Google’s vision for AI: An Introduction
Overview
Will Google I/O 2017 be remembered as a pivotal moment in tech history?
Maybe, maybe not but it did deliver a dramatic shift in Google’s vision of it’s future. In part 1 of this mini-series of articles on Google I/O 2017, we will be focusing on Google’s major shift towards an AI-centered future.
Sundar Pichai, Google CEO declared “An important shift from a mobile first world to an AI first world”, this statement had a much wider implication than initially suspected. It’s manifestation into almost all facets of the Google ecosystem presented at the I/O, from hardware to software certainly raised interest and speculation on this substantial pivot.
AI and Machine Learning
The TPU 2.0 aka “The Cloud TPU” was a major announcement at I/O 2017. The Cloud TPU provides support for training neural networks. The rate in which the Cloud TPU can train neural networks is claimed by Google to be 15x — 30x faster than today’s modern GPUs (Graphics Processing Unit) and CPUs (Central Processing Unit).
What does this mean?
For Machine Learning, Google are able to harness the power of the Cloud TPU to process close to 100 million photos a day, dramatically increasing the rate and accuracy of Machine Learning.
What is a Neural Network?
A neural network is a method of Machine Learning, where machines learn to analyze, identify or perform certain tasks by referencing training examples.
A very simple example would be an image identifying system where a machine would be fed data to determine whether an image was an apple or an orange. The data might consist of colour, texture, weight all represented in numerical form and the machine would make its predictions according to the data set. When presented with an image the machine can check whether it fits the criteria in the data set provided. The predictions would be inaccurate at first, but as more data is fed into the machine and the machine processes more data, the predictions become more accurate.
With the unveiling of the Cloud TPU and the claim to dramatically increase the speed of training neural networks, we can see that the benefits of increasing the accuracy and speed in which a machine would learn. By increasing the processing power and the size of data sets, Google have potentially made a big leap forward for Machine Learning and “early AI”.
It’s accessibility to developers via Cloud gives an incredibly powerful opportunity to developers and entrepreneurs globally.
