News

IBM AutoAI for Watson Studio Aims to Streamline AI Rollouts

• By Michael Desmond
• 06/19/2019

IBM this week extended its effort to make artificial intelligence (AI) and machine learning (ML) more accessible, with the release of AutoAI for IBM Watson Studio.

Available now for Watson Studio on the IBM Cloud, this suite of AI capabilities streamlines and automates many of the thorniest tasks in data prep and preprocessing, including model development and feature engineering.

AutoAI targets what Rob Thomas, General Manager for IBM Data and AI, calls "a foundational step in AI."

"IBM has been working closely with clients as they chart their paths to AI, and one of the first challenges many face is data prep," he says. "We have seen that complexity of data infrastructures can be daunting to the most sophisticated companies, but it can be overwhelming for those with little to no technical resources."

AutoAI is designed to "smooth the process," Thomas says, helping organizations more quickly build ML models and experiments so they can focus on higher value activities around AI, such as designing, testing and deploying ML models.

IBM says that AutoAI includes a suite of powerful model types for enterprise data science, such as gradient boosted trees, and is engineered to let users quickly scale ML experimentation and deployment processes. The tooling also incorporates IBM Neural Networks Synthesis (NeuNetS) for development of deep-learning models. Previewed last fall and currently in open beta within Watson Studio projects, NeuNetS synthesizes custom neural networks that can be optimized for speed or accuracy.

The new offering addresses common pain points for organizations wrestling with data management and quality in their AI projects. A recent Forrester survey found that 60 percent of respondents cited managing data quality as a top challenge with AI deployments, while 44 percent flagged data prep.

These kinds of challenges pose a real threat to AI projects. As the report notes: "Most enterprise AI models don't make it into production, and many stall at the pilot or proof-of-concept phase, even when they show value."