Quick Poll: Will Trust Be An Inhibitor to AI Adoption?

A very nuanced question, of course. I saw this ad on a car and thought, "eh, not so sure a human is a better answer. Could be - but not always." 

I attended the #ARInsights event in San Francisco this week. The final panel discussion focused on how to get started with AI applications and what to expect. More than one panelist raised the issue of trust as a major barrier to adoption. I felt myself disagreeing. 

Trust has so many caveats. Bottom line, the benefits must outweigh the risks, which might include bias, incorrect answers, security concerns, malicious intent or behavior, and more. Trust is a factor. It may not be the biggest one. Humans also make a lot of mistakes. Just because we trust the person, it doesn't mean we don't check their work. 

Many organizations are held to high standards. Examples include regulated industries, the government, the military, first responders, etc. The risks are high. They are also held to higher standards. 

Many of our day-to-day decisions just don't carry that much risk. Keep in mind that humans also see confirmation of existing points of view. 

Made me think of autonomous vehicles. They have been omnipresent in my neighborhood in San Francisco for about six years. Initially, they were a novelty. They made mistakes. I tested them. I stepped out into crosswalks to see what would happen. I tried to open the doors of parked cars. Mistakes forced GM's Cruise to withdraw. Today, I trust a Waymo (Google) more than I do most human drivers. They aren't tired, distracted, or emotional. They always stop in the crosswalk when I am there with my dog. They don't beep horns or shout at anyone. There is no road rage. They follow speed limit guidelines. They don't make left-hand turns from the right lane. 

My friends and colleagues John R Rymer and Rob Koplowitz just published a new piece of research to guide enterprise decisions around AI use cases. You can find our research on their website or mine. julieask.com/reesearch or analysis.tech/research