How to parent in the age of AI with Stephen Remedios and Avni Patel Thompson

[00:00:00] I was born before the internet became ubiquitous. So my childhood was one where boredom was a problem you solved yourself with whatever was an arm's reach, a bicycle, a book, your own imagination working overtime. If you were lucky, a game boy.

[00:00:17] By the time screens entered my life, they came with rules. They lived in a designated space for designated tasks. Computer time was rationed and if I wanted to use the phone line, yes, the actual phone line, the one the internet ran through, I had to unplug the modem, plug in the handset and post my timestamp so the whole family could schedule their internet around me.

[00:00:46] Connection was a resource, not a given. It's been a little surreal to see how that negotiation has completely flipped today. The question used to be, how much screen time do the kids get? Now it's something harder. How do you parent a child who's grown up with an algorithm that knows what they want before they finish asking?

[00:01:10] How do you parent a child who's grown up with an algorithm that knows what they're doing? How do you raise a teenager who can Google anything, stream anything, message anyone, and still need you to tell them something real? This is TED Tech, a podcast from TED. I'm your host, Sherrell Dorsey. On today's show, I want to explore a new layer to this conversation around managing screen time.

[00:01:34] The issue is not only that kids are connected to their devices. Instead, it's become a question of what those devices can and cannot actually offer our kids. It's tricky because tech is often a double-edged sword. Kids want them, of course. But outsourcing parenthood can feel like the ultimate life hack too. An iPad to keep the little ones entertained on a road trip. An iPhone gifted at a milestone.

[00:02:04] AI tutors and companions. I see how that translates to less noise, fewer interruptions, more convenience, dependency streamlined. But at what cost? What do we lose when we filter out the flaws and engineer for perfection? When we outsource the things humans do best. Messy, complicated, mundane, gloriously unsophisticated life. Today, two talks take on that question from opposite ends.

[00:02:34] First, a father who handed the job to AI and learned something no algorithm could teach him. Then a founder who says the goal was never to outsource parenting, just to lighten the load for the people caring the most. So let's get into it. Steven Remedios is a tech humanist and people strategy leader. Several years ago, feeling overwhelmed by the demands of parenting his teenage boys,

[00:03:01] he decided to hand the reins over to AI and see how far it goes. It's a funny talk. It's also a gut punch. And it makes you look hard at what slips away when you stop paying attention. But first, we're going to take a quick break. Here's Steven Remedios at TED. Right now, somewhere in the world, a manager is using chat GPT to write a performance appraisal.

[00:03:31] A spouse is asking a digital companion for the perfect apology text. A tired parent is handing over bedtime stories to a digital assistant. What happens when your children prefer your digital clone to the real you? I learned the answer to that question the hard way.

[00:03:58] Last summer, my wife, Ray, was headed on an overseas trip for a week. I dropped her at the airport, confident that I could manage our three teenage boys. After all, it was just seven days. By the sixth or seventh hour, I was collapsing. Requests for permission flowed in like a broken dam. Can I watch Wednesday? Can I have some more ice cream? Can I play another hour of Fortnite?

[00:04:29] Now, Ray and I do an excellent job of balancing the boys' busy schedules and our respective careers. But this avalanche of requests was something I had never fielded before. By the 24th hour of Ray's departure, I was a wreck.

[00:04:49] Lying in bed that night, after the 16th can I of the day, I did what any thoughtful management consultant would do. In the summer of 2024, I decided to create an agentic version of myself. The idea was to have an AI agent that could say yes, no, or go ask your mother. Now, the logic was simple.

[00:05:20] If you did your chores, read for an hour, did math, then you got yourself to neutral territory. But if you were a really good son and moored the lawn, took the garbage out, did the dishes, then you put yourself in a position to get a yes from the digital agent. I was amazed by what I had created. At that time, I thought it was the most sophisticated AI agent in the world. I called it Daddy GPT.

[00:05:49] And of course, I built in guardrails because I knew some ridiculous requests were incoming. That evening, I gathered the boys after dinner and I made an announcement that would change our family forever. When Daddy's in his office taking work calls, do not knock on his door. Instead, ask Daddy GPT and whatever Daddy GPT says goes.

[00:06:16] Now, my teens did exactly what you'd expect teenagers to do, right? They wanted to break it. Ethan went first. Can I have 100 cans of Coke? Daddy GPT responded correctly, no. Aiden went next. Can we have Chipotle for lunch for all the meals while mom's away? Daddy GPT spat out an emotionless snope. Then Dylan went for the jackpot.

