AT&T claims industry’s first open source AI marketplace
by Martha DeGrasse
AT&T says it is the first company to provide an open platform for developers to combine artificial intelligence applications through a drag-and-drop interface. The goal is to enable AI applications to work together even if they were created using different software languages. AT&T expects AI to expand the number of services it can offer its customers, and has already described many of these use cases.
Acumos, which AT&T developed in partnership with Tech Mahindra, was announced last fall and is now publicly available through the Linux Foundation. AT&T says it will enable developers to edit, integrate, compose, package, train and deploy AI microservices to efficiently create complex and sophisticated AI services. The company expects Acumos to make machine learning apps more accessible to developers who may not have extensive experience with artificial intelligence.
“We’re not aware of anyone else that has this capability and we think it’s really powerful because it’s going to allow all these network services, not just ours but others …. to really take advantage of AI in a much simpler way,” said AT&T CTO Andre Fuetsch. When Acumos was first announced last fall, Fuetsch explained that image recognition algorithms will be among the most important AI applications for AT&T.
TensorFlow, SciKit Learn, RCloud, and H2O are among the toolkits supported by Acumos. Java, Python and R are also supported. AT&T says models built with any of these languages can be automatically onboarded, cataloged, and packaged into an independent, containerized
microservice. All Acumos microservices are expected to be interoperable with one another.