At the recent Google I/O 2021 conference, the cloud provider announced the general availability of Vertex AI, a managed machine learning platform designed to accelerate the deployment and maintenance of artificial intelligence models.
Using Vertex AI, engineers can manage image, video, text, and tabular datasets, and build machine learning pipelines to train and evaluate models using Google Cloud algorithms or custom training code. They can then deploy models for online or batch use cases all on scalable managed infrastructure.
The new service provides Docker images that developers run for serving predictions from trained model artifacts, with prebuilt containers for TensorFlow, XGBoost and Scikit-learn prediction. If data needs to stay on-site or on a device, Vertex ML Edge Manager, currently experimental, can deploy and monitor models on the edge.
Vertex AI replaces legacy services such as AI Platform Data Labeling, AI Platform Training and Prediction, AutoML Natural Language, AutoML Video, AutoML Vision, AutoML Tables, and AI Platform Deep Learning Containers.
Andrew Moore, vice president and general manager of Cloud AI at Google Cloud, explains why the cloud provider decided to introduce a new platform:
We had two guiding lights while building Vertex AI: get data scientists and engineers out of the orchestration weeds, and create an industry-wide shift that would make everyone get serious about moving AI out of pilot purgatory and into full-scale production.
Cassie Kozyrkov, chief decision scientist at Google, highlights the main benefit of the new product, managing the entire lifecycle of AI and machine learning development:
If only machine learning had the equivalent of a Swiss Army knife that was 80% faster to use than the regular toolbox. Good news, as of today it does!
In one of the comments, Ornela Bardhi, Marie Curie PhD fellow in AI and health at the University of Deusto, praises the new service but raises a question about accountability of managed services in machine learning:
It was about time some company was going to create such a platform (...) If the model performs not as intended, who would be accountable in this case? Considering that one of the benefits is "train models without code, minimal expertise required".
Some users on Reddit question instead if the announced platform is simply a rebranding, as user 0xnld suggests:
Not obvious from the article, but it appears to be a rebranding of AI Platform (Unified) which was in beta for the last year or so.
In a separate article, Google explains how to streamline ML training workflows with Vertex AI, avoiding running model training on local environments like notebook computers or desktops and working instead with Vertex AI custom training service. Using a pre-built TensorFlow 2 image as example, the authors cover how to package the code for a training job, submit a training job, configure which machines to use and access the trained model.
The pricing model of Vertex AI matches the existing ML products that it will supersede. For AutoML models, users pay for training the model, deploying it to an endpoint and using it to make predictions.