PegaWorld 2017: Pegasystems Stresses the 'T Switch' Function

LAS VEGAS — On day two of PegaWorld, Pegasystems' annual user conference, Rob Walker, vice president of decision management at Pegasystems, took the stage of the MGM Grand accompanied by a piece of violin-heavy classical music. Only moments later, though, he pointed out that the music just played had been fabricated. "In that piece, every single note was composed by AI," he said, and the technology pulled it off by studying other compositions and stringing together elements common to those pieces. "I'm not a classical music expert, but I believe that was convincing."

In a similar fashion, an auditorium of renowned music critics was recently duped into believing they were hearing a newly uncovered masterpiece, Walker said. And by studying a collection of Rembrandt's paintings, an intelligent system composed an approximate picture that fooled many attendees. "If the audience is confused about which" painting is an authentic Rembrandt and which is the fake, "we're essentially saying that this AI passed the Turing test."

In light of such advancements, it is now becoming more important than ever that businesses make sure they are not putting too much trust in machines when the risk factors are too high and the possibility of failure is considerable. "What we need is a conscious decision on where to use transparent AI, and where to use opaque AI," Walker urged.  

He pointed, for instance, to the TayTweets Microsoft Twitter bot that learned from user data sets only to exhibit offensive behavior in its posts, coming off as racist and sexist in the process. "What if that had been a deep-learning customer service chatbot? That would have been an unmitigated disaster," Walker pointed out, potentially resulting in the violation of corporate policies and tarnished brand reputations.

For this reason, Pegasystems, he said, applies a "T switch" to its arsenal of tools that allows end users—whether they are salespeople, marketers, or customer service professionals—to adjust the settings on a scale of 1-10, with 1 being the lowest degree of trust, and 10 being the highest. This is because for certain processes, functions, and industries, customers demand more explanation from end users regarding how data is being used. With a switch, and a set of checks and balances, users can decide, Walker said, where to allow "opaque" AI—where anything goes—and where to "insist on transparency" and detailed explanations regarding how the machine arrived at a decision.   

Humans cannot, however, be second-guessing AI all the time, Walker emphasized. They must establish an efficient system of collaboration where there are "ethical sign-offs" from humans. Using adequate controls, "the outcomes will not not only be unprecedented, but also safe," Walker said.  

Related Articles

PegaWorld 2017: Pegasystems Sharpens Personalization Focus with AI

At its annual user conference, Pegasystems announces it is revamping its software tool set to provide "intelligent guidance."

CRM Evolution, Day 2: Artificial Intelligence Will Drive Customer Engagement

Industry authorities underscore the AI imperative.

CRM Evolution Day 1: Personalization Requires Technology and the Human Element

'Customers want to be loyal to you…but we don't necessarily roll out the red carpet and make it easy for them to remain loyal to us,' says an expert on brand advocacy.