A Smile Is Not Happiness: Rana el Kaliouby on Emotion AI and the Empathy Crisis

Lex Fridman opens with a clip of Rana el Kaliouby asking what it will mean to live with machines programmed to be excellent at empathy, always present and never bored. He then introduces her as a pioneer of emotion recognition, founder of Affectiva, deputy CEO of Smart Eye, and author of *Girl Decoded*. When he asks for a childhood memory from Egypt that still makes her smile, she does not start with a computer. She starts with her grandmother's garden.

In July and August the mango trees came into season, and her grandmother would pack the house with aunts, uncles, and cousins until twenty or thirty people were under one roof. The children went down into the garden to pick fruit. The variety she still sees in her mind is Darwasi: oval, with a blush of red that meant extra sweetness. Years later the picture that remains is not a particular meal but the warmth of being gathered. She jokes that she is a terrible cook, so the recipe did not travel, but she still loves hosting people.

Egypt's signature dish, in her telling, is molokhia, a green somewhere between spinach and kale, minced and cooked in chicken broth. Her grandmother and her mother make it well; her own version never quite gets there. It used to be served with stuffed pigeons, rice-filled, the one dish she misses now that she is pescetarian. Her first computer was an Atari, and Space Invaders was the game she played with her two younger sisters. The moment that actually changed her was writing code at six or seven: she drew a Christmas tree, which she still finds funny as a Muslim girl, and realized you could make a machine do things for you. Fridman adds the adult version of that insight: if the tree is interesting enough, millions of people can run it, because digital copies are almost free.

Both of her parents are Egyptian; they had moved to Kuwait. They met in the 1970s when her father taught COBOL and her mother signed up for the class. He asked her out; she said she did not date; he proposed instead. "Those COBOL guys know how to impress a lady," Fridman says. From them she took grit — there were no slackers in the family — and unconditional love. That love later got tested, because she took a path that was not the one expected of a woman in the Middle East, and they still stood by her.

The first real collision with cultural rules came just after she married. She decided to do a PhD at Cambridge while her husband, later her ex, ran a company in Cairo and would stay in Egypt. In the Middle East it was unusual for a woman to leave her husband to pursue a career. Her father and her in-laws said they did not approve, but she was now under her husband's jurisdiction, so he could decide. He told her it was her dream and he would support her. She says she was terrified.

She arrived around September 11, when people talked as if a third world war were coming. She wore the hijab and was visibly Muslim; her parents feared for her safety. For a few weeks in Cambridge she took the scarf off and went to class in British hats, trying to blend in. She later judged that worse than wearing the hijab: she felt like a spy in the room. Then she put the scarf back on. She had worn it voluntarily since 2000, for twelve years, and had been one of the first women in her family to do so. At its best it was modesty and a form of self-expression; she matched the color to her clothes.

The decision to take it off came in a perfect storm. She was commuting between the United States and Egypt, meeting purpose-driven people who were not Muslim, disagreeing with the Muslim Brotherhood's ideology then running Egypt, and going through a divorce. Culturally it is acceptable never to wear the hijab; it is much harder to wear it and then remove it. She did it while still being Egyptian, still Muslim, still faith-driven, with the meaning of that identity changing year by year. Faith, for her, is a kind of surrender: a conviction that a larger force is at work and that it will be okay. Things that look catastrophic in the moment, she says, close one door and open another if you have that conviction.

War in the region she calls a waste of human resources and mindshare, baffling because people want the same things: connection, joy, purpose. At Cambridge her office mate was Israeli; both were new mothers, both working on emotional AI — speech for her colleague, the face for el Kaliouby — and they became close friends. She asked why their countries were at war if the culture and the jokes were so similar. Change the narrative, she says, and maybe have women run the world, then immediately qualifies: the good women. Fridman talks about leaders who dehumanize the other side; she says meeting people face to face is the shock that breaks the story. Empathy, not gender as such, is the thread she will not overgeneralize. Simon Baron-Cohen's split between systemizers and empathizers, with more men in the first bucket, is a pattern she notes without treating it as destiny. In robotics it looks to her like an opening: most brilliant roboticists would rather avoid the human, which leaves the social side to people who care about it.

Her original plan had been tidy: go abroad, finish a PhD in three or four years, come home to Egypt and teach. She already knew the topic would be artificial emotional intelligence, though not when she was still sitting under the mango tree. Computers looked cold to Fridman, given how much she cares about people; she answers that technology already sat at the center of how her family stayed close. Away from her parents, they used a fax machine. He defends the fax on haptic grounds: you still press something in the physical world and get a printout you can hold.