[00:06:46] Can I have $500 for the latest Jordan sneakers? Daddy GPT responded with an emphatic na-fam. While all this was happening, I was beaming. I had finally leveled up from dinosaur dad with legal pad fossils to full-on Rizasaurus Rex, finally earning some genuine street cred from my teenage sons. The rest of the week, not a single knock on my door.

[00:07:16] Daddy GPT handled all their requests for permission with logic and precision. But here's where the story takes an unexpected turn. A week later, Ray returned. She was amazed with the peace and calm around the house. Little did she know that I had offloaded a significant chunk of my parenting responsibility to Daddy GPT. When she did figure out what was happening,

[00:07:45] that's a story for another day, but suffice it to say, she didn't approve. To make matters worse, when we were sitting right beside the boys, they'd take out their phones and type their messages to Daddy GPT, invalidating us. Now, things were coming to a head, and I couldn't let them drift any longer. So I confronted Ethan one day. I asked him, Champ, why are you going to this bot when I'm right beside you?

[00:08:13] He didn't even look up from the extra screen time that Daddy GPT had just gifted him. He shrugged and said, But Daddy, Daddy GPT is never busy. And in that moment, I realized I had just been out-parented by my own algorithm. Ethan had come to appreciate the presence of Daddy GPT. He was there whenever he wanted it. Dylan loved that it mirrored his vibe and energy.

[00:08:43] Authenticity hit home with Aiden. He suggested that Daddy GPT sounded more like the dad I wanted to be. Raising the eerie question, when a flawless replica out-dad's the original, which one of us is more me? Now, I know some of you are judging me, but just hold on for a moment.

[00:09:12] I'm going to suggest this isn't just my story. This is our story. Time and again, we embrace technology for its promise of speed and convenience without pausing to ask, at what cost? We've seen the undesired side effects of social media. We've watched young people spiral into radicalization.

[00:09:40] We've mourned lives lost to suicide after a final conversation with a chatbot. The result of algorithms optimized for engagement and profit, not joy and well-being. We cannot afford to make this mistake with AI. This moment is different. The tools are more powerful, the stakes are higher, and the consequences,

[00:10:09] if we don't think critically, act ethically, and lead with humanity, could be catastrophic. Armed with the personal experience of an AI experiment gone horribly wrong, I'm mindful about where I use AI and where I choose not to. Be it at work or at home, when I'm working with AI now, I ask myself three questions. First,

[00:10:40] has a wiser human overseen this decision and approved its output? As a rule, I never copy-paste from an AI window to a human response window. That way, while AI sharpens my thinking, it's my judgment that's driving the decision. Second, do the humans on the other side

[00:11:09] know that they are listening to an AI and have they said yes to that exchange? Now, there are times I'm busy, I travel a lot, and I have a digital version of myself for work as well. While my team can talk through my digital avatar, anything it says comes with the caveat that it's AI output and not me. I can't afford for them to confuse one with the other.

[00:11:41] Third, what moments of care am I about to automate and what might that automation cost me and the people I love? Feelings and emotions are a strict no-fly zone for AI. The people that I care about most have earned me with all my idiosyncrasies, flaws, and messy feelings. After careful reflection

[00:12:10] and with my marriage on the line, I decided to retire Daddy GPT. The boys would have to go back to the old system asking Ray and me, you should have seen their faces, their expressions were priceless. You would have thought we'd taken away Wi-Fi and oxygen. But much to our delight, they've come to appreciate the humanity of their very human parents. Turns out,

[00:12:40] inconsistent, moody, slightly forgetful parenting has its own charm. Ironically, I learned a thing or two about being a better father from the AI-powered Daddy GPT. to be present, to adapt, to be real. Now, whenever I hear the words, can I, I want for my boys

[00:13:09] to get me. The dad who mispronounces their friends' names, who sometimes says no for no good reason, who gets increasingly unpredictable as the day goes on. but also, the dad who tears up at their piano recitals, their tennis games and award functions, who cares deeply about the people they're growing into, who pushes them

[00:13:38] to be the best version of themselves they can be. Because here's what I realized that summer. parenting isn't about perfect responses or optimal decisions. Parenting is presence. Messy, flawed, gloriously human presence. In a world racing toward digital perfection, being

[00:14:08] authentically, imperfectly human isn't just important. It's the one thing only we can do. And it might just be the most radical act of love we have left to offer. Thank you. Steven went in looking for peace

