Conference Industry Research

Fei-fei Li in Google Cloud NEXT ’17: Annoucing Google Could Video Intelligence API, and more Cloud Machine Learning Updates

Between March 8-10, Synced was invited as media guest to attend the Google Cloud NEXT '17 conference in San Francisco.

Between March 8-10, Synced was invited as media guest to attend the Google Cloud NEXT ’17 conference in San Francisco.

The first day of the conference was opened by keynote speeches from senior vice president of Google Cloud – Diane Greene, Google CEO – Sundar Pichai, Alphabet executive chairman – Eric Schmidt, and Google Cloud Machine Learning and Artificial Intelligence chief scientist – Fei-fei Li.

Fei-fei Li joined Google last November to much fanfare in the AI industry. Therefore, it was no surprise that her debut in the Google Cloud Conference turned out to be the center of attention. In her Keynote Speech, Fei-fei Li, on behalf of Google, released several Google Could API products and explained Google’s strategy of “Democratizing AI” . She also announced Google Cloud ‘s acquisition of Kaggle, one of data science’s most prominent community.

Fei-fei Li introduced the application and development of AI by providing several case studies:

  • Retail: Machine learning algorithms helped Google AdSense by providing consumers with more appropriate advice. But there is still room for improvement for things such as supply chain optimization, time-of-demand changes, and the use of UAVs or unmanned vehicles for consumers express goods.
  • Media and entertainment: For example, auto tagging of Google Photos and YouTube’s suggested playlists. Also, today’s VR and AR rely on computer vision for motion tracking, environmental monitoring and gaming. Even news reports will be generated automatically. Artificial intelligence will help us create more personalized content, such as music, video and artwork.
  • Finance: Machine learning is playing an increasingly important role in credit card risk detection, anti-fraud and money laundering.
  • Health care: Artificial intelligence is improving people’s lives in more ways than one, such as intelligent hospitals filled with sensors, and enhanced diagnosis. A few months ago, researchers at Google Brain showed that they could use deep learning to help diagnose diabetic retinitis.

Artificial intelligence is becoming ubiquitous, but not all developers have the ability to utilize AI technologies. Therefore, Fei-fei Li argued that the next step in artificial intelligence must be the democratization of this technology. We must lower the barriers of entry into AI and make it available to the largest community of developers, users, and enterprises so they can apply it based on their own needs. Cloud computing is the perfect way to accomplish the democratization of artificial intelligence. That is why Google is fully invested in the cloud of artificial intelligence and machine learning.

The democratization of artificial intelligence contains four aspects: Democratizing compute, Democratizing algorithms, Democratizing data and Democratizing talent.

1.Democratizing computation.
Deep learning requires a significant amount of computing resource to enable its millions of parameters and billions of connections, this is where the Cloud comes in. Fei-fei Li said “ Cloud Video Intelligence API (now in Private Beta) uses powerful deep-learning models, built using frameworks like TensorFlow and applied on large-scale media platforms . Cloud Machine Learning Engine, now in GA, is an attractive option for organizations that want to train and deploy their own models into production in the cloud. It has the advantages of a managed service for building custom TensorFlow-based machine-learning models that interact with any type of data, at any scale.”

With ML Engine, users can use the TensorFlow libraries they are familiar with, and focus their efforts on their ideas and solutions. Google Cloud will handle the infrastructure and model creation for users: users upload their needs to Google Cloud, and its ML Engine can run large-scale processing, then deploy them on mobile devices. However, for many people. Machine learning is still very complex. To remedy this, Google released a trained API. it is like a switch, you can apply its intelligence to any application to allow it to understand voice, image or natural language.

2.Democratizing algorithm.
Vision API was a new product announced by Fei-fei Li. The improved Vision API contains several new critical capabilities. Number one being the API’s metadata being extended to millions of entities from Google Knowledge Graph. Today, they use the same metadata to support Google’s image search. Another improvement being the enhanced optical character recognition (OCR) function, which is capable to extracted text from text-rich images, such as legal documents.

As to the information in videos, Google Cloud also released another new API – Video Intelligence API, which can identify and help users index objects in videos.

3.Democratizing data.
Like humans, artificial intelligence requires a significant amount of data to provide insight on self-development. Thus, the data set is one of the biggest obstacles in the development of artificial intelligence. Although ImageNet achieved great success, it had to overcome many difficulties, and there are still issues remaining to be solved. Fei-fei Li argued that we need to have more scalable and more effective methods to democratize data. To provide data for more data scientists, machine learning developers, experts in various fields, and even business users.

One of Google Cloud’s main accomplishment on the journy to democratize data is the acquisition of the data science community Kaggle. Kaggle is home to more than 850,000 data scientists with numerous open source data sets. Fei-fei Li stated that after the acquisition of Kaggle, Google Cloud will provide the most advanced machine learning environment for this community, and provide opportunities for marketization.

In fact, Kaggle have already been working with Google before the acquisition. In the past, they have co-hosted the Google Cloud & YouTube-8M Video Understanding Challenge

4.Democratizing talent.
Google Cloud released the Advanced Solution Lab to allow other companies to tap into Google’s wealth of experts to help them solve complex machine learning problems. USAA is a good example, where they were able to solve their technical issues with the help of Google experts.

A fun fact: this year’s conference coincided with the International Women’s Day (March 8th), In her speech, Fei-fei Li also mentioned one of her doctoral students, head of Google Cloud’s artificial intelligence and machine learning R&D division – Jia Li. She said “ She is another badass woman in Stan, CS and AI.” We hope that more and more women will join science and technology to show the world their value, and earn the respect they deserve.


Original Article from Synced China| Author: ChainnZ| Localized by Synced Global Team: Meghan Han, Rita Chen

1 comment on “Fei-fei Li in Google Cloud NEXT ’17: Annoucing Google Could Video Intelligence API, and more Cloud Machine Learning Updates


Leave a Reply

Your email address will not be published. Required fields are marked *

%d bloggers like this: