Tibco's AI triple play

"AI has really surged back," TIbco chief executive Murray Rode (pictured) told iTWire. "It seems to have real momentum this time," probably because the infrastructure to support it is in place.

According to chief analytics officer Michael O'Connor, the company is taking a threefold approach to AI by making its products smarter, providing tools (especially in Spotfire) for machine learning, and helping customers embed AI and machine learning into their operations.

For example, Tibco has created a recommendations engine for analytics that reads metadata and suggests suitable visualisations, said Rode, noting that the company is also working with other vendors on ways to visualise the output of machine learning systems.

{loadposition stephen08}He warned that while robotic process automation is gaining traction in some industries such as banking, the unbounded nature of such systems allows them to learn flawed processes, leading to undesirable outcomes. "The real benefit today... is in rules-driven systems," he asserted.

Chief technology officer and executive vice-president of products and technology, Matt Quinn, picked up on this theme, observing that the downside of AI is that it can be like "providing a sharp knife to a small child." Machine learning is deeply dependent on its training, so there can be unfortunate outcomes if the training data is not truly representative of reality, and so "AI today is still more art than science".

While technology can reveal patterns and correlations, said Rode, it is not currently possible to automate the part where a human asks "what am I missing?"

Even so, "AI has become interesting to people," said Quinn, as we need systems that can help us do our jobs, such as assistants that help us get to the answer more quickly.

"We're on the cusp of something really, really great" thanks to the combination of cheap and scalable compute power, microservices, and mobile interaction, he said. "We all have to work together... to achieve the single biggest shift in enterprise IT."

But Quinn also noted the current "unbridled enthusiasm" that all the world's problems can be solved with AI. That sort of hype is necessary to make progress, but the outcomes will be different to the hype, he predicted.

One of the reasons for that enthusiasm is that while "AI has always suffered from the science fiction problem" in that it is expected to work almost perfectly, widely used implementations such as Siri and Google Assistant have reduced those expectations to a more achievable level.

AI can help companies create value from the already large and still growing number of sensors that are part of the push towards IoT. "We're seeing sensors where sensors never were before," O'Connor observed.

Examples from Tibco customers can be found in a number of sectors. In the Australian oil and gas industry, data from sensors on pumps is being processed by the Spotfire analytics and visualisation software in order to interpret anomalies (such as blocked tubes) and create rules for the StreamBase streaming analytics software which can then generate appropriate alerts when the data deviates from normal.

In some cases, the alert might specify that the equipment must be checked immediately, while in others it might just flag that a particular check should be made the next time that service personnel are at that location.

In the telecommunications sector, a Tibco customer found that applying deep learning to the logs and other messages generated by various pieces of equipment revealed there were signs of impending issues a few hours before outages occurred. The analytics software now generates alerts at the first signs of trouble, so remedial action can be taken before customers are affected.

"It's not just about having the insight, it's about putting it to work," said Rode.

The writer attended Tibco Now 2017 Singapore as a guest of the company.