What My Mother Taught Me About Trusting Machines

Note from the President

May is Mental Health Awareness Month. It is also, in our region, the month when Mother's Day quietly does more emotional work than any official observance. We notice the mothers. We name what we have not always named. We try, sometimes clumsily, to be the kind of family that the women in our families have always built around us.

Naila Rampersad has written a personal piece this month. I asked her for it because I think the question of what an older Caribbean woman makes of a chat interface, of an agent, of a phone that suddenly answers back, is a question we as an industry mishandle when we treat it abstractly. Naila handles it close.

A short technical note. Anthropic released its Claude 4 model family last Thursday. Several of our members have asked what the implications are for the workflows they have set up over the past year. The short answer is that the new models are meaningfully better at long-running tasks and code. For most members, the practical change is small. If you are running production systems on the API, your engineering team should evaluate the new versions for the things they need to be evaluated on. If you are using the consumer products, the upgrade is automatic. Get on with your work.

Happy Mother's Day to the mothers, grandmothers, godmothers, aunties, and the women in our lives who do the carrying.

Adrian Dunkley Founder and President, Caribbean AI Association

---

Feature

What My Mother Taught Me About Trusting Machines

By Naila Rampersad

My mother is seventy-one. She lives in the same house in San Fernando she has lived in since 1979. She raised four children in that house, buried a husband, retired from forty years of secretarial work at an oil company, and made the best stewed chicken in our half of the country. She is not, by any conventional measure, an early adopter of technology. She owned a fax machine after most people stopped using them. She learned email in 2009. She still keeps a hardcover address book on the kitchen counter, and she writes new numbers in it before she puts them in her phone.

In January, my niece set up ChatGPT on my mother's phone. By April, she was using it every morning.

I want to tell you what she uses it for, because the pattern is, I think, not what the AI industry expects, and it has changed how I think about who these tools are actually for.

She uses it to figure out medications. She has three. Her doctor explains them in a hurry. The pharmacist, when she has the energy to ask, gives her the answer that fits the time available. The leaflets that come with the medication are written by lawyers. When she got home in February with a new prescription, she opened ChatGPT and asked, in her own words, what this pill was supposed to do, what it was supposed to feel like in the first week, what it should not be taken with, and what side effects would justify a call back to the doctor. The model answered her in a calm, conversational, patient voice, in language that did not assume she had been to medical school, and it told her to confirm with her doctor or pharmacist, which she had not been able to do. She told me about this on a Saturday afternoon and then said, in the same breath, that the model was the first thing in her life that had explained her medicine to her without making her feel old.

She uses it to read official letters. My mother has spent fifty years of her life dealing with letters from the bank, the insurance company, the National Insurance Board, the City Corporation, the tax office, the utility company, and the various government departments that periodically write her with what we collectively call "regulatory poetry." She has always been competent at this work. She has never enjoyed it. Now she photographs the letter, asks the model what it is asking her to do, what the deadlines are, and what the consequences are of doing nothing. She makes a decision. She writes the reply by hand if a reply is needed. She told me, with a certain amount of pleasure, that the time she used to spend dreading these letters has been redirected to her flowers.

She uses it to remember things. My mother's working memory is not what it was at fifty. Whose is. She has begun, in the past three months, to keep an ongoing conversation with the model about her week. What groceries are running low. What her doctor said. What her grandchildren mentioned that she should follow up on. The conversation is open every morning when she has her coffee, and she adds to it through the day. It is not a journal. It is a kind of extended kitchen counter for a kitchen counter that ran out of room a long time ago. Whatever the AI industry thinks it built, this is what my mother uses it for, and I find it harder to think of a better use.

I want to dwell on what she does not use it for, because the absences are also instructive.

She does not use it to have an opinion. My mother has opinions, formed by seven decades of attention and not lightly given. She does not consult a chatbot for those. When I asked her, last month, whether she had asked the model what it thought of the political situation in our country, she looked at me as if I had asked her if she discussed her marriage with the gardener. She said no, child. It does not know us.

She does not use it for company. This is the thing the technology press has been most anxious about with older adults. The chatbot as substitute friend. My mother is sociable. She speaks to her sisters on the phone every day. She walks to her neighbour's house in the evenings. She goes to church on Sundays. She has, in short, a life. The chatbot has not replaced anyone in it. She does not think of it as a person. She thinks of it as a useful drawer in the kitchen.

She does not use it for big decisions. When my father's small estate was being settled, she did not ask the model what to do. She asked her cousin who is a lawyer, and her oldest sister who buried her own husband six years before. The model was not a candidate for that kind of question. She understood, without anyone telling her, the difference between a tool that can read a letter and a person who can hold your hand.

I tell you these things because the conversation in our industry about who AI is for, and how it should be designed, and what its risks are, has been led, in the main, by people much younger than my mother and much more remote from her life. We have spent the last two years arguing about superintelligence and existential risk and the future of work, and the woman in my home who has actually integrated this technology into her daily life cares about none of these things. She cares about whether the pill explanation is right. She cares about whether the letter from the City Corporation is going to cost her money. She cares about whether the model will remember, between today and tomorrow, that she asked about her grand-daughter's recital.

The lessons I am taking from this, in this month of mothers, are several.

The first is that the design choices that make a chat tool useful to a seventy-one year old woman are not the same as the design choices that make it impressive to a twenty-five year old engineer. My mother does not care about response speed. She cares about whether the answer is the right shape. She does not want a feature shelf. She wants a conversation that picks up where it left off. The product decisions that have made the consumer chat tools workable for her are simple, durable decisions made by small teams. The technical sophistication is in service of the simplicity, not in display of it.

The second is that trust, for an older Caribbean woman, is built the way trust has always been built for her. By the thing being right, repeatedly, about small matters. My mother did not decide one day to trust the model. The model earned its place, one correctly explained pill at a time. The earning took two months. The Caribbean technology buyer, household and business, is going to do the same arithmetic. The companies who win this region's confidence will be the ones who get small things right over and over again. The companies who try to win it with a marketing campaign will be the ones who lose it the first time they hallucinate a child's allergy medication.

The third lesson is one I am not entirely comfortable with, but which seems true. My mother is, in the use of these tools, doing better than I am. She is not asking the model to do her work. She is asking it to be useful in her life. The distinction is not small. The week I noticed this, I went home and changed how I use the same tools. I closed the seventeen browser tabs of half-finished AI-assisted projects, and I opened one conversation about my children's homework schedule. The conversation has been running for six weeks now. My week is calmer for it.

There is a thing my mother said to me, in February, that has stayed with me. We were on the porch. She had just asked the model something and got an answer she liked. She put the phone down, looked at me, and said, "it is not the smartest thing in the house, but it is the only thing in the house that has time for me."

The room got quiet for a moment. I was not sure if she was making a joke or asking me to apologize. I think she was doing both. I think she was also telling me something that the AI industry, in our urgency to talk about what these systems will become, should hear from someone who has already made one of them part of her life. The technology will be measured, in the long run, by whether it had time for the people who needed it. My mother needed someone, in the gap between the doctor's visit and the next one, who would explain her medicine without making her feel old. She got one. The fact that the one was a piece of software does not make the gift smaller. It makes it more complicated, and it asks more of the people who build, sell, and regulate the software, because the obligation is real.

A happy Mother's Day to my mother, who is reading this on the porch, and to yours, who is wherever she is. May the technology in their lives have time for them, and may we, who built it, be honest about whether we have earned that.

---

Naila Rampersad is a journalist based in San Fernando. She writes regularly for The Monthly Intelligence Report and serves on the Communications Committee.

Originally published in The Monthly Intelligence Report, May 2025.