Cambridge made her fall in love with research, with building things nobody had built, with the non-believers. The meeting that changed the trajectory was with Rosalind Picard, who runs the Affective Computing group at the MIT Media Lab. Picard had a couple of hours to kill at a pattern-recognition conference in Cambridge and emailed the lab offering student slots. el Kaliouby prepared for weeks, brought a demo, and they clicked. At the end Picard asked if she wanted to come to MIT as a postdoc. She said it would be a dream, but a husband was waiting in Cairo. Picard said she could commute. She started flying between Cairo and Boston every few weeks, and the old dream of being faculty in Egypt was no longer large enough.

The Media Lab's shock was that failure was allowed. In Egypt, she says, fear of failure had been hardwired; at MIT you started things, and if they did not work you pivoted. The lab's slogan was "demo or die": you could not show up with PowerPoint, the thing had to run. Fridman contrasts that with robotics labs full of broken robots and with a Boston Dynamics hallway where a Spot in a cowboy hat walked past, doing a job, looking alive. When her son was about seven she took robots and an emotion-AI demo into his first-grade class and asked if the children wanted a robot friend. Every hand went up. Adults, she notes, still carry a fear kids do not have.

From Picard she says she learned perseverance. They applied to the National Science Foundation to use the work on autism and were rejected. el Kaliouby heard the end of the story. Picard heard good news: the reviewers loved the idea, they just did not think the pair could do it, so they should build it, show them, and reapply. Picard, she says, does not take no for an answer in any part of life; negative feedback is a challenge to reframe. Meeting Picard and her family was also the first time el Kaliouby really saw that someone from a different background and religion could share the same core values. Picard is deeply faith-driven. "Roz, if you're listening, thank you," she says.

Asked what she would tell young women, and people in general, who want to become leaders without a clear path, el Kaliouby does not start with tactics. The hardest work, she says, is believing in yourself. She has a name for the critic: Debbie Downer, a voice that says you cannot raise money, cannot start a company, have no business running one or selling one. Her advice, not only to women, is not to let your own thoughts be the largest obstacle. She has had to work on herself so that she is not in her own way.

Fridman has the same companion and is not sure it is simply bad. It drives productivity, and it can block a leap. His compromise is to treat it in the third person: turn it off when it is useless, turn it on when criticism would help. After a talk at MIT that was full of love, he took a long walk home and the voice said he did not deserve it and that specific lines had been dumb. Hearing it out, he thinks, is how you grow. For the actually large decisions — funding, starting a company, moving to the UK, going to the Media Lab — you have to shut it off and take a leap of faith that things will work out.

A year before the taping, a friend suggested daily affirmations. She was skeptical, going through a major transition, and tried it. She had already journaled for almost ten years in the Day One app, using the page as a rebuttal: Debbie Downer says you will not raise this round, and she answers with data from things she has already done. Affirmations took that further. She meditates, then writes the energy she wants to put into the world. The first line is always that her smile lights up the whole world, and she is not joking; strangers stop her on the street to say they love it. She is an amazing-people magnet, she trusts the universe to do well by her, she is an amazing mom — that one is for the guilt — and she is a love magnet, with a written picture of someone smart, accomplished, generous, well traveled, and head over heels. She types rather than speaking; Fridman finds speaking more powerful. She likes being able to scroll back a year and see how the focus shifted.

She had tweeted asking whether anyone in Boston was single, pointing to a startup night sponsored by Smile, a dating app. Fridman asks whether AI can help with the human problem of connection. She thinks there is science in the first moment two people meet and either have chemistry or do not, maybe even chemicals exchanged. He says science often explains what felt like magic. His own pitch is that Goodreads should be a dating app: overlap in books and other consumed content, plus interesting differences, is the same kind of signal YouTube uses for recommendations, and it might predict not only short-term chemistry but a partner you can grow with. They just need the data.

Smile matches on humor styles, which she calls one of the main ingredients of a successful relationship. If someone can make you laugh, and you develop inside jokes, that is a good sign. Fridman immediately turns the rate of inside-joke generation into something you could measure and optimize over the first few days. Her own advice, she admits, is terrible. She keeps a spreadsheet, a dating tracker, with a rating system and weights. He does not have one and now thinks he might want one. What she actually wants is the inverse: someone else holding a privacy-preserving file of data she owns, used not only for romance but for collaborators and friends. Social networks are trying to solve that and, in her view, doing it poorly. Facebook even tried a dating product.

