By Jim Lundy
(Aragon Research) – Google held its annual Google Enterprise event called Google Cloud Next 2017 and with it, came a flurry of announcements and partnerships, including a significant one that allows SAP Hana to run on Google Cloud.
That said, there are a number of significant moves to allow more firms to take advantage of machine learning on the Google Cloud. This blog analyzes some of the announcements Google made at Cloud Next.
Cloud Machine Learning Availability
While it may seem like just another service, Google’s Cloud Machine Learning Service means that deep learning (via TensorFlow) is now available to enterprises from Google. Deep learning leverages things such as a neural network to allow for analyzing more complex issues and tasks, such as image recognition.
Google Machine Learning is a full cloud-managed service. Offerings such as HyperTune help enterprises to tune a model quicker and improve the model’s predictive accuracy. This will also allow an enterprise to get a model into production faster. Google Machine Learning also works with Google Cloud Dataflow (data processing), Google Cloud Datalab (data science workflow), and Google BigQuery (SQL analytics).
Why Deep Learning vs. Machine Learning?
Enterprises need to realize that there are different ways to do machine learning. TensorFlow is a library that can be used to write or instantiate a deep learning algorithm (it has other uses, too). What’s interesting is that Google has also produced TensorFlow hardware accelerators (TPUs), which were used in the AlphaGo project and are now finding more widespread use in Google Cloud.
TensorFlow is one of the first major deep learning libraries to be open sourced, and a number of major enterprises are using TensorFlow today. The key point of Google Cloud Machine Learning is that Google is now making it easier to leverage its cloud to leverage existing deep learning models (i.e., see video intelligence below) or use it to develop new ones. Enterprises trying to decide on how to proceed in this area should talk to Aragon, as AI/deep learning is one of our focused topics.
Google Cloud Video Intelligence API
Google is releasing a video intelligence API that allows videos and scenes in videos to be searched and summarized. Cloud Video Intelligence is in beta and firms such as Cantemo are already using it.
Our take is that this is an incentive for more firms that specialize in video to leverage the Google Cloud. As video content becomes the dominant content type, cloud providers like Google are focusing on finding videos and scenes in videos a common capability.
Google Cloud Vision API 1.1
The Google Cloud Vision API, which has been in beta since March of 2016, is an image classification API. It allows for image and facial Recognition.
Our take is that for many applications going forward, image and facial recognition will be vital. One of the things that Google is doing is making it easier for enterprises to try their APIs. We even uploaded a picture (see Figure 1) and the vision API instantly did facial recognition on it (note this was without any tuning).
Cloud Jobs API
Google is helping the HR tech community to offer more advanced predictive recruiting capabilities. The Cloud Jobs API, which runs on the Google Cloud, is attracting more firms, such as CareerBuilder, Dice, and Jibe to leverage the Google Cloud.
Machine Learning and Support
One of the hardest things for any enterprise is getting started with machine learning. Training is vital. Google is now offering a number of choices for support including support from engineers. Options include:
- Development engineering support ($100 per user per month) is for developers or QA engineers and offers a response within 4 to 8 business hours.
- Production engineering support ($250 per user per month) provides a one-hour response time for critical issues.
- On-call engineering support ($1,500 per user per month) pages a Google engineer and delivers a 15-minute response time 24/7 for critical issues.
Key Takeaway
The key takeaway from Google Next 2017 is that Google is much more focused on making it easier to access and use its growing portfolio of machine and deep learning capabilities, powered by Google Cloud.
Google now counts on over 13,000 partners who are offering various aspects of Google Cloud in their offerings. Google is one of the major cloud providers and the depth of its offerings in machine learning means Google is one to evaluate when evaluating machine learning-based cloud platforms.