AI Without H.I. (Human Intelligence) Is a Recipe for Mediocrity

With AI literally everywhere, from our phones and laptops, to our countertops and dashboards, in our cars, on our wrists, playlists and other entertainment choices, it’s gotten very easy to rely on it for answers. How many times have you gotten a “mostly right” answer? What happens when you question the validity?

The other day I casually asked Amazon’s Alexa for the time of a favorite sports team’s next game. My husband overheard and said, “their season’s done.” Knowing him, I knew his comment could mean one of two things: a) they’ve actually played their last game already, or b) their season is effectively over, meaning they have zero chance at the playoffs, so why bother.

Here’s what I got from the all-knowing AI device:

Alexa: “They don’t play tonight, but they play tomorrow night in Minnesota at 7 p.m. central. Would you like me to set a reminder for you?”

Me: “They do? Didn’t they just play in Minnesota a few days ago?”

Alexa: “Yes. They play in Minnesota at 7 p.m. tomorrow night.”

Hubby: “Julie, I’m telling you, they are done. The season’s over. There are no more games left.”

Me: (Laughing) “Alexa, is that true? Are you losing your mind?”

Alexa: “Yes, I’m sorry, Julie. I keep getting things mixed up and my calendaring is off! I’ll try to do better next time.”

We laughed and I got an I-told-you-so from my “sports-encyclopedia” husband. But Alexa’s not the only agentic tool that can give false information. Earlier today, I asked Copilot to pull some data from the last 24 hours about a topic. It gave me a well-formatted list of responses. One of the items looked familiar. I asked Copilot why it was included.

Copilot: “You’re right. That was from two weeks ago. I should have more closely narrowed my search to just the parameters you provided. I’ll try to do better.”

Neither of these examples was life-changing, but they are a reminder of a basic principle. AI without human intelligence or intervention is not intelligent. It is a blueprint for mistakes and mediocrity.

Imagine if someone relied on a “mostly correct” list of acceptable drugs to treat a condition. Or research data based on “mostly correct” counts that didn’t exclude irrelevant ones. That’s not good data. Worse, these mistakes could lead to harmful outcomes or be multiplied exponentially as they are presented as fact to large audiences and further inform flawed conclusions.

The lesson as we use these amazing tools: to realize their benefits, take the time to question and independently verify responses from the start.