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?EU?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.
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…