Google Brain

from Wikipedia, the free encyclopedia

Google Brain is the deep learning division (field of machine learning and artificial intelligence ) of Google LLC . Its headquarters are in Mountain View , California , with additional offices in Cambridge , London , Montreal , New York City , San Francisco , Toronto and Zurich .

history

The project was first introduced in 2011 as a part-time research project of the Google employees Jeff Dean , Greg Corrado and professor of Stanford University Andrew Ng started. The project first received a lot of attention in June 2012, when a computer cluster made up of 16,000 computers, which was supposed to simulate the human brain, first recognized a cat on the basis of YouTube images.

Actions and Research

Health and Death Prediction

Google Brain tries to determine the time of death of patients with the help of deep learning. For this purpose, 216,221 files from patients who had been hospitalized for at least 24 hours were examined by the system. About 46 million data points were taken from the files and the associated doctor's reports. The research team received the data from the hospitals " University of California San Francisco Medical Center " and " University of Chicago Medicine ". This enables the software to predict a death rate in the hospital, a possible readmission within 30 days or even an extended length of stay. This means that doctors can make decisions much earlier. In addition, thanks to the artificial neural network , the software is capable of learning and has so far performed better than comparable software solutions.

Computer evaluation of the quality and aesthetics of media files

In December 2017, Google Brain presented the Neural Image Assessment (German: neural image assessment) Google NIMA , which is able to evaluate images according to their aesthetics. This uses a deep convolutional neural network and is able to give a subjective assessment of the images. It was very similar to the decisions made by humans and is the most extensive and advanced software in this area to date. It can help people with their search and filter larger media data.

robotics

In 2016, Google Brain worked with Google X to do research on self-learning robots . In 2017, Google set the three goals for new robot learning. These should be achieved through reinforcement learning, through their own interaction with objects and through human demonstration. The robots should also be able to solve more complex tasks than industrial robots.

Android and other popular Google software

Some of the technologies researched and developed are already used in Google products and the Android operating system. These include, for example, voice recognition, photo searches on Google+ and video recommendations on YouTube .

Language processing

In September 2016, Google Translate was switched to an artificial neural network, which should enable a more precise and human-like translation and save the manual entry of the required word data.

In February 2018, researchers from Google Brain presented an algorithm that is supposed to be able to collect and extract information from several texts in order to be able to write lexicon articles in natural language and encyclopedic form, e.g. for Wikipedia . However, the developers of the algorithm limit the fact that the quality still deviates significantly from human authors, the algorithm has problems with many sources and is not able to differentiate between trustworthy and untrustworthy information.

See also

  • DeepMind (another company of Google LLC for research into artificial intelligence)

Web links

Individual evidence

  1. ^ Research at Google. Retrieved February 19, 2018 .
  2. ↑ a b Research at Google. Retrieved February 19, 2018 .
  3. ^ Greg Corrado, Senior Research Scientist, Google. RE.WORK Deep Learning Summit 2015, February 20, 2015, accessed February 19, 2018 .
  4. ↑ Using large-scale brain simulations for machine learning and AI In: Official Google Blog. Retrieved February 19, 2018 .
  5. ^ A b John Markoff: In a Big Network of Computers, Evidence of Machine Learning . In: The New York Times . June 25, 2012 ( nytimes.com ).
  6. ↑ A Massive Google Network Learns To Identify - Cats. NPR.org, accessed February 19, 2018 .
  7. ↑ Laura Shin: Google brain simulator teaches itself to recognize cats. ZDNet, accessed on February 19, 2018 .
  8. ↑ Alexander Armbruster: New software in the AI ​​field: How Google wants to predict death . In: Frankfurter Allgemeine Zeitung . January 29, 2018 ( faz.net [accessed February 19, 2018]).
  9. ↑ René Resch: Google's new AI can predict death . In: PC WORLD . ( pcwelt.de [accessed on February 19, 2018]).
  10. ↑ Dana Neumann: Google's artificial intelligence predicts death . ( futurezone.de [accessed on February 19, 2018]).
  11. ↑ Google Claims: We Know Exactly Who Will Not Leave Hospital Alive - Video . In: Focus Online . ( focus.de [accessed on February 19, 2018]).
  12. ↑ David Burger: Scary but innovative: Google's new AI can predict death . In: Chip Online . ( chip.de [accessed on February 19, 2018]).
  13. ↑ Google: AI can now predict whether people find a picture beautiful or not . In: t3n News . ( t3n.de [accessed on February 19, 2018]).
  14. ↑ Björn Bohn: Machine Learning: Google's NIMA can evaluate the quality of images. Retrieved February 19, 2018 .
  15. ^ The Google Brain Team - Looking Back on 2016. Research Blog, accessed February 19, 2018 .
  16. ^ Speech Recognition and Deep Learning. In: Research Blog. Retrieved February 19, 2018 .
  17. ↑ Improving Photo Search: A Step Across the Semantic Gap. In: Research Blog. Retrieved February 19, 2018 .
  18. ↑ This Is Google's Plan to Save YouTube. Retrieved February 19, 2018 .
  19. ↑ Zero-Shot Translation with Google's Multilingual Neural Machine Translation System. Research blog, accessed February 19, 2018 .
  20. ↑ Google uses neural networks to translate without transcribing . In: New Scientist . ( newscientist.com [accessed February 19, 2018]).
  21. ↑ Artificial intelligence: Google Brain writes Wikipedia articles independently. heise online, accessed on February 19, 2018 .
  22. ↑ Peter J. Liu, Mohammad Saleh, Etienne Pot, Ben Goodrich, Ryan Sepassi: Generating Wikipedia by Summarizing Long Sequences . January 30, 2018, arxiv : 1801.10198 [abs] .