What Is AI, Really?

Dr. Singh's rural clinic. Dr. Priya Singh works at a clinic in Sioux Lookout, Ontario — a remote community more than 500 kilometres from the nearest specialist hospital. When a patient needs an X-ray read by a radiologist, the wait can be weeks. In 2024, Dr. Singh started using an AI imaging tool that flags potential concerns and highlights areas for her attention. She still makes every clinical decision. The AI does not diagnose — it says "this area looks different from normal X-rays I was trained on." That flag alone has helped Dr. Singh catch three cases earlier than would otherwise have been possible. The patients are still alive.

Gurbir's grandmother in Brampton. Ninety-one-year-old Harpreet came to Canada from Punjab twenty years ago. Her English is limited. Her grandchildren, born in Brampton, grew up speaking English and Punjabi but are more comfortable in English. For years, family conversations at dinners were warm but shallow — important things were hard to say across the language gap. Last year, Harpreet's granddaughter started using an AI translation app on her phone. Harpreet could speak in Punjabi; the app would translate to English, nearly in real time. Harpreet told her granddaughter things she had never been able to say — about her childhood in Punjab, about what it felt like to leave home. The granddaughter cried. "We had three more years with her," she said. Harpreet passed away in March. "Those conversations were a gift."

Margaret's Wednesday morning call. Margaret, 74, from Ottawa, was washing dishes when her phone rang. The voice on the line sounded exactly like her grandson Tyler. "Grandma, I'm in trouble. I was in a car accident and I need $3,000 right now. Please don't tell Mom and Dad yet." Margaret felt her heart jump. She almost said yes. Then she remembered something she had learned at a library session: Stop. Breathe. Verify. She said, "Tyler, let me call you back on your number." She hung up and called Tyler's actual phone. He answered on the second ring — from his bedroom in Guelph, completely fine. The voice on that call was not Tyler. It was an AI voice clone generated from Tyler's social media videos in about five minutes.

Margaret did not lose $3,000. The 3-Second Rule saved her.

The voice that was not Barbara's daughter. Gordon, 77, from Kanata, Ontario, received a voicemail from a number he did not recognise. The voice was unmistakably his daughter Barbara — the same rhythm, the same accent. "Dad, I'm at the hospital. There was an accident. I need you to send money to this number for the deposit." Gordon sent $2,200. He called Barbara's mobile twenty minutes later. She had been at her desk all morning. She was fine. The voice on the voicemail was reconstructed from clips of Barbara speaking at a community fundraiser that had been posted on a local news website. Gordon lost $2,200. He did not apply the 3-Second Rule. "I was scared," he said. "I just reacted."

The photo that was not real. In 2024, a photo circulated in a small Ontario town's Facebook group showing the local mayor apparently accepting an envelope of cash from a developer. It looked completely real. People shared it. Some wrote letters to the town council. The mayor's office received threatening phone calls. Two weeks later, a digital forensics analyst confirmed the photo had been generated by an AI image tool. The developer in the photo did not exist. The scene had never happened. The mayor had to take two weeks off work. By the time the correction was posted, the original photo had been shared 4,000 times. The correction was shared 80 times.

The lesson: a photo being on your screen does not make it real. Images can be generated. The 3-Second Rule applies to photos and videos too — verify before sharing.

The AI that confidently gave the wrong answer. Donna, 68, from London, Ontario, asked an AI chatbot: "Can I take ibuprofen with my blood pressure medication?" The chatbot gave a clear, confident, well-written answer. The answer was wrong — it did not account for Donna's specific medication, which has a known interaction with ibuprofen that can reduce the medication's effectiveness and raise blood pressure dangerously. Donna took the ibuprofen. Her blood pressure spiked. She ended up in emergency care.

AI chatbots are trained to sound confident. Sounding confident is not the same as being correct. For health, legal, and financial questions — always verify with a qualified professional. AI is a starting point, not the final answer.

The transit times that were out of date. Eleanor, 74, from Stratford, Ontario, had an early-morning appointment out of town and needed to catch a bus she had never taken before. She asked an AI tool for the schedule, the stop, and the route. It answered right away and sounded completely sure — departure time, platform, the whole trip laid out. The times were wrong. The AI was working from information that was months out of date, and it had no way of knowing the schedule had since changed. Something felt off to Eleanor, so she phoned the transit line to double-check. The real bus left earlier than the AI had said. If she had trusted the answer, she would have missed the bus and the appointment. The AI never warned her it might be out of date — it sounded exactly as confident about the wrong times as it would have about the right ones.

The lesson: AI does not know when its information is stale. For anything tied to a specific time — a bus, an appointment, an opening hour — treat the AI's answer as a starting point and confirm it at the source.