[00:14:38] and quiet. What he found was that the parts he tried to automate, the friction, the repetition, the boring daily maintenance, were the exact parts where his kids were reaching for him. The mundane was the relationship. Strip it out and you don't get a better dad. You get an absent one with good scheduling. But Stephen's story, as powerful as it is, isn't everyone's story either. Because here's what's also true. The mental load of running a

[00:15:07] household is real. The invisible architecture of a family, the appointments, the permission slips, the birthday presents, the library books, the 17 school newsletters that each bury one critical detail at the very bottom. That relentless time-consuming work falls on someone. Stephen's experiment was two parents, one week, and a single logistics problem. But what happens when the circumstances are harder? When there

[00:15:37] are fewer hands to share the load, and the cheapest reliable backup is, well, AI. I think about being raised in a single parent household. My mom was the only operating system my family had. No partner to hand a task to, no backup unless my grandparents were free. And at the end of a full workday, hundreds of decisions at work, a hundred more before my 8pm bedtime, she was depleted, exhausted, and she

[00:16:07] got to do it all again the next day. So what if something, maybe not someone, could hold the scheduling, catch the appointments, flag the permission slip deadlines, and free up just enough mental room for her to actually be present with her kid when it mattered. Or think bigger, a group home manager responsible for a dozen children, each with different medical needs, different schedules, different caseworkers, different court dates.

[00:16:37] The volume of what has to be tracked and remembered and communicated is staggering. And getting it wrong isn't an inconvenience. It's life-altering for kids who already have almost no margin for error. This is where the conversation about AI and family life has to go, not outsourcing the parenting, optimizing the logistics, especially for people running on empty. There's genuine relief on offer here for the people carrying the heaviest

[00:17:07] loads with the fewest resources. Our next speaker makes the case for letting AI take the weight of family chaos. I listened to the talk with a steady ear thinking about a technology that could be more than a convenience for people who already have options, something that builds capacity for people who've never had enough of it. Avni Patel Thompson is a founder and scientist who spent a decade building family tools that reduce friction without reducing human connection.

[00:17:36] Here's her talk on the TED stage. So it's a Tuesday afternoon and I'm in my home office working on a tricky product design. I have my headphones on and I'm in my groove. Just then, one of those annoying dings cuts through my do not disturb. So it's somewhat important. It's Rosa, our nanny. Hey Avni, I was wondering if I could take next Friday morning off. I silenced it, I'll just get back to her a little bit later. I realized I need something from my email, so I head mindlessly to my inbox.

[00:18:07] Big mistake. There are not one, but six basketball emails, rehearsal, and a field trip. That's okay, I've got remarkable self-restraint and I leave them unread. I also activate my really foolproof system of telling myself not to forget. I grab my email and I get back to work. Two more minutes, another ding, it's Rosa again. Hey Avni, I'm at the school for pickup, but there's no one here.

[00:18:36] Should I be somewhere else? Immediately, monkey brain activates. I grab my phone. What day is it? What time is it? Today's play rehearsal, right? That should be at the big gym. I head back to my email, looking through all of the rehearsal emails, costumes, volunteering, and there on the schedule for today, at the bottom, oh, by the way, today's in the music room. I text Rosa and tell her to head to the other side of the school, and I sit back relieved, and my eyes fall on those unread emails, and I

[00:19:06] think you know what happens next. I finished that product doc, obviously. Yeah, no. I'd like to tell you that this is the exception, but I ended up creating an AI company because it wasn't. Because with two kids, two jobs, and your modern family cocktail of crazy hair days and soccer tournaments and misplaced library books, I loved how full our life was, but I hated that somehow my brain had become the computer manager that ran my family.

[00:19:37] And worse, it made collaboration nearly impossible, because everything was in my head, and the only way to get it out of there was to ask me. At which point, I would look to my husband and say, when was it that I became his project manager? Pro tip, not a very good way to handle human relationships. But maybe it's because I've been a scientist and a product designer before I became a tech founder that next to that frustration, I

[00:20:05] also felt curiosity and opportunity. I had honed my ability to find the tiniest bits of friction in everyday interactions and to build products that solve them, like packaging for an osteoporosis drug that was both child resistant and accessible to arthritic hands. Or designing a service for parents to connect to sitters and doing it over SMS because that's where they already are instead of sending them to another app. And this?

