Images: the next frontier in data analytics?

Barclay’s global data chief says we’re still at the starting line.


Barclays Bank global chief data officer, Usama Fayyad, says the next great leap forward in data analytics will happen when technology finds a way to make sense of image and video content.

He told the crowd at yesterday’s UTS big data summit that the field of analytics remains constrained by its reliance on text.

Even one of the world’s leaders in leveraging vast volumes of information, Google, still uses keyword labels to drive its image search, he pointed out

“You could easily come up with websites that trick Google by uploading some images with the wrong labels,” Fayyad said. “This is how primitive we still are today when it comes to understanding images.”

The exception, he argued, is faces, a domain being pushed forward by social networking giant Facebook at what he described as an almost alarming pace.

“Facebook is starting to do some interesting things with faces because it has recognised that facial recognition is nothing more than six or seven attributes about the distance between eyes and nose etcetera.

“Today they are working on matches, which is kind of a scary thing. They can auto-tag you so there is nowhere to hide,” he laughed.

But the big data expert, who also has had stints at Yahoo, Microsoft and NASA under his belt, said the “next frontier” and the next Google would be created by whoever can make meaning from non-text content.

“You want to create the next great search engine? he asked. “Look at YouTube, it is still sorted by text descriptions. There is zero understanding of what is inside that movie.”

“Imagine what happens when technology truly starts understanding images.”

Fayyad also pointed out that one of the other major shortfalls of big business is a failure to act on data-driven insights in time.

He asked the audience to imagine a prospective customer who all of a sudden started searching for travel websites, hotels, hire cars and tickets. The longer that activity is tracked, the more certain the algorithm becomes that its target is planning a holiday.

“But then what happens in real life?” he asked. “All of a sudden I’m done, I bought the damn tickets. I don’t want to see anything more about tickets.”

He said many big data systems aren’t reaching their customers before that opportunity falls away.

“In the world of statistics that disconnect is huge.”

Author: Paris Cowan

View Online / PDF

Leave a Reply