It’s increasingly clear that when it comes to artificial intelligence (AI), many organizations will be able to leverage investments made by IT vendors that are being made available as open source code. The latest example of that trend is a decision by Salesforce to make TransmogrifAI, a machine learning library that makes it simpler to consume large amounts of structured data, available as open source code on GitHub.
Shubha Nabar, senior director of data science for Salesforce Einstein, the AI platform developed by Salesforce, says the decision to make TransmogrifAI open source is driven by primarily by a desire to make AI technologies readily available and easily understandable.
“We want to demystify it,” says Nabar.
Nabar says even when dealing with structured data, the whole process of collecting the data that needs to be fed into an AI model is laborious. Salesforce developed TransmogrifAI to automate that process at scale as part of its Einstein AI development efforts. IT organizations that are looking to apply AI models to massive amounts of structured data can now leverage those development efforts.
That approach also makes it simpler for organizations to consistently show how an AI model was constructed, which has significant implications for both being able to collaborate on the building of AI models as well as audit how recommendations were derived, says Nabar.
Over time, it will be interesting to see how enterprise IT organizations wind up consuming AI technologies. Obviously, applications such as the customer relationship management (CRM) applications developed by Salesforce will be among the first places organizations get exposed to AI. But it’s also now only a matter of time before enterprise IT organizations become the beneficiaries of a wide spectrum of open source AI software that they didn’t have to hire a data scientist to necessarily build on their behalf.
