Google’s late entry into the machine learning field will compete with Amazon and Microsoft.
Google Cloud recently released Vertex AI, a machine learning (ML) platform that enables companies to turbocharge the deployment and upkeep of their artificial intelligence (AI) technologies.
Vertex AI incorporates Google Cloud services under one unified user interface (UI), application programming interface (API) and library. In fact, with Vertex, users get access to the very same MLOps toolkit that Google uses to power its own products, including computer vision, language and conversation, and structured data.
MLOps, a relatively new discipline that marries machine learning and information technology operations, features increased collaboration between data scientists and IT professionals working to develop machine learning algorithm products. The Vertex toolkit promises to remain on the cutting edge of this field via continual updating and enhancement by Google Research.
The Vertex platform streamlines the process of building, training and implementing ML models at scale. Within its environment, customers can evolve models from experiments to production-ready products faster, detect and troubleshoot patterns and anomalies, and make more precise predictions and better decisions. This would in turn allow companies to respond faster and with better agility to rapidly changing market pressures.
Vertex aims to provide solutions to the ML development headaches that prevent companies from deploying AI. Currently, data scientists often must stitch together ML point solutions manually—which tends to result in lags and delays in model development, experimentation and testing. As a result, many of these models never make it into production. In fact, a recent report found that data quality issues prevent a whopping 87 percent of companies from implementing any AI technologies.
In addition, most current AI models are developed by data scientists working in isolation from other IT services. And when it comes time to deploy the model, the developers may run into more snags when they struggle to align their products with existing IT processes.
The inefficiencies continue after implementation: models created by data scientists tend to remain static once deployed, so they can fall behind changes in human behavior and channel content.
“Vertex was designed to help customers with four things,” said Craig Wiley, Google Cloud AI product management director. “The first is, we want to help them increase the velocity of the machine learning models that they’re building and deploying. Number two is, we want to make sure that they have Google’s best-in-class capabilities available to them. Number three is, we want these workflows to be highly scalable.… And then number four is, we want to make sure they have everything they need for appropriate model management and governance. Ultimately, the goal here is to figure out how we can accelerate companies finding ROI with their machine learning.”
Google claims that ML products developed with Vertex AI will require almost 80 percent fewer lines of code to train a model compared to its competitors. And the system is designed to allow nonexpert users to create and train custom ML models without having to invest in extensive ML training.
The platform also has MLOps features, including:
- Vertex Vizier, an AI optimizer that automatically tunes ML model parameters—significantly reducing the time needed to tune a model and enabling operators to run more experiments faster.
- A Feature Store—a shared platform where users can serve, reuse and share features and experiments. This feature also provides more effective storage and version control of AI models.
- Vertex Experiments, which enables faster model selection to speed up the deployment of models into production.
- Vertex Continuous Monitoring and Vertex Pipelines, which allow for enhanced self-service model management and repeatability, streamlining ML workflow from end to end.
Vertex aims to make ML development and deployment easier and bring AI and IT functions together—a trend that is gaining in importance. Applications using AI models are becoming more common, which means that the conventional approach of creating AI models separate from the rest of the IT shop is becoming obsolete. It’s becoming imperative to align model development with the DevOps process. In addition, since AI models can be subject to drift, DevOps operators will need to replace those models as new data sources become available or as the assumptions that the AI model was built on become outdated. We may soon reach a place where AI models are treated and managed the same as other software—even if they have been developed differently.
“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,” said Andrew Moore, Google Cloud’s vice president and general manager of Cloud AI and Industry Solutions.
Vertex AI has been road tested at companies such as personal care giant L’Oréal and Essence, a data-driven global marketing firm. L’Oréal used Vertex to train the AI models of its ModiFace augmented reality AI technology for the beauty industry. Essence deploys Vertex to integrate the workflows of data scientists and developers.
“Vertex AI gives our data scientists the ability to quickly create new models based on the change in environment while also letting our developers and data analysts maintain models in order to scale and innovate,” said Mark Bulling, SVP of Product Innovation at Essence. “The MLOps capabilities in Vertex AI mean we can stay ahead of our clients’ expectations.”
Google Cloud introduces Vertex AI.
Google seems to be betting that machine learning management systems and services, such as Vertex, will see sustained demand in an increasingly complex world. Companies are putting more resources into remote working technologies and infrastructure—another indication that the pandemic is bound to have a long-lasting impact on the way collaborative work is done.
This is particularly interesting because the company lags behind its competitors—Amazon and Microsoft—in the ML field. But Google has taken definite steps to close the gap. Vertex AI isn’t Google’s first foray into ML (that would be the Cloud ML Engine, which it released in 2016), but Vertex AI technology seems ready to compete with other more established products such as Amazon SageMaker and Microsoft Azure Machine Learning.
Data scientists and DevOps engineers alike should feel encouraged by Google’s renewed efforts in the ML market, and Vertex looks to be a worthy addition to this growing field. The worldwide machine learning market is predicted to be worth $96.7 billion over the next few years. There has been an exponential growth in data generation—and the need to harness that data to forecast and predict outcomes is growing along with it.
“We are very proud of what we came up with in this platform, as it enables serious deployments for a new generation of AI that will empower data scientists and engineers to do fulfilling and creative work,” said Moore.
Vertex is available now.
For more about recent developments in artificial intelligence, read: Siemens set to incorporate Tangent Works’ AI tool InstantML into MindSphere IoT tech.