The keynote for the Google I/O developer conference just wrapped up, but the company has already made it clear that artificial intelligence and machine learning are two of the main themes of this year’s event. CEO Sundar Pichai kicked off the event by saying the company is focused on leading the drive to AI within the industry, comparing the shift to AI as similar to the earlier shift from desktop to mobile devices.
Pichai said the company is rethinking all of its services in order to apply AI and machine learning technology across as many aspects of Google's business as possible. The company also introduced a new initiative it is calling Google.AI, which will work to centralize all of the company's AI efforts under one roof.
Cloud Tensor Processing Unit
In addition, Pichai announced that Google will be working on the hardware side of AI technology equation with the launch of its new Cloud Tensor Processing Units, or Cloud TPUs. The chips are similar to the company's existing TPUs, but are optimized both for training neural networks and allowing neural nets to make inferences about the data they are interpreting. Pichai said the Cloud TPUs are designed for deployment in company data centers, and will be available on Google Compute Engine.
On the software side, Pichai said Google is working on developing neural nets capable of developing their own neural nets in an automated process he described as machines "learning to learn." The new iterative process will be powered by the Cloud TPUs.
Pichai identified several use cases on the organizational and enterprise level for the company's AI approaches. For example, Google's platform could be used to develop neural nets that are able to make more accurate diagnoses in diseases such as breast cancer, which is notoriously difficult for humans to diagnose.
In chemistry, Pichai said the company's tools could be used to develop synthetic molecules for the drug discovery and material science industry that can be tailor-made to behave in certain pre-defined ways.
Machine Vision and Natural Language Processing
Google highlighted two specific areas today in which it has applied AI technology, machine vision and natural language processing. In the case of machine vision, the company announced its new Google Lens initiative, which aims to help machines understand the world in a way similar to how humans do. In once example, Pichai demonstrated how Google Lens could be used on a smartphone to snap a picture of a WiFi password and then connect itself to the network automatically.
Most of the natural language processing advances, meanwhile, are being put to use on Google Assistant, an AI assistant similar to Apple's Siri and Amazon's Alexa. The company announced that Assistant is now available on the iPhone, and that it is working with consumer appliance manufacturers to integrate the platform into their devices with the newly announced Assistant SDK. The SDK will enable third parties to develop appliances that can be controlled verbally using Google Assistant.
The company also announced Actions on Google today, a new platform that will support third party integrations that will allow users to make transactions through Assistant. For example, a user could call, order, and pay for a takeout meal verbally without having to fill in their account information or street address.
The 2017 Google I/O developer conference runs May 17 - 19, at the Shoreline Amphitheatre in Mountain View, California.