How is Google Enabling AI?


hen Google started as a research project in a Stanford Lab in 1998 its objective was to index all the world’s information, from that time till now we have seen multiple transformation of Google and IoT now has offering ranging from GSuite to Google Compute. Having dominated search on the web, Google is set to take the Artificial Intelligence (AI) experience to the next level. While AI has been around for some time, a backing by Google is likely to bring about some massive developments in this domain.

Google has an old relationship with AI. Image search, translation, maps, and Google Assistant, all involve the use of AI. Recent years have seen a rise in the extent to which Google is incorporating AI in its system. A large number of developments in Google’s AI sphere have taken place over the last couple of years. At this year’s I/O Developer Conference, Google’s Sundar Pichai made a number of interesting announcements, signalling the importance of AI. Leveraging AI to solve problems for its customers, Google is taking upon itself to bring the AI benefits to the world.


TensorFlow was released in 2015 by Google Brain team and is a Python, C++ and CUDA library that was created to support Deep Learning Neural Networks. It leans heavily on the Python based Theano library. TensorFlow supports the definition and execution of a Computational Graph along with supporting Symbolic Differentiation for calculating gradients in the back which propagates learning phase of a Neural Network.

On another note TensorFlow cannot claim to be the first Distributed Deep Learning framework, other platforms like H2O, DL4J, Spark MLLib, Petuum, Singa, etc provide similar functionalities. In fact, prior to Tensorflow Google had a distributed framework called DistBelief. The benefit that TensorFlow brings to the table is that it reduces the complexity of deploying Deep Learning solutions as compared to previous system.

Cloud AutoML Vision

Even before Google brought the Cloud AutoML capabilities into its fold. AutoML has been around for many years (which includes multiple open-source AutoML libraries), in May 2017 Google adopted the term AutoML for its neural architecture based search.

That’s why we’ve created an approach called AutoML, showing that it’s possible for neural nets to design neural nets

– Sundar Pichai at Google I/O conference 2018

Cloud AutoML, provides organizations, researchers, and businesses who require custom machine learning models a simple, no-frills way to train self-learning object recognition. Cloud AutoML has expanded into natural language processing (with AutoML Natural Language) and translation (with AutoML Translate)


DeepMind, was founded in September 2010, by Demis Hassabis, Shane Legg and Mustafa Suleyman and was acquired by Google in 2014 and is now at the forefront of development happening in the AI domain within Google.

ML and AI now make tasks like image and pattern recognition possible within vast amounts of data, by leveraging these development DeepMind is looking at how machine learning and other AI-related technologies can be applied to areas such as protein folding and quantum chemistry, helping government as well as agencies address aspects like patient deterioration, cancer diagnosis, battery management, energy supply and Generating Google Assistant voices.

Google Duplex

Google Duplex is one of the most talked about applications by Google, still under development. The key feature of this application is that it has the ability to make phone calls on your behalf. It is aimed at tasks like fixing appointments, scheduling travel etc.

Machine Learning Kit

Google is also launching a machine learning kit for app developers who aren’t skilled at it. The software development kit (SDK) supports text, images, barcode scanning, landmark recognition etc. and is available on both iOS and Android.

Google’s Increasing Foray Into AI

With sophisticated machine learning technology, Google is also looking at solving environmental problems. By combining cloud computing, geo-mapping, and machine learning, the company is giving rise to some interesting use cases.

For instance, Project Sunroof, launched in 2015, trains Google’s systems to examine satellite data and identify homes with solar panels mounted on their roofs and those that don’t. The technology combines Google Earth satellite images with meteorological data to assess if a particular location is ideal for solar panel installation along with the energy savings that result. The technology has the potential to be useful for not just home-owners but also city governments.

The Verdict

With so many AI projects under its ambit, Google seems to have taken the pole position in furthering AI’s potential. Considering that most of its previous experiments with AI have been successful, Google is likely to continue leading other successful developments.