AI novelties at Google Next 2019
“AI is a tool. The choice about how it gets deployed is ours.” Oren Etzioni
In a similar article, I talked about new analytic releases at Google Next 2019 conference. In this article, I will describe a few of the new AI beasts that have been announced. The focus this year has been on ML democratization with the enhancements of BigQuery ML and AutoML and the arrival of a full end-to-end AI platform. It’s definitely a very exciting period for AI and needless to say that Google is pioneer in making it easily accessible to everyone. Far away is the time when you needed advanced statistical and mathematical background to approach it.
Note: this is a cross-post from the Fourcast blog. Find the article there.
- AI platform: Streamline and simplify ML projects from ideation to production and deployment. End goal is to have one platform to make it all. AI platform is backed by Kubeflow which uses kubernetes in the backend. This enables each part of your ML pipeline to be portable and sit on a microservice of its own. This platform also leverages the best of Google technology with TensorFlow, TPUs, and TFX tools as you deploy your AI applications to production.
- BigQuery ML: It was introduced last year at NEXT 2018 and since then it keeps on evolving at a very fast pace. It is now generally available and includes all the following new models: K-means clustering ML, tensorflow DNN classifier and tensorflow DNN regressor. This places BigQuery ML as a serious player to develop ML models straight from your data warehouse and increases the importance of BigQuery as the backbone of the analytical infrastructure. This is an excellent news due to the ever-evolving efficiency of BigQuery.
- AutoML Tables: The AutoML suite also gets bigger with the arrival of AutoML Tables that will enable you to build and deploy state-of-the-art ML models on structured data with only a few clicks. Same concept as for the other products of the AutoML suite, you just need to feed AutoML with the data and it will then do the rest and build you an optimal model by carrying out automatic feature engineering, grid-searching on various different models and a combination of hyper-parameters fine-tuning.
- Recommendation AI: Data scientists know too well that building recommendation systems can be a cumbersome and time-consuming task with results not always guaranteed. GCP is tackling it with this new product that is based on years of experience at Google on recommending content across flagship properties such as Google Ads, Google Search, and YouTube. “Recommendations AI uses Google’s latest machine learning architectures, which dynamically adapt to real-time customer behavior and changes in variables like assortment, pricing, and special offers.”
- Document understanding AI: uses ML to understand and analyze any kind of documents. A lot of companies loose time in non-added value and annoying tasks that consist of taking data from a document and putting it in your IT system. Document Understanding AI comes at the rescue by capturing, classifying, enriching, and visualizing documents in both physical and digital formats. It will save your organization time, money but more importantly will make your process less error-prone. Use cases go from “Document and content management“ to “Robotic Process Automation” and “Procure-to-pay automation”.
It has never been a more thrilling time to work in AI fields as all the technology around it gets easier to handle day by day. Thanks to the enhancements of its core AI building blocks and some few brand-new products, GCP is keeping its position as a leader in doing AI on the cloud. If you think of any AI applications, rest assured that GCP has the solution for you and that you can get something working withing a few days. This is somehow the magic of the cloud; you get things done quickly and in case you fail, you fail quickly and at low cost as you leverage the “pay as you go” pricing policy of every GCP products.
While the entrance barrier is getting lower and lower, we acknowledge that this barrier still exists and for this reason our GCP consultants will be glad to accompany you in your different AI journeys on GCP.
Contact us at email@example.com and feel free to visit our website on www.fourcast.io.