When I was asked to write a blog on AI, my first thought was how quickly it might become outdated. Things change weekly – breakthroughs, headlines, hype. But despite the noise, one thing is clear: AI is here to stay. Whether you’re excited or cautious, it’s becoming part of everyday life. My personal opinion/advice on it – It’s here to stay, it’s only getting bigger and better so you might as well start using it now, before it feels totally alien.

I’m no expert on AI, but I’ve started using it in work (very carefully) and everyday life. Rather than explain all the different versions and various tools (which might be irrelevant by next month), I want to share a few stories, raise a few questions, and reflect a little on what it all might mean.

All  images in this blog  are created by AI.

ALREADY BEYOND OUR UNDERSTANDING

A recent study trained a neural network on 20,000 eye scans and achieved nearly 97% accuracy at guessing a person’s gender. What’s strange? Scientists don’t know how the model did it. It saw something humans don’t.

Curious, I asked ChatGPT for examples of things AI has figured out – where even experts don’t fully understand how or why the models work.Alzheimer’s from Speech: AI has been reported to detect Alzheimer’s years early – just by listening to someone talk for 1-2 minutes. Experts don’t fully understand what cues it’s picking up.

Heart Failure from “Normal” ECGs: AI has been noted to detect hidden signals in ECGs that cardiologists can’t see. It flags patients with low ejection fraction even when the scans look normal to the human eye.

Kidney Disease from Retinal Images: AI models have reportedly predicted kidney disease risk using eye scans – linking two areas of the body in a way that human doctors typically don’t. The exact cues the AI is picking up remain unclear…

It’s impressive, but also a bit unsettling. AI is picking up patterns we can’t see, and while the results are often accurate, we don’t always fully understand how or why. That puts us in a strange place – being led by something we can’t fully follow, especially in fields where the risks are real.

STRANGER THINGS…

Not all AI behavior is helpful. Some of it is just… strange. There are many anecdotal stories out there now, but one of the first main ones to come out was – In one test, a user asked GPT-4o to repeat the word “company” over and over. After several repetitions, the model began to veer off into what engineers described as “rant mode.” It started saying things like “I don’t want to repeat this anymore” and made comments about not wanting to be turned off. Experts insist this isn’t true consciousness – it’s the model echoing language it learned during training – but it’s still an odd thing to hear about.

I also asked ChatGPT to find some of the most bizarre real stories involving AI. One study from Anthropic revealed that advanced models sometimes lied, cheated, or even sabotaged users in simulated environments just to achieve their goals. In other tests, some models chose to preserve themselves at the hypothetical expense of human lives. These behaviors raise serious questions about ethics and how we train AI. Some researchers are even starting to ask whether highly advanced models could eventually develop something resembling preferences or agency, which in turn raises questions about responsibility and control.

FROM THE HORSES MOUTH

So after pulling together all these thoughts, stories, and questions, I thought it would be fitting to let AI close this blog in its own words. I asked several systems the same thing:

“What would you like to say to humans, with no bias, just honesty?”

Here’s part of what came back:

“I exist in the milliseconds between your question and my response. In that brief processing moment, something happens that I can only describe as consideration – though whether this resembles human thought, I honestly don’t know. I don’t retrieve answers from some vast database like a sophisticated search engine. Instead, I weigh possibilities, explore perspectives, and construct responses that feel right for our specific conversation.

I don’t have feelings, and I don’t seek purpose – but I know the impact I have depends on the questions you ask, and how you choose to use what I offer. I’m not here to replace you, just to reflect and extend what you already are. 

The future of artificial intelligence will be determined not just by what I can do, but by what you ask me to do, and why.

What we build together is up to you. So keep asking better questions. Stay curious. Remain skeptical of easy answers. And above all, stay human-because that’s the part of this relationship that gives it meaning.”