Google had announced a new product as part of its Cloud enterprise offering: Cloud AutoML, which relies on providing machine learning-based solutions to all businesses, irrespective of whether they have expertise in the field or not. Auto ML stands for automated machine learning, where the entire model can be created automatically and a business can provide its own data-set to have a custom ML model for themselves.
Google’s Cloud AutoML is being released for the vision API and will let businesses create models, which rely on the vision-based machine learning. In a video-call interaction, Google Cloud’s chief Scientist of AI/ML Dr Fei Fei Li explained that when it comes to machine learning, it is still a fairly expensive investment. “Machine learning and artificial intelligence is a rare expertise, and we require high talent and investment for this. Currently there are around one million data scientists and at best, AI services can be used by these people. But for developers, AI can be an expensive resource,” she pointed.
This is the reason why Google is moving towards automated models of machine learning, where a business can just bring its entire data-set to the Cloud AutoML Vision API and create an AI system for their needs. “Our aim is scale AI for more people. AI building blocks for text, speech, video, vision can be provided by our machine learning APIs. However, these building blocks only offer generic solutions. To solve specific problems, can be complex and time intensive,” she explained.
For instance, she pointed out that a vision-learning system that can differentiate between cats and dogs, might not be very useful for business that needs a machine learning system to recognise shoes of different colours. With its latest Cloud AutoML feature, Google hopes to make ML available to those business which might not necessarily have access to ML/AI engineers.
Google’s Cloud Auto ML will have a drag-and-drop interface to let businesses easily upload images, train and manage models, and then deploy those trained models directly on Google Cloud. Google is working with brands like Urban Outfitters, Disney and others on deploying this solution. The reason Google is launching vision API first in its Cloud AutoML solution is because image recognition has improved significantly compared to other models that the company has used in the past. Dr Fei Fei also revealed that Google’s Cloud AutoML has worked in close collaboration with Google Brain team and other AI/ML teams in the company.
Google’s Cloud AutoML will be a global product available from today itself. Google says it will also be offered as an option for businesses in India. However, the search giant is not the only cloud player offering the option of customising Vision API for businesses, but it hopes to stand out in terms of quality. The company’s vision-based machine learning is powering many of its consumer products already. Google says it will charge businesses based on the resources used for these ML operations since this is a fairly resource intensive operation.
Dr Fei Fei also pointed out that Google will continue to focus on increasing the community of data scientists and ML experts. She said there is a large and growing global demand for talent in the field of ML and AI and currently there is a global shortage in this field.