Conversation Technology, Chip Performance, and the Vertex Platform - This Week in AI

Welcome to our This Week in AI roundup. This week we have several announcements from Google's annual developer conference, Google I/O.

4 months ago   •   2 min read

By MLQ

Welcome to our This Week in AI roundup. Our goal with this roundup is provide an overview of the week's most important news, papers, and industry developments.

This week we have several announcements from Google's annual developer conference, Google I/O.

Google's Breakthrough Conversation Technology

The AI team at Google created LaMda, which stands for "Language Model for Dialogue Applications." It's the latest development in the company's mission to better people's interactions with technology.

With a series of open-ended questions, Google's LaMDA can start conversations and can talk about an almost limitless number of topics in a free-flowing manner. The technology has the potential to open up new areas of useful applications and more natural ways of interacting with AI.

The technology was created with the aim of improving the speed and accuracy of chatbots that respond to voice commands. LaMDA is a modern language-trained conversational model that can predict what words will appear next in a conversation.

Read the full story.

Google Doubles Chip Performing with TPU Vx4

Google's CEO Sundar Pichai also discussed the company's new TPU v4 Tensor Processing Units, which have improved the performance of its TPU hardware by more than twofold.

More than one exaflop of floating point output can be delivered by a single TPU pod of v4 chips. Analysts are praising Google's TPU v4 announcement and what it means for businesses dealing with ever-increasing ML training demands.

Read the full story.

Google launches AI for skin conditions

Google has unveiled a platform that uses artificial intelligence to help patients identify skin, hair, and nail condition based on images uploaded by user. In Europe, the software has earned a CE mark for use as a medical tool. It took three years to create and was trained on a dataset of 65,000 photographs of diagnosed diseases.

Medical devices, including those with AI at their core, must strike a balance. The question that the author highlights is:

Do you focus on catching everyone who has a disease or on ruling out those who are healthy to avoid unnecessary worry or treatments?

One is often at the expense of the other.

Read the full story.

Google's New Vertex Platform

According to a survey, the global AI market is projected to rise from $27.23 billion in 2019 to nearly $267 billion by 2027. Artificial intelligence, data processing, and other advanced workloads have been made more accessible and easier to use for businesses by tech giants.

Vertex AI is a framework that provides a single user interface and API for a variety of existing machine learning services. Vertex AI allows developers to train an AI model with nearly 80% fewer lines of code than platforms from other cloud providers.

As Craig Wiley, the director of product for Vertex AI wrote in a blog post:

Today, data scientists grapple with the challenge of manually piecing together ML point solutions, creating a lag time in model development and experimentation, resulting in very few models making it into production.

Vertex would "build an industry-wide change" that will "force everyone to get serious about putting AI into development," according to Google.

Read the full story.

That's it for this edition of This Week in AI, if you were forwarded this newsletter and would like to receive it you can sign up here.

Spread the word

Keep reading