Google Opens Access to Machine Learning

A saying remind us that artificial 
intelligence is totally different
from Human Mind

Google runs some of the most advanced computing systems in the world, running everything from its search engine to its many other projects. Now Google is opening up its computing power to all by making its machine learning software open access.
Google made the announcement Monday that they will be allowing anyone access to TensorFlow, their newest machine learning system. The software is an artificial intelligence that gradually learns to recognize patterns, helping it to identify things it has not yet seen. For example, after seeing pictures of cats, it may learn to recognize a picture of a cat it has not seen before based on general characteristics.
Google currently uses machine learning to run different applications like Google Translate, Google Photos, and speech recognition for search. Now, individuals, researchers, and other companies will be able to use the power of Google’s software for their own projects. Instead of others trying to create their own machine learning software, they can instead work on applying the software to problems in the real world. Others may also be able to make improvements to the software, making it better for all, including Google. In the past, others have tried to make their own versions of Google’s software, but they may not have gotten everything right. They would have spent time and energy trying to fix problems that Google had already solved in their software. By simply releasing their software to everyone, Google gets past this duplicate work to allow the industry to grow.
The software will be scalable, so that a hobbyist can run it on a smartphone while a researcher harnesses the power of a group of computer servers. The idea is that researchers could use the power of machine learning to advance research into protein folding, astronomy, and anything else that involves huge amounts of data.
Other groups, including researchers at New York University, had already developed and released machine learning software to the world. However, Google’s is considered the most advanced system.
Before those that work in search engine optimization get too excited, Google is not revealing all its secrets. They are still withholding parts of their software as well as the hardware that runs it. This means that people will not be able to use Google’s software to game the system and rank higher in search engines. However, it does open up the ability to crunch data, which could help people like internet marketers better target their site.
Machine learning, or deep learning, involves using hardware, computer processors, that act similarly to neurons in the human brain, linked by the TensorFlow software. The machine can then use these different “neurons” to process huge amounts of information. Once they are fed huge amounts of words, they begin to recognize sentences and understand what is being said, for example. The machine can then learn how to have a conversation with someone. While the machine’s abilities still remain below that of a child, the technology is always improving. With more people working on the basic software and making improvements, the technology may be able to improve even faster.
TensorFlow is the software that runs Google’s applications, such as their newly-released Smart Reply. Announced last week, this software analyzes email in a Gmail inbox and automatically writes three suggested replies. If all looks good, the user can just hit send instead of having to type a reply from scratch. Email can take up a disproportionate amount of time during the day, and Google is hoping their Smart Reply can help users save time replying to mundane emails. As a user continues to use the system, it will gradually learn patterns that it will help it create better email replies.
Source : jilard