Artificial intelligence has become quite accessible for enterprises thanks to cloud services such as Google LLC’s Vision API that offer pretrained machine learning models on-demand. But building a custom AI optimized for a company’s specific use case is still a costly endeavor.

Google has set out to change that by introducing today a new cloud toolkit called AutoML that provides a drag-and-drop interface for training AI models. According to the technology giant, the service aims to address the fact that most enterprises can’t afford the talent necessary to build a machine learning model from scratch. Plus, AI development is often taxing even for the organizations that can.

“There’s a very limited number of people that can create advanced machine learning models,” Fei-Fei Li, Google’s chief scientist for Cloud AI, and Jia Li, head of research and development for Cloud AI, wrote in a blog post. “And if you’re one of the companies that has access to ML/AI engineers, you still have to manage the time-intensive and complicated process of building your own custom ML model. We believe Cloud AutoML will make AI experts even more productive, advance new fields in AI and help less-skilled engineers build powerful AI systems they previously only dreamed of.”

The first tool in the AutoML lineup is called AutoML Vision. As the name indicates, it’s designed to ease the creation of models that process images.

To start a new project, a user must upload a sample dataset reflective of the photos that the AI is expected to process. It’s also necessary to tag each image with appropriate keywords — such as the names of the species in a wildlife photo — so that the model will make the necessary associations. The user then simply needs to hit the “train” button in the interface and the service will take it from there.

Google said the entire optimization process is performed automatically without the need for coding. Since AutoML makes it possible to train a model in processing the specific type of data that a company wishes to analyze, the results can be more accurate than those produced by a general-purpose product such as the Vision API.

This should theoretically also apply to any future tools Google will add to AutoML. The company didn’t share any information about the roadmap in today’s announcement, but its existing selection of AI services may provide an idea of what to expect. The Google Cloud Platform currently has tools for speech recognition, natural-language processing and automated translation among others. 

One early adopter of AutoML is Urban Outfitters Inc.’s data science team. The retailer is employing the service to automatically extract attributes such as neckline style from fashion catalog photos, data that it uses to help online shoppers search for items more easily.

Meanwhile, other providers are also working to lower the entry barrier to AI development. Last year, H2O.ai Inc. launched a platform that can pick the best machine learning model for a task from a ready-made selection and then automatically optimize its accuracy.

Image: Google