Google announces launch of open AI division

Google announces launch of open AI division

CEO of Google, Sundar Pichai, launched a giant open AI division during Google’s I/O developer conference. Pichai announced that Google is making available a new machine-learning system called TensorFlow through an online open-source library.

The new Google.ai site and division focuses on artificial intelligence, deep learning and building the tools to make their new focus on AI for both its services and research, possible.

Pichai announced that that the Tensor Processing Units (TPUs) it uses to train machine learning models are available in the Google Cloud Platform for anyone to use via the Google Compute Engine today. “We want it to be possible for hundreds of thousands of developers to use machine learning,” Pichai said.

Pichai explained that TensorFlow allows Google to build and train neural nets up to five times faster than their first-generation system, known as Distbelief. Used internally at Google to do things like identify and automatically label items in YouTube videos and photos, and improve speech recognition in Google apps, Distbelief is narrowly targeted to neural networks and is tied tightly to Google’s internal infrastructure.

TensorFlow is the second-generation machine-learning system, which Google claims is twice as fast and more flexible, and can be run on an individual smartphone or across data centers. Pichai said that Google will be using the neural nets it creates to build other neural nets with AutoML. A candidate set of neural nets is reiterated using a reinforcement-training approach until the best one is found.

This will not only be used in Google’s own products but also to assist health care providers in their work to improve care and prevent medical incidents. Pichai gave examples of the possibility of AI helping to sequence DNA and helping pathologists locate things like the spread of cancer.

Source: Engadget