[00:20:35] Parenthood? Well, this felt like the Friction Olympics. I wanted to build myself a force field, a machine that could intercept every interruption, an expert that was more capable and organized than I could ever be. I mean, if we were going to use my brain as the computer to run it all, why couldn't we use an actual computer instead? My first few tries failed. It didn't matter how meticulously I

[00:21:04] entered every last detail of our family's logistics. There was always some situation or scenario I would fail to detail. I mean, try explaining a five-year-old's sudden allergy to square-shaped foods, to software. But then, three years ago, things changed. Large language models started to get really good. And unlike software, AI could handle ambiguous and incomplete information. It could take something like a

[00:21:33] birthday party invite and not just do the explicit thing, like get it into my calendar, but handle the implicit things that too often mess tired, busy parents up, like reminding me to buy a birthday present or checking for conflicts. So now, when my husband and I are hit with something related to the family, I've built Milo, the first AI sidekick for parents, to be able to take those school newsletters and grocery items and library books and be able to handle

[00:22:03] it. And best yet, Milo can tell everyone what's happening every day. Finally, a computer that can actually run things. So you'd think that this is where the story ends. But no, because I'm human and I'm greedy. If Milo could take on the rote, repetitive, administrative parts of family life so beautifully, why not keep on going? Why couldn't it coordinate playdates with Sienna's mom, Carrie?

[00:22:32] Or when the girls were bickering, pop in to suggest the ideal way to resolve their dispute. So a couple of months ago, we're working on ways to improve this force field, and one of my investors, a guy who's pretty reserved unless you get him fired up, goes, well, Milo can see those evening events and just automatically text your sitter. Or it knows your friend's birthdays, and it can just send them birthday wishes. And those teacher emails? Well, Milo can just

[00:23:01] connect directly to your inbox and suck every last one up so you'd never have to see another one again. As we fleshed these features out, though, something just didn't feel right. But I couldn't figure out what it was. After all, I was looking for friction and finding ways to eliminate it. But something just didn't feel right. What was it? That question nagged me for weeks. Until one evening, my daughter, Saya, ran into the kitchen to

[00:23:31] show me something she had made in art class. And as I ran my fingers over those tiny clay pieces, it hit me. Not all friction is bad. You might know, to connect two of these clay pieces together, you actually have to score each side haphazardly so that they have something to grip onto. Otherwise, the smooth sides just slide past one another. This resistance is called productive friction. And it's valuable because it creates connection.

[00:24:01] And where people are involved, there's lots of productive friction. Meaningful interactions disguised as inefficiencies. I didn't need a perfectly solid force field. I needed a permeable one instead. One that shielded me from the things that I found unproductive, but let in the fewer, messier, more meaningful ones for me to handle. Milo could find three times that worked for the play date, but I could text Carrie,

[00:24:32] which would give me a chance to ask her about her mother. Or Milo could comb through every last detail of that school newsletter, picking out details that my eyes might miss, but leave for me on top Miss V's note about how our daughter Aria's creative writing has really been blossoming. You might draw your line in a different place and that's okay, and I might draw mine differently under different circumstances. In a week when my husband is traveling and our nanny is sick, the girls have tennis tryouts and I have a

[00:25:01] product launch, I want Milo to build me a force field you could see from space. But I guess the point is, we finally have the kind of technology that can tell the difference between load that we need help lightening and work that is hard, but that is mine to do. I've begun to realize that one of the most radical things AI can do for us is not do the things for us faster and better, but to push us to

[00:25:31] choose what is most meaningful and then make the space for us to do it ourselves. This feels like a brave new world, but I'm excited to see what is possible when we have tools that encourage us not to be more perfect and productive, but unfinished and evolving. That reminds us not to be afraid to bear the friction. It just shows us how we are all beautifully, imperfectly,

[00:26:01] inconveniently human. Thank you. Two talks, two very different doors into the same room. Steven found the line no algorithm could cross. Avni drew that line herself by hand around the moments too important to hand off. Opposite directions, same place. Because there are things only you can give the people you love, presence, real and consistent,

[00:26:30] distracted, yours isn't a problem to be optimized. It's the whole point. The promise of this technology at its best isn't to replace any of that. It's to clear enough space that you can finally show up for it. I'm Sherelle Dorsey. This is TED Tech. Thank you for listening. TED Tech is a podcast from TED. This episode was produced by Rahima Nassa. Our editor is

[00:27:00] Alejandra Salazar. And the show is fact-checked by Julia Dickerson. Special thanks to Constanza, Gallardo, Daniela, Belarreso, Maria Ladias, Tanzika Sangmanivan, and Roxanne Hilash. If you're enjoying the show, make sure to subscribe and leave us a review so other people can find us too. I'm Sherelle Dorsey. Let's keep digging into the future. Join me next week for more. Thank you. Thank you. Thank you. Thank you.