Google’s TensorFlow – which is now free – helps train computers in artificial intelligence so that one day, computer systems may solve problems through planning, reasoning, and abstract thinking much like humans.
Some business leaders and computer scientists believe that a breakthrough in the field of artificial intelligence may soon take place, but others are more sceptical and say that although computers will continue their evolution in certain area, they may never be as intelligent as humans.
On Monday, Google open-sourced part of TensorFlow – a deep learning engine – in hopes of taking a step toward building artificial intelligence.
TensorFlow powers Google’s Photos app, speech recognition, translation services, etc. In can be defined as a software library of algorithms that is able to train computer systems to learn and think the way humans do.
Neutral networks carry out intricate mathematical calculation on tensors, or arrays of data, which allows them to find various relationships and patterns as they go through all the information that is available to them.
For instance, with the help of the neutral network, the Google Translate app can provide better translations based on its knowledge of word usage in regular conversations. In a similar way, the Google Photo app can identify a cat within a new picture, based on similar images that it has previously seen.
Sundar Pichai, Chief Executive Officer (CEO) of Google Inc, wrote in a blog post that by open-sourcing TensorFlow they hope that people in the machine learning community – engineers, academic researchers, hobbyists – will take interest in this new project and that they will help accelerate the research.
This particular machine learning engine can work on people’s computers because TensorFlow is not that tied to the company’s hardware, according to Google.
Rajat Monga, TensorFlow technical lead and Jeff Dean, a computer scientists and software engineer who is currently a Google Senior Fellow in the Systems Infrastructure Group, said that TensorFlow is more configurable, more flexible, and about two times faster than DistBelief – the predecessor of the new deep learning system.
Eric Emerson Schmidt, the executive chairman of Alphabet Inc., said in September that machine learning may revolutionise the way in which people will approach problems regarding genomics, climate science, and energy in the future.
Image Source: recodetech