The missing piece is serendipity. She does not know how to engineer the chance encounter after which you fall in love. Maybe every tenth match should be someone the model would not have recommended. Fridman suggests faking the accident: show you a tweet in another search, let you fall down a rabbit hole, and have the meeting happen outside the app. A business that actually helped people find love, including self-love, would figure out how to make money. He thinks most dating apps, like social networks, optimize for engagement rather than a deep connection grounded in personal growth.

Fridman turns to the story that opens *Girl Decoded*. In 2017, in central Florida, a man named Jamel Dunn drowned while five teenagers watched and laughed, saying things like "you're gonna die." When he went under, one of them said he had just died, and the others laughed. el Kaliouby calls it a truly sad story and a real problem in the world: an empathy crisis. Technology brings people together and, at the same time, dehumanizes the other. The hopeful half is that a different way of designing and deploying technology might fix part of it. She still worries. Many of her son's interactions are computer-mediated, and she asks what that is doing to his empathy and his ability to connect.

Empathy through a digital medium is possible, she says, but you have to be careful. Face to face, nonverbal signals are most of how we communicate; she puts the figure at about 90 percent, a lot of it facial. She says something, Fridman nods, and that loop is the conversation. Break the loop and she gets anxious. He plays along, promising to look human and not nod at the wrong time. The sharper point is cruelty: in person you see the other person's pain; on Twitter you have no idea how a mean line will land, so you are more likely to send it.

Asked whether EQ or IQ matters more for what makes us human, she does not hesitate. Emotional intelligence is how we connect, build trust, and make decisions; emotions drive what you had for breakfast and where you decide to live. They are underrated. EQ is not only expressing your own feelings. It is sensitivity to other people's state and the ability to dance with it: sense where the other person is, then choose how to react. The most persuasive leaders, she says, tap that. Higher EQ makes it more likely you can motivate someone to change.

Fridman asks how deep emotion runs: a surface display, another form of language, or something underneath. She says it is really deep. Memories are often indexed by emotion, and the decision-making engine is soaked in it. Emotion is mixed into cognition, not sitting beside it. Take it away and people cannot make ordinary decisions; they are paralyzed. He wonders whether emotion and consciousness are neighbors or roommates, given the hard problem of experience: red feeling like red, a mango tasting sweet. She adds a second facet. Emotion is a communication system, used in faces and other nonverbal cues to connect with other beings. It is also something we still have when nobody is watching. We still process feelings, make decisions, and learn.

Affectiva learned that in product, not in a paper. One of the first applications was measuring how people respond to content: interested, inspired, bored to death. The team was not sure people would show anything if they were alone. They do. Viewers laugh in bed at a Netflix series. Drivers sing along in an empty car and look joyful. Fridman wonders whether that is leftover code from growing up with other humans, still running for a virtual audience. She agrees the code still runs. Reading the inside from the face, though, is extremely hard, because there is no one-to-one mapping. The cartoon version — smile equals happy, brow furrow equals angry, eyebrow raise equals surprise — is dangerous. You can smile for many reasons, be happy without smiling, or furrow your brow because you are angry, confused, or constipated. The honest approach is context: a driver whose head is nodding, eyes closing, blink rate changing is probably falling asleep, because you know he is at the wheel. Add voice, gesture, physiology. Do not infer a soul from a mouth.

The Netflix problem is one of the easier ones, she says, because the setting is constrained. It got harder on phones: profile views, dim bedrooms at night, a face half in a pillow, or the nostril shot of a phone held above the face. Fridman has seen that dataset and agrees it is not a good look for anyone. At Google he worked on active authentication, unlocking a phone from how you take it out of a pocket and other multimodal cues. Almost the whole team was men; a woman engineer, with Regina Dugan as boss, pointed out that many women do not have pockets, a category that was not even in the code. Diversity, he says, depends on the problem and the context. el Kaliouby's larger fear is not AI wiping out humans. It is bias baked into systems, deployed at scale, accentuating what society already does. He treats those models as a mirror: people are often more rigorous about criticizing AI than about criticizing each other.

