AI Speak: Same Words, New Meaning

I admit there are things I didn’t know, things I now know — and plenty of things I suspect I will never know.

There was a time when cache, cookies and spam were words I understood perfectly well before technology gave them new meanings.

Cache? Something involving a hiding place.
Cookies? Something warm from the oven.
Spam? Something that came in a can.

Then came cell phones and an entirely new vocabulary. I learned to swipe, scroll, text, stream and FaceTime.

Now we have artificial intelligence, or AI, and once again, there’s a new language to learn.

Some of the words in the AI dictionary aren’t entirely new — they’ve been hanging around medicine and science for years. AI has simply given them another meaning.

Let’s look at a few of my favorites.

Intelligence
The word intelligence seems straightforward enough. But AI has complicated it.

AI systems can generate text, write code, analyze data and recognize patterns at remarkable speed. But speed and impressive performance aren’t the same thing as human intelligence.

Today’s AI systems are designed and trained by people. They don’t have consciousness, lived experiences, personal goals or a moral compass. They don’t know what it feels like to be happy, frightened or embarrassed — or to step on a Lego barefoot.

Human intelligence also has an extraordinary ability to adapt. We can learn something in one situation and apply it somewhere completely different. AI can be capable within the tasks it has learned, but it doesn’t automatically make that leap.

So when we call something “artificial intelligence,” it is worth remembering the artificial part.

Hallucinations
In medicine, a hallucination is when the senses report something that isn’t actually happening — seeing or hearing something that isn’t there. The brain, in other words, is telling a very convincing story — but the story isn’t real.

In AI, a hallucination is something different but equally concerning: the system makes something up. A generative AI model can produce an answer that sounds polished, confident and completely plausible — while being entirely wrong.

That may be harmless when you’re asking for a funny poem about your pet. It’s considerably less funny when you’re asking about a medication, medical study or diagnosis.

The lesson? Sounding sure doesn’t make something true.

Neural Networks
Here’s another term with a medical connection.

In biology, a neural network refers to interconnected nerve cells that communicate through electrical and chemical signals.

In AI, a neural network is a computer model made up of interconnected nodes that process information and identify patterns.

The resemblance to the brain gave the technology its name. But the comparison has limits. An AI neural network isn’t a tiny artificial brain sitting inside your computer.

Remember, this is “artificial” intelligence.

Training
We train for marathons. We train doctors. We train our dogs — although the dogs sometimes have their own ideas about who is training whom.

AI gets trained, too.

Instead of running laps, studying for exams or sitting for treats, an AI model processes enormous amounts of data to learn patterns.

No running shoes. No written exams. No dog treats. Just data.

Cookies, Clouds & Corgis
Who knew cookies would become digital? That clouds would move into our computers? That spam would become something we delete instead of eat?

Now AI is giving familiar words entirely new jobs.

I may never understand everything about AI. I’m OK with that.

I don’t need to know how my phone finds a restaurant, how Gemini turned a photo of my corgi into a professional tennis player, or exactly what happens inside an AI model to appreciate what it can do.

But when it comes to health care, I want more than an answer to my prompt. I want context. I want nuance. I want someone who knows my story.

AI can be part of the conversation. It just shouldn’t be the only voice in the room

Happy reading,
Suzanne Daniels, Ph.D.