DataOps, automated trust and transparency in AI will form the basis for the next big trends in the data world, according to IBM Data and AI vice-president of marketing Scott Hebner.
Hebner (pictured left), who leads the global marketing and strategy for IBM's Data and AI business unit, said that the three areas would become ever-more important as they would act to encourage businesses and organizations of all ilks to move toward AI adoption.
The importance of DataOps is that it will reduce the approximately 60% of a data scientist's job that revolves around preparing data for analysis and use.
"You need to have virtualized data pipelines that are high-quality production grade from which data scientists can build and train their models off, and not have a data scientist uniquely build each one," Hebner remarked. "The developing trend of DataOps is going to help facilitate that and the technologies that help do it."
The second trend on Hebner's radar, the notion of using AI to build AI, is based on a technology developed by IBM Research and integrated within IBM's Watson Studio. AI Automation (AutoAI) uses AI to generate AI, providing people of lower technical skills with the ability to build AI models and generate them in such a way that they can be regenerated as the models learn.
Not only does the technology address a major issue within the data world – that of the generational skills gap – but it will also reduce the time required for data scientists to hand-code models. And it will radically improve the quality of those models, which can self-evolve (regenerate) as they learn.
The third and final trend highlighted by Hebner presents what he described as "a huge opportunity" in building trust and transparency in AI models.
"Some businesses are beginning to rely on AI models, but you have to provide them with the ability to understand if there's any bias in the model, so you can explain what the model is telling you and how it came to that conclusion," he said.
Watson OpenScale, which came out late last year, will help to build trust and confidence into AI models, as well as a new level of transparency, as businesses and organizations look to climb what IBM has designated as the "AI Ladder".
"If I make a decision and someone questions it, I use OpenScale to go back along the journey and explain why it has recommended a decision," explained Hebner. "It is about operationalizing intelligence around what an AI model is telling me to do, so I feel confident, and have trust and transparency in how I make the decisions."
Describing each of the three trends as "the three legs of a stool", Hebner noted that progress in each area would augment moves toward a "world of trust and transparency" around predictive analytics.
"Predictive analytics is a tricky thing, because you're predicting the future," Hebner remarked.
Hebner said that he believed his three trends – DataOps, AutoAI and OpenScale – would form the next three hills data scientists would need to climb in order to advance data innovation, with all three technologies potentially tied together.
"DataOps creates a virtual, high-quality production grade pipeline of data, ensuring you feel more comfortable that the data is the right business-ready data, and is compliant with all policies and regulations," he continued. "Then you use AutoAI capabilities to generate and build great models, quickly and with lower skill. And this is followed by OpenScale, which can detect whether the AI that built the AI model infused bias into the model, or if you want to understand the linage of how that model was created by the AI."