Google Cloud on Air

So, I have spent my day today listening to this global online live conference by Google. I know quite a bit about Amazon AWS platform, but didn’t really had much idea about Google cloud platform before joining this conference. And needless to say, google cloud on air, was awesome. Well, some sessions were not as good, but most of them were very informative and they encourage you to explore the google cloud platform even more. The sessions were not only talks and presentations (which were extremely informative anyways), but they were filled with demos and hands on exercizes to get us started on the platform. 

‘Cloud on air’ had 3 parallel tracks. It was a tough choice to choose between them. But then the cloud on air was being broadcasted in 3 timezones, so you can easily watch the session you have missed in the subsequent broadcast. The three tracks were Machine Learning/ AI, Big Data and Industry Solutions. And all of them were amazing. 

Right from developing the services hands on in the google cloud platform to finding the right use cases for the actual industry applications ..it was pretty much all in one. It gave the picture of the transformation cloud platforms have brought in the industry today to the the way technologies are used, designed using google cloud services.

The machine learning/ AI track pretty much explained many concepts and business use cases of AI implementation. Google invented TPUs (Tensor Processing Units) to make the deep learning models compute faster. The sessions explained how the model can easily be overfitted, if it only focuses on memorising the stuff like it does with a linear classifier. To make the models generalised and be able to learn, deep learning models are gaining more popularity. Deep learning model learns the way a human learns. The nodes in the deep learning model are called neurons for the same reason! Just as a child would learn to identify cats and dogs with experience, how the parents correct the child when he is wrong, same way the deep learning networks work. They learn from experience. But they also require lots of computation power, and hence the need of multiple CPUs, GPUs or TPUs.

The big data track covered the need of moving to the cloud. How even the most regularised industries like healthcare and banking also are moving to the cloud for greater security! Yes, cloud and security do go hand in hand. Google services are compliant with many regulation standards including PCI DSS. New regions are available in almost all parts of the world, which is good for countries like Australia, where the regulations do not allow the data to be stored outside the country. Cloud data warehouse using google big query can make a data warehouse implementation much easier and faster. Real time streaming is the way to go in this age. Now you can get insights from your data in minutes than in days.

In today’s world 90% of the data in unstructured. It could be from social media text and pictures, or from IoT devices. You simply can not store it in rows and columns in a relational database. To be able to handle this vast amount of data and be able to process this data, gain insights from it, to apply AI on this data, going cloud is mandatory. And google cloud platform provides easy to use services for each of these steps the data needs to go through to make it usable for the businesses. 

I was able to correlate many of the services from google with the amazon aws services. And it’s really great to see these platforms providing such excellent services. Currently I don’t really have sufficient knowledge to be able to compare them though!.

TensorFlow is a Google machine learning library, which is open source and is at number one on Gihub for machine learning. And you don’t need to be a data scientist expert to be able to use this service. If you are, great, but even if you aren’t an expert in building and tuning the models, the self learning deep networks will auto improve the models through computations and can even give results better than a human can. That is AI doing AI!

The cloud platform makes it cost effective, as you only pay for what you use. You can create a cluster in minutes use it for computations and then delete it , so you won’t be charged for it anymore. Increasing cpu power, memory, all on the go. Your cluster, current running job won’t be affected by these changes. No downtime at all! 

The conference educated people that machine learning and AI is not only a hype! It can actually provide tremendous business opportunities, will improve customer experience, enable faster data driven decision making and develop unseen revenue generation opportunities.