Truly Intelligent AI
Prepare for truly intelligent AI within the coming decade, predicted Gregory Dorman, VP, distributed systems and analytics, for MariaDB. “Artificial intelligence will become what its name says—intelligent,” he said. “Today, it isn’t. When vendors claim AI, they’re really describing mere machine learning—the use of an algorithm that calculates an outcome based on patterns in accumulated datasets. Real intelligence is a different animal altogether.”
For example, Dorman elaborated, “I can argue that driving a car does not require intelligence—it requires a good reaction. Self-driving cars today possess a programmed artificial instinct, with which they react to environmental conditions relayed by sensors. That’s not intelligence.” The AI technologies of the future, he explained, “will become capable of relating to human emotion, picking up on non-verbal cues during communications, deriving appropriate meaning, and understanding context during interactions, and even developing their own sense of preference.”
The engines that power AI, the machines—similar to humans—learn from experience to make predictions (supervised learning), draw similarities between objects or scenarios (unsupervised learning), and make better decisions through trial and error (reinforcement learning), said Silver. “To date, AI has proven to be very capable of solving problems where there is a lot of labeled training data, such as for supervised learning.”
Unsupervised and Reinforcement Learning
Silver also foresees significant innovation in unsupervised and reinforcement learning. “Rather than relying on large amounts of labeled data, which can be expensive and time-consuming to curate, AI systems will be able to learn more effectively based on limited sets of data,” he explained. “By adopting these next-generation AI solutions, enterprises will be able to react more quickly and effectively to changes in the business landscape.”
Edge analytics will also increase in sophistication over the coming decade, especially with increased connectivity and the use of 5G, which “will require analytics, everywhere,” said Day. “In the future—and now, to some degree—everything we have or use will be a connected device.” Enter IoT, which is a technology that will be pervasive and transparent, Day said. “The market is requiring automation because we cannot keep up with the pace of innovation and change without it. Analytics at the edge has seen significant adoption over the last 24 months. Intelligent decision making at the edge, powered by IoT, will set the bar for how organizations operate and respond to digital transformation. The decisions we make on our data will continue to move closer to the edge and where the data lives—or streams—versus a traditional data warehouse.”
Automated and integrated intelligence may even create two tiers of businesses—the haves from the have-nots. “Today’s fragmented infrastructures comprised of best-of-breed transactional and analytical technologies will become endangered species as integrated solutions rapidly mature,” said Dorman. “Winners will be those with a single, totally integrated and highly responsive platform delivered as a service. The gap separating the haves and have-nots will quickly widen between companies that are still operating on dispersed, disconnected, slow-to-react infrastructures and those that benefit from integrated transactional and analytical capabilities that produce actionable intelligence in real time.”
Quantum Computing and Intelligent Clouds
Quantum computing—still being developed in vendors’ labs—may eventually make a mainstream debut. “Significant issues like the lack of quantum storage need to be resolved before its use becomes practical,” said Andrew Mendelsohn, EVP, database server technologies for Oracle. “That said, the theoretical potential of superposition is promising when it comes to accelerating data processing.”