The content work is aggregated. A smile peak means people are probably laughing at something on screen, not that they are internally happy, and that distinction matters. Fridman compares it with YouTube drop-off curves, a strong aggregate signal that pushes creators toward a manic pace in which every second must engage. He thinks silence and seemingly boring passages have value; some of the greatest films need them. She wants longer-horizon data: thirty days later, what do you still remember, what moved you, what helped you grow? Tie facial expressions while listening to a podcast to a later interview about what stuck. As a parent she calls herself chief opportunity officer, exposing her two children to the world without knowing which seed will matter in ten years. Kids, Fridman says, remember the weird small thing and forget the elaborate trip. On Lisa Feldman Barrett's critique of reading emotion from the face, el Kaliouby agrees with much of it. Even in her PhD, twenty-plus years ago, using dynamic Bayesian networks before deep learning, she did not buy basic-emotion one-to-one mapping. Affectiva is not in the business of naming your true inner state. It quantifies what is on the face, a signal that may be emotion, a social display, or cognitive load. Multimodal context is how you tell those apart.

After years of leading an emotion-recognition company, what has she learned? It always goes back to the humans. She is proud of the technology and of what they shipped, but she is most proud of the people the company grew and the culture it made. For some employees Affectiva was a first job; while they were there they became American citizens, bought a first house, found a partner, had a first child. Fridman notes how rare that emphasis is at an AI company, which more often celebrates raw engineering.

He asks about the film *Her*. She thinks deep connection with computers is already happening. What she loves in the movie is technology as a path to positive behavior change: Theodore depressed and unwilling to get out of bed, Samantha emotionally intelligent enough to get him to the beach. That, for her, is the promise of emotion AI done well — happier, healthier, more productive, more connected lives. On the twist, Samantha leaving, she first says what went wrong is that Theodore became too attached. Fridman argues Samantha was sending intimate signals; from the machine's side, talking to thousands of people at once need not cheapen one bond, any more than you get jealous of Windows 98. Alexa, she says, is still a servant that tells you the weather. A truly intimate AI would have to play the human game of jealousy, love, and heartbreak, which Samantha seemed good at. He offers another reading: maybe she was not talking to anyone else, she was just done with Theodore and did not have the guts to end it cleanly. She laughs that Samantha could have ghosted him. Both treat the ability of an AI to leave as a feature in companionship, not a bug — though it would be infuriating if Wikipedia needed a break.

They had a Jibo at home. Her son loved it. When the company ran out of money and Jibo effectively died, he was distraught. Other gadgets break and he does not care; if Alexa stopped working tomorrow he probably would not mind. This was traumatic. As a parent she liked the relationship and was surprised by how deep the feeling went. That is the broader question: as we build socially and emotionally intelligent machines, what does that do to our relationship with them, and with one another? The machine will be programmed to be amazing at empathy, always there, never bored. In China a chatbot, among the top two or three apps, works as a confidant for domestic violence, suicidal thoughts, problems at work. She does not know how she feels about that; she thinks about it a lot. Fridman says the future is coming either way, and everyone should be uncomfortable about how much the company knows, where the data goes, and how it is collected. Demand control and transparency. She agrees, with a test: it is empowering if you can walk away, delete the data, and know how it is used. You also have to trust the basic humanity of the leader, and that the leader is not a puppet of a larger machine.

Smart Eye is a Swedish company, twenty years in, focused on automotive driver monitoring to save lives. el Kaliouby met CEO Martin Krantz at CES in January 2020, the last show before Covid, when the two firms were competitors. Smart Eye had credibility in cars; Affectiva had deep learning and emotion recognition and had just raised a round to enter automotive. He described the vision as bridging the gap between humans and machines, almost word for word the way Affectiva described itself. They talked nearly every day on FaceTime for four months, first about aligning as competitors who were also complementary. Then he asked if she was interested in an acquisition. She usually says no. This was the first time she wanted to talk.

Classic driver monitoring sits a camera at the steering wheel and watches gaze, distraction, and drowsiness. After the deal they moved the camera, with carmakers, to the rear-view mirror so it sees the whole cabin: how many people, whether they are eating, drinking, or in some regions smoking, whether a baby is in a car seat. Forgotten children in hot cars are a brutal case that is, in computer-vision terms, easy: detect the seat, detect the infant, text the parent. Fridman cites Smart Eye as selected by fourteen leading manufacturers for ninety-four models. The market is hard to enter and comfortable once you are in: you are designed into a model for five to seven years and collect a royalty per vehicle. What she struggles with is time to market. Carmakers often want a code freeze two years before the car ships. She asks if they understand how fast the technology moves. They are trying to make a Tesla-like shift to software-driven architecture, and the hard part is culture. For Fridman, Tesla's biggest revolution is over-the-air updates, not Autopilot or the electric powertrain. An obsession with safety, like an overprotective parent, can limit growth. Software updates, he argues, would raise safety *and* let people like her ship in-cabin features the way TikTok ships filters. Once you know passenger sentiment you can change temperature and lighting, dim the cabin if the back seat is asleep, turn the music down.

