⁂ Article
Confidence Is Not the Same as Truth: What my mother, social media influencers, and AI hallucinations have in common
I have a lot of great memories of mom, aka M'dear. She had an infectious laugh that would fill a room until its echoes were bouncing off the walls.
I was very curious as a young child and M'dear would answer my every question with unshakeable confidence.
Even if she was wrong.
I have a lot of great memories of mom, aka M'dear. She had an infectious laugh that would fill a room until its echoes were bouncing off the walls.
I was very curious as a young child and M'dear would answer my every question with unshakeable confidence.
Even if she was wrong.
I don't remember the exact question, but I do remember looking up into her eyes as she answered. I could actually see the confidence in her answer. It wasn't until several days later that I figured out she was wrong and confronted her about it. Her answer was simply, "I didn't know the answer".
I can understand why she did it; the earnest questions from her child were just too much pressure. She did not want to let me down by saying, "I don't know". Maybe she thought I would think less of her? Feel disappointment?
When I learned that mom didn't have all the answers I was a bit disappointed but I learned also that mom and dad were simply human. Like me. It was forgivable because ultimately, she was only doing her best to be a good mom.
Social media influencers are also human but are driven by different needs, such as the need to acquire that lucrative contract and the need to have an ever growing audience.
I was on social media the other day watching video shorts. I stopped to watch a social media influencer tell me how some foods were processed before being sold to us consumers. He looked directly at the screen and delivered his message.
His confidence was palpable. The comments section was full of "thanks for this info" and "I'm so glad you took the time to research this for us". It was easy to imagine hundreds of people nodding along, sharing with friends, and accepting this information as fact.
But that was just it. Everything didn't quite add up. A quick web search confirmed it. He was spreading pseudo-science to hundreds of people.
Don't get me wrong, I don't believe the vast majority of influencers start out with malicious intent. Maybe they were seen as "experts" in their real-life social circles. When they got online and their audience grew, so did the pressure to always have the right answer.
That pressure, that drive can easily lead them to the point of deception, where the influencer's online persona has forked from their real life persona. The danger is that their audience is unaware of the split.
I felt that same confidence myself, recently. A few months ago, I was researching ideas for a print-on-demand business. I wanted to be able to boot-strap this business with little-to no upfront costs. Everything would be handled by the POD website.
As I got started, things were going great. Every question was answered by the AI with total confidence. Every scenario, every brainstorm session went smoothly, without the hint of doubt.
It was when I was pretty deep in designing the process of order fulfillment that I realized me and the AI had parted ways several steps ago.
Those assumptions snowballed until the gap between what I wanted and what the AI thought I wanted was huge.
At first, I had no idea where things had gone wrong. I even blamed myself until it occurred to me to simply ask.
So I did.
I asked the AI what assumptions it was making during the course of our planning. I was surprised at the number of assumptions made.
AI delivers that same confidence and we eagerly gobble it up. But there's no one behind it.
It has no emotions. It doesn't feel any pressure to not disappoint me like my mom felt, and it doesn't care about always appearing right or infallible in front of their audience like influencers.
AI is simply providing the most-likely plausible output to our input.
AI is not maliciously deceiving us. Not even on accident. I would go so far as to say that every hallucination an AI gives us is our own fault.
In many ways, AI is a magnification of the humans who programmed it and a magnification of all the content it consumed while AI was learning. Yes, programmers wrote the code, but it was trained (or fed?) on the best and worst society has to offer.
Confidence does not equal truth. We must override the instinct to trust confidence itself. Expect hallucinations, our own and AI's. Take five minutes. Just ask.