Potential challenges: “AIOps is still a bit of the ‘Wild West.’ Vendors are over-promising and under-delivering,” said McDowell. “Customers must assess their cost of failure. Where the cost of failure is high, customers should look to AIOps for recommendations. Where the cost of failure is low, auto-tuning can be used. AIOps ISVs, to be successful, must limit their scope rather than attempting to make predictions and recommendations with an endless number of variables and without regard for customers cost of failure.”
Future prospects: “AIOps will be table stakes in less than 5 years,” McDowell predicted. “All monitoring and Ops software [still in existence] will have a degree of AIOps capabilities.”
‘BIG CONTENT’ POWERED BY AI
Big data, along with the tools to process and analyze structured data through business intelligence applications, non-relational databases, and ETL platforms, is mainstream at this point, said Chad Steelberg, chairman and CEO of Veritone. The next area of focus is what Steelberg referred to as “big content.” This, he said, represents five times more data than big data alone and includes unstructured information, comprised of mostly audio and video content. “The explosion of this data, whether from medical devices or the proliferation of IP cameras, is filling the cloud at a rate of nearly 1 zettabyte per year and growing exponentially.”
Emerging or widespread? Big content technology is still in its emerging phase, Steelberg pointed out, “with less than 5% of the use cases defined. Much of the cognition required for the machine to understand big content and unlock 100% of its value is still emerging and in its experimental stages.”
Potential challenges: Big content increases reliance on AI, which “will be the key to our survival, not the cause of our demise,” Steelberg said. “Humans are going to have to learn to trust the wisdom of the machine.”
Future prospects: “Cognitive services powered by AI and new databases designed to store, retrieve, and understand this unstructured content at superhuman levels is the future,” said Steelberg. “In 5 years, AI will have reached a point where machines can see, hear, and understand the world at superhuman levels. Our unstructured society will be covered in sensors, feeding these intelligent services with streams of raw data.”
AutoML
AutoML will have the greatest impact on enterprises’ ability to process and manage data, predicted Alex Ough, CTO architect at Sungard Availability Services. “AutoML automatically selects the best algorithm and model for the intended purpose, enabling individuals with limited ML knowledge to successfully build and apply ML technologies. All you have to do is upload your labeled data, then AutoML will find the appropriate algorithm and return the best results.”
Emerging or widespread? “AutoML is still in the emerging stages,” said Ough. “Google’s AutoML is currently in beta, and another similar open source technology called Auto-Keras is in pre-release stages.”
Potential challenges: “Data silos can present major challenges for AutoML technology,” said Ough. “To reap the benefits of AutoML, you must have a unified source of accurate data across the enterprise and find the truth of source.”
Future prospects: “I foresee this technology being used across all industries in 5 years,” Ough predicted.