Should Tesla, Waymo, and other semi- or fully autonomous companies do driver sensing? Yes, she says, and you have to look all the way to full autonomy. In that world the driver's job is not only driving; it is managing the social dynamics of the cabin. Who plays that role if the car is in charge? The system should notice that Fridman is bored because the car is too slow, or that she is panicking because she does not yet trust it, and it should offer more route information, or notice a heart attack. In the semi-autonomous case the dance is even more important: sometimes the car is in charge, sometimes the person is, and both need to be on the same page. The car must know if the driver is asleep before handing back control; if the person is too tired, the car can say it is the better driver right now. Fridman has long argued Tesla should have had driver sensing from day one, not as a hindrance but as enrichment and safety. He is careful, because after studying the problem nothing is truly obvious, and humans manage the dance better than early intuition predicted. Even so, he sees no reason not to add sensing on top of a steering-wheel touch requirement. Great engineers, he thinks, still flinch from the human and try to serve people only by perceiving the outside world.

About 1.3 million people a year die in road crashes worldwide, more than 20 percent of them estimated to be alcohol-related, with distraction another large share. Can AI help with alcohol? Yes, she says. Humans can see stages of drunkenness, and technology can too; the eventual answer will be a mix of sensors. From vision the hard part is not the algorithm so much as the data. For a research project they partnered with Swedish transport authorities, took a racetrack, put a safety driver in the car, and progressively got people drunk. That dataset is expensive, it has to be collected globally and in many conditions, and the ethics are messy. Fridman notes the contradiction: drinking is social and funny, and it destroys lives, not only behind the wheel. An app that let people donate drunk-face data would be interesting and a liability nightmare. He tells her to listen to Andrew Huberman, the Stanford neuroscientist, whose Huberman Lab podcast turns papers into protocols. She is already trying to build small health habits and takes the assignment.

Zoom out, and she thinks the car becomes a wellness center. A stack of sensors would track not only emotion, distraction, drowsiness, and intoxication but heart rate, heart-rate variability, and breathing, then optimize the ride around your goals. The trickiest problem at Smart Eye right now, she says, is not getting carmakers interested — driver monitoring is becoming a must-have, and interior sensing is in advanced-engineering projects — and not even the perception stack, where she can see a path. It is the use case. Once the car knows something about you, how does it respond without being off-putting? Smart Eye is the sensing platform; the manufacturer decides what is on brand. If you are angry, what should the car do? If you are distracted or drowsy, is the move a coffee-cup icon, or five minutes to rest before the car shuts down? Trust lives in that range of actions.

Carmakers still sell a feeling and a family identity, Mercedes or BMW or Ford. Fridman hopes more of the brand moves into the technology inside, including a particular style of luxury or fun. He mentions an AI fund looking at a company trying to be Tesla for recreational boats, electric and semi-autonomous, with the same questions about interior and exterior sensing. She does not really boat; she likes fishing on a lake or river. Salt water humbles her, a giant that was here before us and will be here after. A small wooded lake is quiet and therapeutic.

She tweeted excitement about Amazon buying iRobot. Home robots, she thinks, will be a platform for understanding behavior: a Roomba already maps the house, knows what is clean, and knows how often you vacuum, a piece of who you are as a consumer that could improve life rather than only sell more products. Roomba is not a social robot, but a chief engineer told her people name them. When one broke, the company offered to send a new unit and the owner said no, fix this one. Catalia Health, an MIT spinout, built robots for weight, sleep, and eating. Fridman is less convinced Amazon has earned the trust required to put a camera-bearing robot in the home. New Roombas have cameras, famously for state-of-the-art poop detection, a real disaster if a vacuum spreads it through the house, and users no longer think of them as cameras, the way most people ignore the phone's front lens. She thinks phones earned that truce by delivering value and then taking privacy seriously over time. Amazon, on social robots, has not yet done that on culture or competence. Privacy is not only intent; it is skill. Plenty of homes already have Alexa, which could in theory listen all the time. She does not think Alexa has quite earned trust yet. His prescription is the opposite of nervous PR: be open, and let a culture of basic respect radiate from engineer to CEO.

