Google Cloud has launched a new managed Artificial Intelligence and Machine Learning platform called Vertex AI, bringing together all of its AI tools, including AutoML and AI Platform, in one unified API.
Vertex AI has been set up to make model development easier and comes with pre-trained APIs for computer vision, language, structured data, and conversation, Google Cloud said, as hyperscaler segment heats up.
The company claims Vertex AI requires nearly 80% fewer lines of code to train a model, and Google Cloud says it will give developers access to the AI toolkit used internally to power Google for the first time.
The aim: "Get data scientists and engineers out of the orchestration weeds, and create a industry-wide shift that would make everyone get serious about moving AI out of pilot purgatory and into full-scale production,” as Andrew Moore, VP and general manager of Cloud AI and Industry Solutions at Google Cloud put it.
Vertex AI: Google Cloud aims to edge ahead in a competitive space.
Google Cloud's AI chops are undisputed: the company is heavily engaged in open-source community projects that originated in Google -- think TensorFlow, BERT and Kubeflow.
Gartner's 2021 Cloud AI Developer Services Magic Quadrant notes approvingly of Google Cloud's services that "developers without extensive ML expertise can quickly infuse AI into their applications, rapidly prototype and increase the speed of deployments", while AWS's SageMaker Autopilot is dubbed potentially confusing, with the company deemed by Gartner more focussed on developers than data scientists.
See also: Netflix open sources its “battle-hardened” Domain Graph Service framework
(Microsoft is among the "more flexible providers of CAIDS in terms of deployment options", Gartner notes approvingly. "Its services are deployable in the Azure cloud, a virtual private cloud or on-premises via containers, based on enterprise customers’ needs", but the research firm warns that "Microsoft focuses its citizen developer AI offerings on its Power Platform; for instance, PowerBI and Power Virtual Agents. While Power Platform has prebuilt connectors to industry systems, the separate positioning of the citizen platform may force upon companies an enterprise roadmap for which they were not planning.")
Vertex AI: What can you do?
Vertex AI, in brief, lets users train and compare models using AutoML or custom code training, storing all models in one central model repository. Customers can use GCP's BigQuery ML to create and execute machine learning models in BigQuery using standard SQL queries on existing BI tools and spreadsheets, or can export datasets from BigQuery directly into Vertex AI.
The platform integrates with TensorFlow, PyTorch, and scikit-learn, among other ML frameworks via custom containers for training and prediction; you can also connect to VMs using JupyterLab, the standard workbench for data scientists. (VMs come pre-installed with deep learning frameworks and libraries.) The company put forward L’Oréal subsidiary ModiFace, as a customer example. ModiFace cleverly creates new services for consumers to try beauty products such as hair color, makeup and nail color, virtually and in real-time.
ModiFace uses Vertex AI for its skin diagnostic, which is trained on thousands of images from L’Oréal’s Research & Innovation, the company’s dedicated R&D arm to deliver deliver tailor-made skincare routines, in a compelling example of the use of AI to deliver targeted customer offerings. As ModiFace COO Jeff Houghton put it, the technology "allowed us to create technology that is incredibly close to actually trying the product in real life.”
The full list of Vertex AI features is here. Pricing is here.
