Video produced by Steve Nathans-Kelly
In this clip from Data Summit 2026, Cal Al-Dhubaib, principal technologist at Rubrik, discusses the many ways sub-optimal human interaction complicates AI performance, raising piracy concerns through over-sharing, manipulating or hacking agentic exchanges, and the lack of a shared frame of reference that would help to assess the relevance of different results for questions with more than one correct answer.
“We have weird machines or algorithms that get fooled by edge cases and we have weird humans who can’t even agree on what correct is. And it is these two things together that’s actually the problem we need to solve for when building an AI system,” Al-Dhubaib said.
Humans are unpredictable and often don’t know what they want; share too much and trust too much; relentless hackers and less forgiving of AI.
“These tools have new ways of introducing the notion of privacy and how we think about sharing information,” Al-Dhubaib said.
Several studies highlight how humans hold AI and these machines to a higher standard and when they make mistakes.
“When chatbots are named after people…individuals are way more frustrated when these Ais systems then make mistakes and they have these human-like attributes,” he noted.
The annual Data Summit conference returned to Boston, May 6-7, 2026, with pre-conference workshops on May 5.
Videos and clips of presentations from Data Summit 2026 are now available for on-demand viewing on the DBTA YouTube channel.
Save the date: Data Summit 2027 is returning to Boston on May 12 -13, with pre-conference workshops on May 11. Register now to attend.