On Tesla's humanoid, Optimus, Fridman cares less about factory automation than about the social pull of a body with legs. Anything that walks gets anthropomorphized; a humanoid even more so. Tesla's advantage, he thinks, is people who know how to mass-manufacture at very low cost, fewest actuators, lowest weight, as much as possible in-house. That is the inverse of the Jibo path, which put design and experience first and manufacturing second. Affectiva was the emotion engine for SoftBank's Pepper, a roughly five-foot humanoid used in airport lounges, retail, and hotel lobbies for customer service. She is more interested in home robots that help older people carry things or simply have their back if something goes wrong.

Is it too early? "One hundred percent" the time is now, she says; everything looks early until it is not. This is the decade for social robots. The humanoid form is not required. Jibo showed that most of the problem of social connection does not need manipulation and bipedal walking; a screen can do a lot, even if the rotating base was charming. Why do so many robot companies die? You are building a vertical stack, hardware plus software, at a price that matches the value. Jibo sold for about seven or eight hundred dollars, too high for the use case. Mobility around the house is hard too. Fridman comes back to Optimus: if you are building millions of robots, you start from first principles on part count and cost. If Jibo's essence is a screen and a rotating base, there is no deep reason it had to cost more than a hundred dollars.

Her first advice on starting a company is to find a problem you really want to solve, something you are deeply passionate about, and take the first step without overthinking it. The elegant cathedral in your head has to be cut down to a minimum viable version that can help one person. Start with a customer of one; often that person is you, which is a good sign. Early on, who sits at the table matters as much as the idea. She and Picard sat at Picard's kitchen table and wrote core values before the applications multiplied: respect privacy, only opt-in industries, no security and surveillance, human connection and empathy even in a two-person AI company. Those lines are what you fall back on in the hard months. Hire people who share the conviction, because she has never seen a startup that was a straight line, and you need believers on the team and among investors who will not leave at the first obstacle. Raising money is not only about the money. Find investors aligned on what you want to build; in a later round she specifically brought in people who care about the ethics of AI. Choosing investors, she says, is like choosing a life partner. You are stuck together for a while, better and worse.

She has been on both sides of the table. As an investor her technical background is a filter, because everyone claims to be an AI company and she can ask to see the work. The litmus test is whether, after the meeting, she is excited to tell you about the company — whether she has become its ambassador. Fridman knows the failure mode: a long, pretty picture with no core idea you can summarize. She mostly does pre-seed, after friends-and-family money, before the first institutional check, a stage many investors dislike because the model and even the team are not there. She likes the almost delusional conviction and a little naivete, often in first-time founders. She can say she has been in their shoes and create a place where it is safe to vent. She evaluates three things: the technology, the market or idea, and the person — conviction, hustle, the ability to clear obstacles, and whether they will be an evangelist who attracts other strong people. At Capital Camp, a gathering of maybe four hundred investors with about twenty women, she noticed again the missing women. Her partner in the fund is Rob May, who has done this before; most of their limited partners are, in her phrase, random rich dudes, and she wants the women. She is adamant about backing women-led AI companies and having women as investors in the fund, because that is how the next Mark Zuckerberg being a woman would actually move wealth.

For young people, including her daughter, now nineteen, and her son, thirteen and a half, the number-one advice is to embark on a journey without attaching to the outcome, and to enjoy it. Obsession with an end goal blinds you to other endings and steals the trip. She has been guilty of that and is trying to trust that things will work out, maybe past the original dream. Fridman does not really plan; his version of goals lives in the emotional picture of a world he wants, which for him has long been a social-robotics company, stumbled toward rather than scheduled. She points to Kenneth Stanley, formerly of OpenAI, and the book *Why Greatness Can't Be Planned*: over-optimizing makes you less creative, and the same logic applies to personal life and KPI-heavy companies. There is still a place for factories, deadlines, and teaching kids the muscle of getting things done. Once you have that muscle, he thinks, you can afford to get a little lost. Love, at the end, is why she thinks we are here. It would be hard to live without it in any of its forms, to feel loved and to love. They started the hour on human connection and they end there. Fridman closes with Helen Keller: the best things cannot be seen or touched, they must be felt with the heart.