‪‪Google‬, ‪Apprentissage automatique‬‬

Google‬, ‪Apprentissage automatique‬‬

Deep Learning in Action

At places like Google, Facebook, Microsoft, and Baidu, GPUs have proved remarkably important to so-called "deep learning" because they can process lots of little bits of data in parallel. Deep learning relies on neural networks-systems that approximate the web of neurons in the human brain-and these networks are designed to analyze massive amounts of data at speed. In order to teach these networks how to recognize a cat, for instance, you feed them countless photos of cats. GPUs are good at this kind of thing. Plus, they do not consume as much power as CPUs
"Machine learning is not magic, but is just a tool, a new way of building software," Greg Corrado, senior research scientist at Google, said during a joint press conference here with reporters from Asia-Pacific regions, Tuesday. "This is a new field where computers learn from data and experiences."
Google organized the "Magic in the Machine" event to introduce its software that will be used in the machine learning framework, which the company believes will be a building block to establish infrastructure for the next-generation information technology (IT) industry.
The move comes as the company is seeking to leverage its machine learning platform as the new growth area amid surging demand for efficient data management that can be applied to tasks in real life.
"For real applications of machine learning, it takes billions of repetitions," he said. "The old way of machine learning was to write a computer program with explicit rules to follow, but the new way helps a computer program itself by learning from examples and repetitions."
Machine learning is a type of artificial intelligence which helps computer software make predictions when exposed to new datasets. A growing number of multinational tech giants, including IBM, Apple and Microsoft, are betting big on developing machine learning-based software as a pillar for their new revenue sources.
Asked about Google's partnership with Korean companies, the Google executive anticipates that Google will extend its partnership with Korean tech giants such as LG over the development of Android-based machine learning software.
"I think all of the companies recognize that machine learning is helpful and innovative for products," he said. "All of the companies are already using the technique and I believe that it will continue."
Top challenge with Internet
Eric Schmidt, chairman of Google parent company Alphabet, stressed the importance of getting all the people around the world connected through the Internet.
"I think the greatest challenge is getting every human being on the Internet," he said during a keynote speech at the press conference. "The Internet was designed for everyone, and is a modernizing force that will make the world happier and healthier. We should also have the goal of getting everything connected, not just certain people."
Regarding machine learning, he expressed excitement over the future that machine learning will bring.
"This technology is good at predicting the sequencing of things," he said. "We have more than one hundred teams using the machine learning. Google is the leader in this technology."