nearly.html
on “NPM and Left-Pad: Have We Forgotten How to Program?”Submitted Announcing SyntaxNet: The World’s Most Accurate Natural Language Parser as a story · 1083 points
HN discussionFor those looking for more context, here's the Google research blog post: http://googleresearch.blogspot.com/2015/11/tensorflow-google...
Comment linking /2015/11/tensorflow-googles-latest-machine_9.html
on “TensorFlow: open-source library for machine intelligence”Submitted Inceptionism: Going Deeper into Neural Networks as a story · 797 points
HN discussionThere is also a substantive blog post at http://googleresearch.blogspot.com/2016/01/alphago-mastering... , via https://news.ycombinator.com/item?id=10981729 .
Comment linking /2016/01/alphago-mastering-ancient-game-of-go.html
on “Google achieves AI 'breakthrough' by beating Go champion”Submitted Lens Blur in the new Google Camera app as a story · 644 points
HN discussionI don't think you appreciate how much of this is good algorithms, and how little you need sheer computing power to get good results. If you look at http://googleresearch.blogspot.com/2016/01/alphago-mastering... you'll find that Google's estimate of the strength difference between the full distributed system and their trained system on a single PC is around 4 professional dan. Let's suppose that squeezing it from a…
Comment linking /2016/01/alphago-mastering-ancient-game-of-go.html
on “AlphaGo beats the world champion Lee Sedol in first of five matches”Submitted Computer, Respond to This Email as a story · 562 points
HN discussionSorry, you're off base a bit. The Atari system did use a Deep Neural Network / Reinforcement algorithm, but as the original poster was trying to point out, the rules of Go were very much hard coded into AlphaGo. From what this [1] says, multiple DNNs are learning how to traverse Monte Carlo trees of Go games. The reinforcement piece comes in choosing which of the Go players is playing the best games. While the highe…
Comment linking /2016/01/alphago-mastering-ancient-game-of-go.html
on “AlphaGo beats the world champion Lee Sedol in first of five matches”Submitted DeepMind moves to TensorFlow as a story · 420 points
HN discussionDeep Blue: Massive search + Hand-coded search heuristics + Hand-coded board position evaluation heuristics [1] AlphaGo: Search via simulations (Monte Carlo Tree Search) + Learned search heuristics (policy networks) + Learned patterns (value networks) [2] Human strongholds seem to be our ability to learn search heuristics and complex patterns. We can perform some simulations but not nearly as extensively as what mach…
Comment linking /2016/01/alphago-mastering-ancient-game-of-go.html
on “AlphaGo beats the world champion Lee Sedol in first of five matches”Submitted How Google Translate squeezes deep learning onto a phone as a story · 403 points
HN discussionI've tried something similar once. I got asked about the binary search algorithm. After some high level discussion on the idea behind it, I told the interviewer I hope I won't have to implement it. The algorithm is known for being easy to get wrong [1]. Nevertheless the interviewer asked me to write it down on paper (using python). I've done so, but made a few mistakes. I still got the job offer, so the mistakes wer…
Comment linking /2006/06/extra-extra-read-all-about-it-nearly.html
on “I will not do a tech interview”Submitted Building a deeper understanding of images as a story · 220 points
HN discussionIf anyone wants to send this to their non-dev friends, here's the write-up I sent to mine! https://medium.com/@stripenight/seeing-how-computers-might-t... --- tldr: To figure out how computers "think", Google asked one of its artificial intelligence algorithms to look at clouds and draw the things it saw! There's this complex Artificial Intelligence algorithm called a neural network ( https://en.wikipedia.org/wiki/A…
Comment linking /2015/06/inceptionism-going-deeper-into-neural.html
on “Inceptionism: Going Deeper into Neural Networks”Submitted Machine Learning in the Cloud, with TensorFlow as a story · 218 points
HN discussionWhy would I give my time for free to Google? Because my entire career- well before I started working here- has been dependent on things that Google has given to me for free. Like Google Search. Search helped me learn to run linux clusters effectively (it was far better than AltaVista for searching for specific error messages) which ensured I had a job, even in the dotcom busts. It helped me learn python, which also…
Comment linking /2012/12/millions-of-core-hours-awarded-to.html
on “Google’s “20% time,” which brought you Gmail and AdSense, is now as good as dead”Submitted Announcing TensorFlow 0.8 – now with distributed computing support as a story · 214 points
HN discussionEven if you abstract away the indexes, when working with datasets that large you have to worry about whether the standard implementation makes the same class of error. For example, using Java's binarySearch on arrays of length over 2^30 was broken until 2006[0][1]. [0]: http://googleresearch.blogspot.com/2006/06/extra-extra-read-... [1]: http://bugs.java.com/bugdatabase/view_bug.do?bug_id=5045582
Comment linking /2006/06/extra-extra-read-all-about-it-nearly.html
on “Gangnam Style breaks YouTube viewer count”Submitted Google voice search: faster and more accurate as a story · 188 points
HN discussionThat's not how it works. Here's the google research blog post about it. http://googleresearch.blogspot.com/2014/10/hdr-low-light-and...
Comment linking /2014/10/hdr-low-light-and-high-dynamic-range.html
on “Google Nexus 5x and 6P”Submitted Low Light and High Dynamic Range photography in the Google Camera App as a story · 175 points
HN discussionHere's the google research blog post about it. http://googleresearch.blogspot.com/2014/10/hdr-low-light-and...
Comment linking /2014/10/hdr-low-light-and-high-dynamic-range.html
on “Google Nexus 5x and 6P”Submitted Running your models in production with TensorFlow Serving as a story · 170 points
HN discussionWhat about binary search? http://googleresearch.blogspot.com/2006/06/extra-extra-read-...
Comment linking /2006/06/extra-extra-read-all-about-it-nearly.html
on “Judge Alsup learns to code and schools Oracle”Submitted Nearly all binary search and merge sort implementations are broken (2006) as a story · 148 points
HN discussionDial-up internet involves converting bits into sound (both in audible and ultrasonic spectrums) and then back on the other end of a phone line. The Google Tone extension ( http://googleresearch.blogspot.com/2015/05/tone-experimental... ) uses both ultrasonic sound and an audible DTMF-based codec. Audio tones are also used for setting up Google OnHub routers; the router emits a tone and the Google On setup app listen…
Comment linking /2015/05/tone-experimental-chrome-extension-for.html
on “Beware of ads that use inaudible sound to link your phone, TV, tablet, and PC”Submitted Building a natural description of images as a story · 146 points
HN discussionGoogle actually already does something along these lines: http://googleresearch.blogspot.com/2015/10/improving-youtube...
Comment linking /2015/10/improving-youtube-video-thumbnails-with.html
on “What convolutional neural networks look at when they see nudity”Submitted Lessons learned while protecting Gmail as a story · 143 points
HN discussion
